Last Autumn I taught a course on Online Marketing & Analytics at the University of Washington MBA program. As part of the class I assigned three books. While there was some complaining  about the amount of reading (“Do we have to read the ENTIRE BOOK?” I was asked more than once), overall the books were very well received. At the end of class a few students asked if I had other book recommendations.

This post will be the first in a series on some of my favorite data-driven books. Most are not specifically marketing books, but I learned a lot from all of them.

To start with the three books I required for the course:

How Brand Grow, Byron Sharp

I reviewed this book last year and I recommend reading that earlier (more thorough) summary. The basic point of the book is that most of marketing ‘best practice’ is wrong, but by looking at the data you can find some things that actually work. It is a summary of what we actually KNOW with respect to marketing vs opinions.

Some of the things we KNOW:

1. Brands get big by growing penetration, but by increasing share of wallet
2. Niche brands exist, but brands you consider niche, aren’t
3. Within a category every brad shares the same customers (Mountain Dew and Diet Coke are drunk by the same type of people)
4. Differentiation of product is generally not a good idea.  But you need to differentiate your brand (i.e., unique name, color, tag line, jingle, etc.) so that you are memorable and recallable
5. Loyalty programs are generally a bad idea
6. Customer acquisition is far more important than customer retention

That’s just the start. As I told my class: Read the entire book.

Everything is Obvious (Once you know the Answer), Duncan Watts

By now, if you read non-fiction at all you will have read the Tipping Point. Now you need to read Duncan Watts. Duncan Watts is the science side of the Tipping Point (vs. Malcolm the storyteller). The difference is that Malcolm tells a great story (the best) and Duncan tells you the truth. Watt’s earlier books were interesting, but nowhere near as entertaining at Gladwell’s. This book though is fantastic. It’s a fast, entertaining read and it will put the record straight on the whole idea of mavens and connectors.

Watts is both an academic expert on network theory, and an actual practitioner (he worked for both Yahoo and Google). The title of the book refers to the fact that when we look back at the past we are very good at explaining everything that happened as stories. That causes two problems:

1. It gives us confidence that we can do the same into the future. And evidence suggests we can’t
2. It makes us think that when we see two things one after the other we can assume that A causes B. It often doesn’t

In the world of experiments you can prove this (I’ve long lost count of the number of times I’ve been surprised by the result of an A/B test). In the real world you can’t. So instead we use the characteristics of successful things and use those characteristics as explanations for why that thing was successful.

Some examples:

• Harry Potter was successful because it had a young protagonist who lived in a world that was similar to our own, but more exiting. He was an outsider with a destiny. He was up against impossible odds. But he had a core group of friends that any kid could relate to. It mixed high adventure with the challenges of coming of age. In other words: Harry Potter was successful because it was like Harry Potter
•  Michael Jackson was successful because he started young with a lot of support. He understood the music business from the inside with his entire family. But he was an outsider who needed to breakout on his own. He came along at a time when we were ready for a King of Pop. In other words, Michael Jackson was successful because he shared the characteristics of Michael Jackson

Watts drives home this point with two very compelling stories. The first is about the Mona Lisa which is we are told is the most famous painting because it has the characteristics of Mona Lisa – but there is a twist. It turns out the Mona Lisa was not famous for a long long time. It wasn’t until about 100 years ago that the Mona Lisa was stolen and its fame blew up. It was the theft that made it famous. Now that we have forgotten the origin of its fame, we attribute its prestige the same way we do everything else: by describing it and defining success as that description.

The second story is a music experiment.

Watts used an early-days social network and divided into separate test groups. Each group was given access to the same alternative music. In every group except one there was a real top-10 list of the music that was listened to the most. If you believe that the music at the top of the charts is there because of intrinsic characteristics, then each Top-10 list should be pretty similar.

They weren’t.

Every list had a different #1 hit. If a song was a #1 hit in one world, it tended not to be at the bottom of the chart in the other worlds, so there was some correlation, but that correlation was very small. Basically if we were to rewind the world to 1975 and run it forward again it is unlikely Michael Jackson and Madonna would be superstars a second time.

(Which is a very hard thing to get your head around).

Read the entire book.

Zero to One, Peter Thiel

This is the most recent and most popular book on the list. A lot has been written about Zero to One in the past six months. I don’t think it is a valuable use of my time to re-tread over this ground. But I will share a one of my take-aways that I haven’t seen mentioned very often.

Thiel talks about ideal price points for new products. He says there are four:

$1 products can spread virally. Your marketing plan is PR, SEO and social

$100 products can be marketed with paid online marketing. They work too.

$10,000 products are sold by professional sales people. You need to actually talk to someone at this price, and at $10,000 you can afford the commission of a good sales person.

$1M products are sold by the founder and CEO.

There are two big gaps:

$10 products will not spread virally, and you don’t have enough margin to do paid marketing. They tend to fail.

$1000 products don’t sell without a salesperson (like $10,000 products), but they don’t have enough margin to afford a good salesperson, so they tend to fail.

He talks a fair amount in the book about $1000 products and how they often involve selling to small businesses (and failing). He spends less time on the $10 product, but it is an equally awkward price point.

I wish I had read his book before starting my start-up.

We created a product that connected consumers to restaurants. We filled restaurant  seats when they would otherwise be empty at a deep discount for the customers. It was a great product with better benefits for both sides of the arrangement than any other product on the market (much better than Groupon, Restautant.com, Coupon Books,  Yelp ads, etc).

The issue?

We charged the consumer $10 to get 30-50% off their meal.

We had to hire salespeople to signed up single-location restaurants.

Effectively we went after the $10 and the $1000 price point at the same time…

More book recommendations next week (or when I’m next inspired…). If you have favorite books you can recommend please list them in the comments.

In the early days on facebook (actually not-so-early, but still many years ago) they opened up the ability to claim domain names for pages. For example, my facebook url is facebook.com/nevraumont – I got my last name because I claimed it first. A buddy and I also jumped at the chance to claim a whole bunch of other facebook urls.

If you had a “page” and that page had at least 20 “likes” you could claim any url you wanted. It was like the early days of the internet. A buddy and I got to work. We build pages, added ~5 links to them with relevant topics, bought facebook ads to get 20 likes, and then claimed urls. We have hundreds of them, about a dozen are really good.

One of my favorites is facebook.com/travel.

Over the years the two of us have tried to build side-businesses on top of the urls. It’s a lot harder than it looks. I am more than confident it’s possible if one were to work on it full time. Both of us have way too high opportunity costs to do that. If you (or someone you know) is interested in building a business on top of a premium facebook domain, let me know. We have names in the travel, weddings, automotive, education, financial and legal spaces (among others). Basically the categories that were making money though Google search at the time.

But I digress.

One of the businesses we build was StopoverTravel.com. The idea was to create a repository of tour companies, and then link to them for affiliate revenue. For fun I started writing content on the site. Once a week I would write about a cool thing to do in the world. The site still exists. If you are looking for travel inspiration, check it out. While this was happening, Google launched Google Surveys. It’s a very inexpensive way to survey people on the internet. I had some free credits so I gave it a try. My goal was to get results that could conceivably go viral (I will talk about why that is basically a terrible idea in another post).

This is that story.

What Google Doesn’t Like
The headline I wanted to run was something like: “People have a lot of sex on small group tours” I wanted to get data on how often it happened, how long the hook-ups lasted, and for even more fun: who it happened with.

My first survey design was set up like this:

Screening question: Have you ever gone on a small group tour? Anyone who said “No” would be eliminated and the next questions would only be asked to people who said “Yes” First question: “While on a small group tour, did you ever have sex with someone you met on tour?” The possible answers were things like, “No”, “Yes, with a fellow traveler”, “Yes, with a local”, “Yes, with the tour guide”

I thought it would be great.

Virality here I come!

I submitted.

Then I got this email from Google:

Thank you for using Consumer Surveys. However, your survey has not yet begun running.

We do not allow surveys to run with your submitted content per the Nudity, Obscenity, and other Adult Material.

We don’t allow surveys that contain nudity, obscenity or sexually suggestive material.  Surveys should not relate to porn, dating with a sexual or mature nature or sexual aids & devices.

Please remove all references to adult material and re-submit.

Oops. Time for a new word for “sex”. I asked my partner. Here was his list:

• “Hook-up”
• “Casual relationship”
• “Physical relationship”
• “One night stand”
• “Make-out”
• “Go all the way”
• “Score with”
• “Become intimate with”
• “Biblically know someone”

(My partner is great)

I re-submitted the survey with the word “sex” replaced by the word “Hook-up” (in quotes).

I hit submit.

A day later I received another email from Google:

 Thank you for using Consumer Surveys. However, your survey has not yet begun running.

We don’t allow surveys that contain nudity, obscenity or sexually suggestive material.  Surveys should not relate to porn, dating with a sexual or mature nature or sexual aids & devices.

Please remove all references to adult material and re-submit.

Apparently no “hook-ups”. I tried again. This time replacing the offending language with “Did you ever begin a relationship (even if very short term)?” Google wrote me back:

Thank you for using Consumer Surveys. However, your survey has not yet begun running.

We do not allow surveys to run with your submitted content per the Nudity, Obscenity, and other Adult Material.  The second question of your survey has innuendo of dating with a sexual/mature nature.

Please remove the part from your second question that reads “(even if very short term)”.

This question could work if it were about non-mature content.  For example,

“While on a small group tour have you ever begun a relationship?”

Answer Choices: Yes – I met my future Spouse Yes – I made lifelong friends No

Answers that point to specifically who in the group (tour guide, locals, etc) and the language about “even if very short term” make the current question seem like it is about mature content.

Still a no. But at least they were being helpful now. I was speaking to a real human being who was making judgement calls. I made some more changes and wrote him back:

Thanks,

Oops. I think I pushed too far. I took a different tactic:

Hi,

Radio silence. I wrote back again:

I recently wrote an article for MediaPost’s Boomer section. The ask was for an article on how Boomers are using the internet. I shifted the focus a little (as I tend to do) to talk about how many companies focused on seniors are just now realizing how important the internet is (shocking I know…). Unfortunately this is happening just as Boomers are shifting from the internet to mobile.

Seniors (and Boomers) have tended to be late adopters. That’s helpful for marketers as we can learn from other industries and we don’t need to be overall innovative (just innovative within the senior space). The issue is that it seems those marketing to seniors also seem to be late adopters.

There are companies trying to do cutting edge stuff in the senior space. The issue is those folks are TOO ahead of where their market is. It’s a fine line.

I call it “fighting the last war”

Here is the full article:

http://www.mediapost.com/publications/article/241352/are-you-fighting-the-last-war-in-your-boomer-mar.html

My first real specialty in marketing was customer lifecycle management (CLM). That’s a fancy term for caring about the total profit you generate from a specific customer. That can mean everything from channel management (figuring out which sources of customers are more valuable for you than others) to cross-selling and upselling; From customer acquisition to customer retention. It was a pretty broad way to look at marketing – and a very analytic way. It was a nice base for the rest of my career, and I highly recommend it to students looking for an initial foray into marketing.

One sub-specialty within CLM that I spent a lot of time with was Loyalty Programs. Loyalty Programs have become very popular over the years. A big reason is that they are so visible. When a company has an excellent CRM system or distribution system or save desk it is basically invisible to the public and their competitors. But a Loyalty Program is obvious to everyone. When a CEO sees a competitor (or even a company in a different industry) with a Loyalty Program they will often ask their CMO: “Why don’t we have a Loyalty Program? I’ve heard that Loyalty is more important than customer acquisition, so shouldn’t we have one of those?”

When a CEO asks for something that involves spending more money on marketing, most CMOs say yes.

The result is more companies with Loyalty Programs – which causes more CEOs to ask, “Why don’t we have one of those?” It’s a flywheel.

A trend like that can get something started, but most businesses are pretty good at shutting things down if they aren’t working. Somehow Loyalty Programs have stuck around, and if anything become more prevalent (especially in travel and retail). And yet I argue that most of the time they are a bad idea (in retail – travel Loyalty Programs are generally done right). The reason is Selection Effect.

First: What do I mean by a Loyalty Program?

“Programs” that drive loyalty could mean many many things. Amazon Prime drives loyalty, but in general it is not considered a Loyalty Program (but maybe it should be). In general people mean one of two things when they call something a Loyalty Program:

1. A program that gives you points when you spend at a company. Those points sit in your account and, after they accumulate, you can redeem them for product or merchandise  (usually the same company’s products, but often other companies as well)
2. A tier-based program. Tier programs give customers defined benefits after they have performed specific activities in a set time period (usually something like a specific number of stays, number of miles traveled or dollars spent)

I will use a separate post to talk about the second type of Loyalty Program. For now let’s stick with the classic “earn and burn” points-based program.

There are definitely different varieties of points-based Loyalty Programs, some of which at first glance don’t look like point programs at all. The most common of these is the “Frequent Visit” stamp card. You’ve definitely seen these: Buy ten coffees, and your next one is on us. Another example (if you are as old as I am) were those Subway stamp “passports”. These examples are very different from the more a standard Marriott Rewards Points, but the principle is the same: Spend now to earn points (Marriot Points/”Coffee Points”), then later spend those points for stuff (Hotel stays, free coffee). The only difference is how many points you earn when you spend and how much those points are worth when you spend them.

What’s the value of a Loyalty Program?

It’s definitely not to drive Loyalty. Point based Loyalty Programs are effectively complicated discount programs. In exchange for being tracked (sometimes companies don’t even ask for that), a customer gets points today, that can be used for discounts tomorrow. The difference between paying $0.90 for a coffee and paying $1 for a coffee, with a 10-stamp loyalty card is almost the same – or for that matter a $1 coffee that earns you 10 points, with each point being worth 1-cent on future coffee purchases. Here are the differences:

1. Breakage: Some percentage of people will not use the points you give them. Those un-used points are called “Breakage”. Depending on the design of the program breakage rates can vary from 0% to 80%. In the above example, if your program has 33% breakage (pretty common), you could give you 30-points on that $1 coffee purchase and it would be the cost to you as offering the coffee for$0.90.
2. Psychology: Generally people value things that aren’t cash at a lower rate than cash (Duh). People don’t buy gift cards for more than their face value – especially for their own use. But sometimes if a program is designed right and you pull the right strings you can get people to act on points when you would not be able to on dollar discounts. Usually this isn’t done with the base program, but rather with promotions. “2x points” sounds a lot better than “An extra 1% off”, and you would not be surprised to know it gives a much bigger sales lift.
3. Selection Effect: Only some people will join your Loyalty Program. Most people will just not care enough. Who joins? You should not be surprised to know it is to some extent price sensitive people, but mostly it is people who are your highest use customers.

Apart from those subtle differences (that are all worth exploring as I wrote more about Loyalty Programs on this blog), basically Loyalty Programs are price discounts. And price discounts do not generally drive loyalty.

So if the value of the Loyalty Program isn’t loyalty, what is it? There are a number of uses, but the most common one is to collect data on your customers. Without something like a loyalty program, Marriott would only know people per visit. They wouldn’t know who their best customers were in order to treat them differently.

But that is NOT the reason most companies build Loyalty Programs. Most build them because they think they will drive Loyalty (and because everyone else is doing it).

Why do companies think Loyalty Programs are Driving Loyalty when they are not?

Most senior executives aren’t stupid. Why would they think that their Loyalty Program is driving Loyalty when it really isn’t? The answer is the Selection Effect.

If you take all your customers and break them into two groups: Those on your Loyalty Program and those not on your Loyalty Program, it will be very very clear that the customers on your Loyalty Program have (1) Longer Tenure with your company, (2) Spend more, (3) Churn at a lower rate.

Wow.

That sure sounds like the Loyalty Program drives Loyalty.

But it’s a classic case of correlation not being causation. In this case the impact is the reverse:

Customers who are planning to spend more with your company in the future (and usually have spent more in the past) are the most likely to join your program. Why? Because they have the most to gain.

There is a cost to joining a program. The cost usually isn’t very high, but it does usually mean signing-up with some forms and keeping a card in your wallet. But that small barrier to entry is enough that many customers will not join your program – specifically customers who aren’t planning on spending very much with you in the future. Meanwhile the heavy users would be crazy not to sign up and get 1% or more off all of their purchases.

So when you compare Loyalty members to non-members, the loyalty members always look more loyal. But the kicker is it’s not the program that is making them loyal, it’s their loyalty which is getting them in the program.

Most companies stop right here by the way. The head of the Loyalty Program shows the bar chart showing members spend more than non-members and the CEO is happy and they keep spending money on the program.

Some companies try to get more sophisticated. The first thing analysis will do is compare Loyalty Members before and after they join the program. They eliminate all the members that were not purchasing for at least 1 year prior to joining, and then look at their spend before and after joining. The really good analysts will try and create a control group of similar customers who did not join the program. It’s commendable to try, but this has a similar, if slightly more complicated problem.

Imagine two customers that are identical. They have both been shopping at your store regularly for a year. Then one day they both come into the store. Customer A is planning to move to Phoenix in the next few weeks. Customer B actually lives across town and was only visiting when he visited his mom who lives nearby. As luck would have it, Customer B just bought s new home near his mom, so he is planning to visit the store more often in the future.

Guess which customer signs-up for your Loyalty Program that day?

And guess how much impact the program itself has on his increased spending?

It’s another example of selection effect. You can’t get around it.

Getting Around Selection Effect

There actually is one way to get around selection effect. It’s called A/B testing. Take a group of customers and randomly assign them to two groups. Then do something to Group A but not to Group B. Watch and see what the long term impact is across the two groups.

Because they were randomly assigned there was no Selection Effect.

The problem is that this only works if you do things “Below the Line”. As soon as something is public and everyone can see it, you can’t keep Groups A and B separate anymore.

This makes A/B testing great for website landing pages or email campaigns or direct mail campaigns or even television campaigns (if done regionally). But if you want to create a company-wide Loyalty Program, it’s really hard to only offer it to 50% of your customers.

But it can be done.

The best time to do it is when you are first planning to launch the program. Instead of launching it fully-formed to the public, you can try launching it in ‘beta’ as a below-the-line test to a select group of your customers. And instead of offering it to your best customers, offer it randomly. Take your entire customer database and divide it in two. Email half of them with an offer to join the program and don’t send anything to the other half.

Then, instead of measuring the lift of people who joined the program vs those who did not, just measure the lift of the group that was OFFERED the program vs the group that wasn’t. If you did your job right in randomized group selection any difference you see will be driven by the existence of the program. You may even see a lift in the people who you offered the program to, but they didn’t join (this is the best case – increased sales without any increased cost).

Any time you create any program without doing a proper A/B test you are at risk of having your results invalidated by the Selection Effect. Loyalty Programs are the biggest example, but I’ve seen it with everything from add-on services (Customers who buy the security package have lower churn, so let’s give away the security package for free), to customer service (customers who visit the branch more often are less likely to churn, so let’s drive more customers into the branch).

Beware the Selection Effect.

I am writing this on December 26th. In much of the Commonwealth today is the biggest shopping day of the year. December 26th is Boxing Day, which was originally a day to give gifts to the poor (and your servants), but has long since morphed into away for merchants to unload excess inventory from Christmas at deep discounts.

As far as pre-planned deep discount days go, Boxing Day makes a lot of sense. A lot more sense than other days like Black Friday or Cyber Monday. The reason has to do with Cannibals.

As I talked about a little in my Attribution post, any sales you attribute to a marketing activity are either over- or under-stated. If you under-state the impact (Best example: Television Advertising) we use the term “Synergy”. The spend is helping other marketing activities in a way you aren’t giving it credit for. If you over-state the impact you are claiming the activity (or channel) is doing something, when it is really just stealing credit from a different activity you are ignoring.

Getting attribution perfectly correct is impossible. Smart companies do the best that can, then they guess whether what’s left is synergistic or cannibalistic and use that to decide roughly if they should spend more or less on those activities.

This doesn’t mean never spending on cannibalistic channels or throwing money at synergistic ones. If you have a channel that you measure as creating $1 in marginal profit for each $0.10 you spend, but you are pretty sure it is very cannibalistic, you are likely okay. Unless the channel is 90% of more cannibalistic, you are still making money and likely shouldn’t stop.

Which brings us back to discounting.

Discounts by their very nature are cannibalistic.

Let’s use a very simple example:

Imagine a world where there are 100 people and they all buy toothpaste every week. Everyone buys one tube from their normal brand for the normal price, use it up over the week and then repeat. Lets say there are two brands: Crest and Colgate and they each have 50% of the market. They each sell their tubes for $1.

Crest decides they want to run a discount to take share from Colgate. They discount their price by 10%. Crest now costs $0.90 per tube.

The first thing that happens is that the 50 people who were buying Crest keep buying Crest. Only now they are paying $0.10 less than they were before. Crest has immediately lost $0.10 x 50 = $5 in revenue. Even worse, they have lost $5 in profit, since there was no costs associated with the price being $0.10 higher.

If Crest were able to produce toothpaste for free (i.e., no marginal costs of production – obviously unrealistic for something like toothpaste), they would need to sell $5 more worth of toothpaste to break even. In this example that would mean 6 additional tubes (6 x 0.9 = $5.40). If their cost of production and distribution was $0.50 per tube, they would need to sell 13 tubes (13 x 0.4 = $5.20). They would need to move their market share from 50% to 63%. And that wouldn’t get them ahead – it would just get them to break even.

And it doesn’t end there.

They get hit two other ways.

The first is Colgate’s reaction. Imagine if Crest did this and you are Colgate. Are you just going to sit there while Crest drops your market share to 37%? No way. You offer your own price discount. And when you do you steal share right back from Crest.

Either we get to a new equilibrium where both tubes are $0.90, both brands are back to 50% share and everyone is making less money, or they go back and forth offering discounts with their shares jumping around from week to week. In either scenario Crest is in a lot worse place than they were before. (This would be solved by collusion between Crest and Colgate – which is why it is illegal, but still happens from time to time)

The second way they get hit has nothing to do with Colgate. Some of the loyal Crest buyers may ‘buy forward’. Instead of buying one tube when it’s on discount, they buy two (or ten!). They don’t increase their consumption, they just move it. So instead of one week of cannibalization, you could end up with a half dozen weeks. Those pre-buyers just don’t buy the next week (or the next ten weeks).

If ten percent of your loyal buyers decide to buy ten weeks in advance, you lose another $0.10 x 10% x 10 = $0.10 per buyer (on average). That means you need another 13% increase in market share just to break even. Do you really think your 10% discount will move your market share from 50% to 76%, and not cause your competitor to react. Even if you do, that just gets you to break even. We haven’t even talked about supply chain costs of causing a spike in demand yet.

Discounting is ugly.

Should you ever discount?

Yes. Let’s walk through cases where discounting makes sense.

Zero Marginal Costs

If your product costs you nothing to make, discounting isn’t so bad. In the above example, if Crest was free to make (“Virtual Crest”), you only need half the market share gain to hit your break even. This is the case with a lot of online services.

Trial

Want to try Netflix? You can sign-up and get your first month free. Netflix wants you to try their product. They know that some percent of the people who try it will stick around. Since their marginal costs are pretty close to zero, it doesn’t cost them very much to let you try before you buy. To a lesser extent this can work with any product. Maybe Crest wants Colgate buyers to try Crest just once. They are so convinced of the superiority of their product they think most of those people will stick around. If those 13% keep buying Crest (and don’t even switch back to Colgate when Colgate discounts to $0.90 the next week), they will have achieved a big win. Unfortunately most products just aren’t that much better.

Low Market Share

If you have no base business to cannibalize, discounting can make a lot of sense – especially combined with ‘trial’. If Crest only had 1% market share to Colgate’s 99%, then discounting to $0.90 to gain 13pp of share sounds like a fantastic deal. They only need to get their share up to 1.06% to break even.

Expandable Categories

People generally only use so much toothpaste, but that isn’t true for chocolate. If you have a 50% share in the chocolate category and you get people to buy a bunch more of your chocolate in a given week, even if they ‘buy forward’ it is unlikely to hurt your sales significantly in future weeks. People will eat what they buy. The issue happens when you start convincing yourself your non-expandable category is expandable. When I was at P&G we told ourselves (and had market research data to back us up) that if people bought more toilet paper, they would use more toilet paper (I will leave it to your imagination how that could happen). I believe that is likely true on the margin (“Oh. I’m running low, guess I better use less squares per wipe until we get to the grocery store…”), in no normal world is toilet paper an expandable category. But it makes sales people a lot more comfortable when they run deep discounts.

Excess Inventory

If you have product that is going to be thrown away, you don’t have a cannibalization problem. In these situations you want customers to forward buy their toothpaste. If they don’t you are going to throw it away. This is really an extreme example of zero marginal costs. In fact it’s negative marginal costs, since you will likely have some costs to bear on the disposal. The best example of this is the fashion industry. Whatever you don’t sell at the end of a season is worthless for (or worse), which is why you see big blowout sales of post-season clothing. The last 15 years have also seen the growth of a fashion-industry business model that exists with just-in-time inventory so they don’t have the excess inventory at the end of the season. This lets them charge lower prices for their non-discounted product during the season and steal share.

This is also the reason why, in general, Boxing Day sales make more sense than Black Friday sales.

Discounts and Prejudice

In my experience many business people have in-going prejudice on whether discounts are good or bad. Some people are convinced discounts drive their business. Others think that discounts dilute their brand image and just degrade their margin without any upside. The truth is more subtle than the two extremes.

I did a lot of work in telecom trying to reduce customer churn (including building those save desks that everyone hates). We would often do this with discounts. We would get a lot of push back from telecom executives that thought we were just throwing money away. But let’s look at the math:

Let’s say you offer a 50% discount for a month to a customer. Let’s say you can target that discount so it hits people that are 50% likely to churn if you don’t do something. Let’s say that gets 50% of those customers to stick around for another month. Let’s also assume that if you get them to stick around past this ‘churn’ period their future churn rate is much higher than your average (say 4x higher), but definitely not 100%. Let’s say these customers are spending $50/month on your service and the average customer lasts 4 years. We’ll say there are 100 customers like this for calculation purposes.

If you do nothing: 50% of your 100 customers go away. You are down 50 x $50 x 12 months x 4 years = -$120,000

If you discount 50% to all 100 customers you end up with:

50 people discounted that you didn’t need to. This cost you 50 x $50 x 50% = -$1250

25 of the 50 churners you lose anyway = -$60,000

The other 25 you discount for one month (25 x $50 x 50% = $625). This keeps them around, but they still churn in a year (instead of your 4 year average), so you are still down 25 x $50 x 12 months x 3 years = -$45,000

So with your discount program you are down $106,875. But that compares favorably to the $120,000 you would be down without doing anything.

Obviously all these numbers are made up, but they are all fairly reasonable.

Discounting is Easy?

The first step in winning in the discount war is understanding cannibalization. Then you need to understand why you are discounting. If you are doing it just because everyone else is, you are likely making some pretty big mistakes. But if you take a few minutes to think about why you are doing it and determine if it fits into the five reasons I give above for discounting, then do some basic theoretical math on what you you would have to believe for it to make sense, you can start making some much smarter judgement on whether a specific discount is a good idea.

Most of the time the answer should be fairly obvious. Like most things, if the answer is really hazy and unclear, it might be ROI positive, but there are likely better things for you to do with your time and effort. It should be obvious discounting is a good idea, otherwise go and put your effort somewhere else.

Disagree? Know other reasons why discounting is a good idea? Comment below.

Last week (November 19^th) an article was published in Geekwire that got my blood boiling. The article in question was written by Jeff Reifman. Jeff was a program manager at Microsoft and is currently an independent technical consultant living in Seattle. Earlier this year he wrote a (very good) post on his personal blog about how the growth of Amazon is so significant that it has measurably influenced the gender ratio of the single people in Seattle. The post went viral and Jeff made a local name for himself as a writer.

This, I assume, gave him the platform which allowed his latest piece to be published in Geekwire. Where his first piece was an interesting data-driven look at the impact of a single company on the demographics of a city, this new post is just a diatribe. The internet is full of ignorant venting, so why did this particular post bother me so much?

1.  It’s getting a lot of airplay – so it’s ignorance with an audience
2. It’s coming from someone who previously produced a nice piece of data-writing, which is disappointing
3. Its latching onto a two trends I’m seeing more and more recently: “Big Company = Evil” and “Change is bad”. Both of those memes have been around for a long time and will never go away, but it’s disheartening when the tech media starts stroking that fire rather than educating
4. In the three days after his post I heard slightly different versions of the same argument from two very different sources. One was from a friend’s facebook feed complaining about similar problems in Vancouver. The other was from a Lyft driver who was lamenting the changes in South Lake Union.

For those four reasons I’m putting aside my standard data-driven marketing blog post this week for a monstrously long (for me), 8000 word rebuttal to Jeff’s piece. This article started out as a point-by-point rebuttal of Jeff’s article (and still ends that way), but it has expanded in its final form as a sort of manifesto against the idea that growth, change and improvement are bad (because change leaves people behind). Change has been leaving people behind long before automobiles replaced the carriage. The point that no one cries for all the unemployed blacksmiths these days has been made ad nauseum, and I won’t spend any more time on it than I already have. Instead I will use Jeff’s core arguments as a structure to explain why I believe Amazon is doing a ton of good for the city of Seattle (and the world for that matter).

On with the essay.

Jeff’s Arguements

Jeff starts with the fact that Amazon’s office space, if fully utilized will cause it to employ 45,000 people, or 7% of Seattle’s workforce. He then states that this will cause (or ‘fuel’), “an unaffordable traffic-filled metropolis dominated by white males and devoid of independent culture.”

The rest of his article tries to explain why this will happen, why it is Amazon’s fault, and why this is bad. His arguments are internally contradictory and based on hyperbole. Before I dive into tackling these arguments one-by-one, I’d like to start with taking on the complaints themselves:

• Unaffordable
• Traffic
• White Males
• No culture

All of those words are very charged. Obviously no one wants a place that they can’t afford. No one likes traffic. Unless you are in the KKK, white males are a sign of non-diversity (which is bad). And if you had to choose between culture and no culture, I’ll bet north of 90% will take “Culture please!”

But before we assume that all those things are bad, it might make sense to examine them a little critically (maybe with different ‘non-charged’ words).

• Unaffordable

An unaffordable city means that people can’t afford the rent or real estate prices. Is this possible in a free market economy? Of course not. If no one can afford rental property, the owners will reduce the price so it doesn’t sit empty. If no one can afford to buy a house, house prices will come down until someone can (or the people that own it decide to sit on it and not sell).

People can obviously afford rent and housing prices in Seattle. In fact, I’ve heard from my friends looking to buy that houses priced under $1M (most of them) are often turning into bidding wars and they have to make offers on the spot if they have any hope of getting the place.

So people can afford to live in Seattle. In fact given the population increase and the utilization of housing, MORE people can afford to live in Seattle than ever before.

But that’s obviously not what he means. He means that prices are higher than they were before. And that prices are so high that people who are lower on the income scale can’t afford the housing costs in the city.

This is bad, right? Gentrification driving out people who have been renting for years and now can’t afford the new prices that the new Amazon employees are able to pay.

What is the alternative?

Easy. Prices do not increase.

Well, we know what happens when prices go down. Investment banks collapse and the government needs to bail out the automotive industry. Pretty sure that’s not good.

It’s not always true that the opposite of bad is good, but I think in this case it is.

Why do prices go up in a market economy?

Two reasons:

1. Demand has gone up
2. Supply has gone down

Since supply isn’t dropping (A quick look at the Seattle skyline tells you it is going up), demand must be going up faster than supply in increasing. What causes demand to go up?

1.  Existing ‘customers’ are willing to pay more for the product
   1. Due to a higher quality product; or
   2. Due to more disposable spending (i.e., they have more money and choose to spend it on this thing they like)
2. New ‘customers ‘ wanting to buy the product

1(1) seems like a pretty good thing. If the city were to make the city better without raising taxes then hopefully we would all be in favor of it.

In practice this isn’t always true. When I was living in Center City in Philadelphia a developer was looking to build a move theater nearby (there were no movie theaters in the neighborhood). Even if you don’t love movies, a movie theater is still better than an empty building (which it would be replacing). And yet the street poles were covered with pamphlets protesting the development. The argument was that if the neighborhood got a theater, then it would be a better place, which would allow landlords to increase rents, which would make it less affordable for the people living there.

Basically their argument was: “Don’t make the city better.”

If there is interest I can expand on that in another post, but for now I will assume that readers here prefer making cities (and the world) a better place rather than a worse place.

1(2) is less clear. If people have more money because asset prices are inflating, this could just be an artificial bubble (which we saw leading up to 2008). We may or may not be able to identify these bubbles in advance, and it may or may not be true Amazon is creating a bubble (it is one of the arguments Jeff throws against the wall in his tirade about the company) but let’s agree that bubbles are not good and if Amazon is creating one, that is also not good. So if the increase in demand is due to an asset bubble we have reason to pause. Let’s come back to that later when I work through Jeff’s specific arguments.

The other reason 1(2) happens is that people get truly richer (as opposed to fake ’bubble’ richer). When people get richer they have two choices: Spend the money or save it. In practice people usually do a bit of both. Whatever they choose to spend their money on will experience an increase in demand. Often when people get richer they choose to live in nicer places (or move out from their parent’s basement) – when they do that they will increase the demand for real estate and housing (and rental costs)

I know some people are against the idea of some people getting richer than others, but for now I’m going to assume that making a significant number of people in the city more well off is a good thing (vs. stagnating and/or decreasing incomes)

Finally, (2) basically says “If your city is a better place, people who don’t live in the city currently may want to move there now that it’s a better city.”

Unless we want to prevent internal migration, that also seems like a pretty good thing.

At this point, assuming demand is not increasing due to a bubble, you should be FOR increased demand (unless you are (1) Against making the city a better place, (2) Against making current residents richer).

In fact it’s even more simple that that. (1) and (2) are two sides of the same thing. If a city has opportunities to make people richer (great jobs for example) that’s part of what makes a city ‘better’. A big reason people move to a city (why they think it is ‘better’ than their alternatives) is that they can get a great job there.

Assuming flat supply of housing, raising real estate prices are a sign of a city that is becoming a better place to live – at least defined by what people are willing to pay for.

It’s easy to get cheap housing in the United States, just move to one of these cities: Top 10 Cheapest Cities to Rent an Apartment

The median rent for a two bedroom apartment in Wichita, Kansas is $650/month. You could live by a river and work for an airline-parts manufacturer. Or move to Tucson Arizona to live by the mountains for only $628/month.

But you don’t want to live in Wichita you say? You want to live in Seattle? Why? Because Seattle is awesome! How do I know Seattle is awesome and it’s not just my personal opinion? Because people want to live here. They want to live here so badly that they have driven up the price of real estate!

When people say they want lower prices, what they really mean is that they want to live in an awesome place but not pay what living in an awesome place costs. A Complaint about higher prices is really a complaint about higher demand which is really a complaint about change.

Let’s Talk about Supply

There is only one thing that will keep rents down while a city becomes a better place to live: Supply.

If you build more housing it will, all things being equal, reduce the value of existing housing. This is a big reason why existing home owners almost always protest new development. They say things like, “It will increase traffic” or “There won’t be enough parking”, but it all comes down to the fact that it will reduce the value of the housing that is next to the new development.

If yours was the only waterfront home and someone wanted to live on the waterfront, they would have to buy from you and you could extract as much as he or she would be willing to pay. But if that same person could build a new house next to yours on the waterfront, you now can’t charge more than the cost to build that house.

Increased housing supply drives down existing housing values (and rents).

That is part of the reason why entrenched interests make it hard to develop. San Francisco is the worst for this. Environmental regulations, building permits, zoning, etc. all make it more difficult to build. When it is more difficult to build, less gets built – which keeps prices high.

You really want housing to be more affordable in Seattle? Then ask the government to make it easier to build. Not just low-income housing. Any housing. Imagine if there were 100 more buildings in Seattle like the Escala. You can bet that the demand for units in the Escala would decrease (and be spread among the 100 additional buildings) and the price of units in the Escala would come down. If the Escala is cheaper, someone who was looking to buy in 98 Union might say, “Well, if it’s only an extra $20,000, why don’t I live in the Escala instead of 98Union?” Now 98 Union [and buildings like it] needs to lower their prices to compete. And lower prices at 98 Union means that someone who was planning to move into Capitol Hill might reconsider now that 98 Union is so affordable, which drives down prices in Capitol Hill.

But isn’t Seattle building like crazy? It sure looks like they are from all the cranes I can see from my office window. I think there are two reasons this increased supply is not (yet) driving prices down:

1. It didn’t happen fast enough. Development stopped in 2008. It took a while to get going again, and it takes a long time to complete a building.
2. Demand is growing even faster than the cranes are building

We are in an adjustment period. Assuming that development is allowed to continue (a big assumption) then prices will come down to an equilibrium.

But given that the real estate footprint itself is limited (there is only so much land within a mile of Pike Place Market), and assuming that demand for living in a specific place in the city will go up as the city gets better, prices will still go up. It’s just that if you are willing to compromise to live a little further away from your ideal location, more supply should spring up to make that no less affordable than living downtown was before.

(The concepts of using property prices as a way to understand the quality of a city were taken from one of my favorite professors at Wharton: Robert Inman)

• Traffic

But wait. If Seattle is getting so much better, why the crazy traffic? Traffic is definitely making the city worse. I haven’t seen firm data, but it would not surprise me if traffic is objectively worse in Seattle than it was 5 years ago. The fact traffic is worse is driving DOWN real estate prices (and rent). That should be obvious. If you had two choices of places to live that were identical and one had a really bad commute to work and one was really easy, it’s pretty clear which you would choose. But what if the really bad commute was a lot cheaper? Maybe you would change your mind. How much cheaper? It would depend on your personal preferences and how bad the traffic was. But we can be pretty certain no one would pay more for worse traffic.

So traffic drives down property values. But property values are still going up. Why?

Because the city is getting better faster than the traffic is getting worse. People are willing to put up with the traffic because of all the other good things Seattle has to offer (like jobs).

But couldn’t we make things even better by building great public transport?

Before you get too excited, remember: If we manage to build better public transport (and we can do it without raising taxes), then that will make the city better. Then what happens? Demand goes up and property prices (and rents) go up…

I keep saying ‘without taxes going up’. I do this because if taxes go up (without adding benefits), it makes the city worse. Go back to your two choices: Two identical places, one with no taxes, the other charges you $1000/year. Pretty clear which choice you will make.

Let’s say we can’t build more public transport without raising taxes (a realistic assumption). So the city raises taxes and invests those dollars in better public transport. Let’s say two breakeven. The taxes reduce property values but the better public transport raises property values. Basically the “people” (those who generate or reduce the demand) decide that the city got what it paid for (ideally the city is finding things to spend on that are better than that. They spend $1 in taxes to create $2 in value for the city – which would eventually increase demand for their city – and raise property values. Unfortunately most politicians are more interested in getting re-elected than they are in creating value for their citizens)

Maybe Seattle can do this.

Before we do though, it’s worth looking for a benchmark to see how we are doing right now. Everyone ‘knows’ Seattle has terrible public transport, but is there a metric we can use to understand just how bad? Like most things, it’s easier to go from terrible to bad than it is to go from great to awesome. There is diminishing returns on everything – including public transport. But if Seattle’s public transport is really really terrible, then we might have the opportunity for some low hanging fruit.

One way to measure the quality of public transport is to look at how much it’s used. If no one is using public transport it’s pretty clear it’s not meeting anyone’s need. If everyone is using it, it’s a sign that it’s likely doing pretty well. The good news is we have this data. The American Community Survey asks the question across the country annually: “How do you usually commute to work?”

It turns out that about 5% of people in the US commute using public transport. In Seattle the number is about 20%. That sounds great, but it is still pretty low compared to cities we think of with great public transport systems:

• New York: 55%
• Washington DC: 37%
• Boston: 35%
• San Fran: 32%
• Chicago 27%
• Philadelphia: 25%

So Seattle has a significant gap to the leaders, but we are still well ahead of most of the country:

• Kansas City: 1%
• Cincinnati: 2%
• San Antonio: 3%
• Atlanta: 4%
• Denver: 8%
• Portland: 12%

Seattle has the 7^th highest rate of public transport commuting in the country. We may not be awesome, but we are a long way from terrible.

But why aren’t we better? What’s stopping Seattle from having the public transport usage of New York City?

One word: Density.

Here is a chart of the ACS data combined with urban density data (granted this is from 2008, but the story still holds):

The x-axis has the share of commuting done by public transit. The y-axis is the population density of the city. Not surprisingly, the denser the city, the better the public transit.

But it’s not a perfect fit. The cities that are below the line are cities that have managed to achieve more use of public transit than their density would suggest. #1 on this ‘over-achieving’ metric is Washington DC. #2 is Portland. #3 is Seattle.

For its population density, Seattle is the third highest user of public transit.

Maybe we aren’t so bad after all? (And maybe the reason we complain so much about it is that we are using it at a much higher rate than comparable cities. People in Houston don’t complain about their public transit – they just ignore it…)

But let’s not rest on our laurels! How do we make public transit even better?

The obvious answer from this chart is to increase population density.

Which is exactly what Amazon is doing (and what Jeff is complaining about).

It’s the old argument that the world would be a much better place if there were a lot less people. If there were no people I would be able to get tickets to the Football game without paying a huge amount of money and I wouldn’t have to wait in line at the Starbucks. But this forgets the fact that without the people there wouldn’t be a football team – or a Starbucks.

It’s easy to move somewhere where there are less people. It just turns out that that is NOT what people want. They want the advantages that come with lots of people more than they dislike the disadvantages that come with lots of people. They just like complaining about it.

There is one more way to improve public transport (I know I said “one word” but there are actually two): Increase the number of jobs in the downtown core:

Guess what else Amazon is doing?

(Charts in this section were from Charlie Gardner on his blog oldurbanist)

• White Males

Jeff says Amazon only hires white men. Quoting from his post: “While the company reports 63% of its worldwide workforce is male, it’s likely closer to 75% male in the company’s Seattle technology headquarters; that’s the company’s overall managerial ratio and close to Microsoft’s technical ratio”

First: From walking around South Lake Union, I’m pretty sure Amazon hires a lot of Indian men. Either that or that ethnic group really likes the fruit at Portage Bay Café. Let’s assume that his complain about diversity is really a complaint about only hiring “White AND Asian men.”

Rather than addressing this directly I will throw it to (Founder of Netscape and Silicon Valley darling) Marc Andreessen. This quote is from an interview with New York Magazine. When he was asked about why Silicon Valley is so un-diverse he replied:

 “I think the critique that Silicon Valley companies are deliberately, systematically discriminatory is incorrect, and there are two reasons to believe that that’s the case. No. 1, these companies are like the United Nations internally. All the diversity studies say that the engineering population is like 70 percent white and Asian. Let’s dig into that for a second. First, apparently Asian doesn’t count as diverse. And then “white”: When you actually go in these companies, what you find is it’s American people, but it’s also Russians, and Eastern Europeans, and French, and German, and British. And then there are the Chinese, Japanese, Koreans, Thais, Indonesians, and Vietnamese. All these different countries, all these different cultures. To believe in a systematic pattern of discrimination, you’d have to believe that we’re discriminatory toward certain people without being discriminatory at all toward an extremely broad range of ethnicities and religions. Because of Pakistanis, we’re seeing a higher-than-ever proportion of Muslim employees in a lot of our companies.

No. 2, our companies are desperate for talent. Desperate. Our companies are dying for talent. They’re like lying on the beach gasping because they can’t get enough talented people in for these jobs. The motivation to go find talent wherever it is is unbelievably high.

There are two fundamental problems that are resulting in what a lot of people believe is discrimination, and these are the problems that I think need to be solved. One is inequality of education. If you come up through a path that’s sort of a stereotypical upper-middle-class American path and you go to Stanford and you get a really great technical education and your professors really care about you, then you come to Silicon Valley and you’ve got the skills and you’re golden.

But, of course, most people in the world—including most people outside the U.S. but also people in the U.S., like where I grew up in rural Wisconsin, or people in the inner city—never have access to that kind of education.”

Amazon isn’t discriminating any more than Silicon Valley as a whole is discriminating; they are hiring all the talent they can find. The real issue is the education long before people get to Amazon. But asking Amazon to solve the US education system might be a little bit much?

• No culture

So maybe I convinced you that Amazon isn’t discriminating. And that while it might be raising property values, that’s only because it’s making the city a better place to live. And while traffic might be getting worse in the short run, it’s still a better place to live than the alternative, and in the longer term the increased density will make public transit even better.

But what about culture? You can’t measure culture!

And even if they aren’t discriminating, they are still increasing the white-ness (plus Asian-ness) of the city. They are increasing the male-ness of the city. And they are driving poor artists out of the city. Eventually we will turn into New York.

Wait a minute. Does New York not have culture? Did the high rents created by the jobs in the finance industry destroy the culture of New York City?

Not so much.

When the World Cities Report on Culture chose a single city from each country to compare, America’s city was New York. When PropertyShark looks at number of cultural attractions New York had the most (2,693) (Seattle had the most per capita). TravelAndLiesure took their ranking system to the public and just did a survey. New York came on top again (followed by DC, Boston, Chicago).

“But Ed, those are all based on tourist attractions and the impressions of tourists. That isn’t real culture. Real culture is the number of people working as artists in a city. And real artists can’t afford to live in a place like New York.”

Except they can.

The American Community Survey has a category called “artists and related workers” – it basically means both employed and self-employed visual artists. Being a visual artist is a tough road in America. There are only 237,000 of them in the country. Guess where they live?

210,000 live in cities.

Which city has the most? For all the talk of the prices in New York driving out artists, it has more artists than any other city in the country.

“But Ed, that’s just because it has more people in total.”

Exactly! As a city does better it attracts more people, but it doesn’t eliminate artists, it just increases everyone else around them. New York might not have the highest per capita artist total, but it has the most people in one place doing art. And that allows a community to form.

Santa Fe, New Mexico has the per capita number of artists in the country. But when I was doing stand-up comedy, no one dreamed of going to Santa Fe. If you wanted to really learn stand-up and/or improv comedy you moved to New York, or Chicago, or LA. No one cares about per capita.

(Seattle by the way has the 5^th highest total of artists in the country. Anyone think that if Seattle grows it’s population 50% due to Amazon we will have less artists living here?)

Jeff’s Post

I just spend >4000 words on my introduction. It should cover most of Jeff’s arguments, but there are a few loose ends to tie up. Let’s work out way through them:

In “Seattle Today” he starts by complaining about the number of construction projects and the poor public transit (both complaints in the same paragraph). I think I’ve covered that in enough detail. Jeff: More construction will make public transit better. Just think about it.

He then complains about all the white people Amazon is hiring (covered) before throwing in the barb that women in the country are paid less than men. (Yes. But what does this have to do with Amazon, or even with Seattle? And the data is not nearly as clear as it is often made out to be.)

The Future is Obvious

Seattle will have more tall buildings (bad? Yes, according to Jeff because it will mean worse views for many buildings). And people will be wealthier (Also bad because it will drive out artists). I think I’ve covered both issues, but he covers himself by saying, “Some may find this exciting… But it’s the nature of the changes that concern me.”

Let’s dive into these natures.

Wealth Gap

Across the country (nothing to do with Seattle again) 5% of Americans control 60% of the wealth. He then claims this is what is driving an increase in crime in Capitol Hill and that February had the highest number of robberies ever (and two gun murders at almost the same time)

Wow. That sounds bad.

Let’s talk about crime for a few paragraphs:

“The most robberies”

Does the most matter? The city with the most robberies in the US in New York. I don’t need to look it up. It’s the most because it’s the biggest. Just like it has the most artists. The difference is, the absolute number of artists matter (number of opportunities and variety of art available, number of connections an artist can make), but in crime what you care about is the chance you experience it. You care about the per capita crime rate.

Guess what? Per capita crime in Seattle has been going down for more than a decade.

Here’s robberies (Robberies per 100,000 population):

Two things you should notice:

1. It’s going down
2. It’s noisy. 2009 saw a big increase in robberies vs. 2008. That wasn’t  the start of a trend, it was just random fluctuation. 2010 had the lowest robbery rate in recorded history.

It’s just not true that crime is increasing in Seattle. And looking at a single month in a single neighborhood is a nice way to scare people, but it’s a terrible argument.

Note: One of the issues with crime statistics is that there is a reporting problem. What if crime is going up, but police aren’t recording it? That’s a big issue with rape for example. It could be increasing, but being covered up. Or it could be decreasing, but being brought out in the open causing the number of reported incidents to go up. The normal way to solve this is looking at murders as a control. Murders are really hard to ‘not report’ so if the murder rate is moving in a different direction than your other crime statistics you might have a problem. Since Jeff complained about the murder rate in Seattle too, let’s take a look at that graph:

Yep. Looks like that is dropping too (and yes it went up from 2011 to 2012, but the ‘n’ is very very low. You have to be careful with low ‘n’ statistics as you know if you have been a regular reader of this blog)

Now that we’ve put crime to bed, let’s move on to his next argument:

“I’m not saying that Amazon shouldn’t grow and that others shouldn’t benefit from the opportunity, I just believe the company’s growing irresponsibly and [sic] beginning to have an irrevocably damaging impact on Seattle’s character and quality of life.”

Got it. So all those earlier complaints about growth were just to get people’s blood boiling. You actually have no problem with growth. You have a problem with their irresponsibility.

What exactly is Amazon responsible for?

Political Influence

Now we get to the heart of his argument. “Microsoft avoids paying taxes”. “Boeing gets tax breaks as a form of “legalized corruption””. This gets right into the argument that companies should stop holding their assets overseas and just bring them back to the US to be taxed. And that they shouldn’t use tax “loop-holes” to avoid paying tax. I wonder how many people who believe that choose not to deduct their mortgage from their taxable income or choose to forgo their deductions for their dependents. Asking companies not to use ‘loopholes’ that lawmakers create is asking managers of those companies to steal from the owners. If I owned 100% of Amazon and I asked the management to minimize my taxes, and they said, “No. that would be legalized corruption and unethical” I would say, “Fine. Minimize my taxes and then pay the difference out of your own pocket.” It’s really easy to tell someone else to give away their money. It is a lot harder to throw away your own.

I fully agree that politicians should eliminate deductions and have a much simpler tax system, but calling people who follow the rules unethical is pure political bating.

Here is a nice old quote I found from Bryan Caplan’s blog that is still relevant:

“I want to ask a question. What is a loophole? If the law does not punish a definite action or does not tax a definite thing, this is not a loophole. It is simply the law. Great Britain does not punish gambling. This is not a loophole; it is a British law. The income-tax exemptions in our income tax are not loopholes. The gentleman who complained about loopholes in our income tax – he did not refer to the exemptions – implicitly starts from the assumption that all income over fifteen or twenty thousand dollars ought to be confiscated and calls therefore a loophole the fact that his ideal is not yet attained. Let us be grateful for the fact that there are still such things as those the honorable gentleman calls loopholes. Thanks to these loopholes this country is still a free country and its workers are not yet reduced to the status and the distress of their Russian colleagues.”

–Ludwig von Mises, Defense, Controls, and Inflation

But back to Jeff.

He jumps back onto his public transit complaints and adds that we should be taxing development more to pay for better public transit. I’m fine with the idea of better public transit, but remember, if you make development harder you will get less of it. If you get less development your density will be lower and you will get… wait for it… less public transit. Not so easy is it Jeff?

This Jeff doesn’t like Jeff Bezo’s politics. He doesn’t like that Bezos gave $100K to fight a new income tax in Washington. Granted: That tax would hit Jeff personally and cost him a lot of money. But it would also hit Amazon. One reason Amazon can attract employees here is there is no state income tax. If the income tax in Washington had been has high as California I likely would never have moved here (or realistically the company that brought me here – Expedia – would have had to pay me a higher salary. And on the margin those higher salaries that they would have to pay would reduce the number of people they could hire).

So income tax in Washington would be pretty bad for Amazon. $100K doesn’t seem like enough of a donation – regardless of personal politics.

Amazon also gave $25,000 to fund an initiative to improve bus service (public transit in the city). The initiative failed. Somehow Jeff insinuates this is a bad thing too as it means even Amazon thinks current public transit isn’t good enough. But why Amazon donating to the cause signals their irresponsible growth is a lot less clear.

Back to diversity

Jeff has already shared all the data on diversity he has to offer. Now he dives into colorful stories. A woman got an interview at Amazon but did not get a job offer. She takes to local media to claim Amazon’s interview process is flawed (among other things it was five hours long! Oh dear! And no one offered her a bathroom break. Um. Maybe Amazon is looking for people with the initiative to ask if they can go to the bathroom?)

Other women complain that they don’t like dating engineers. I can understand why having more engineers move into the city would be bad if you hated engineers. I personally don’t understand this as I married an engineer and she is wonderful. I also wonder if this woman would be complaining about dating investment bankers in New York, service workers in Vegas, insurance agents in Hartford, entertainment folks in LA, and farmers in Idaho. Turns out there is a lot of diversity within any job class. When the problem is “everyone else” maybe it’s time to look in the mirror?

(Or to be fair, maybe her personality just doesn’t mesh with the common personality traits of engineers. Maybe the best thing for her love life would be to find an oil-man in Dallas. It might be worth her long term happiness to try, rather than complaining?)

Jeff has one last diversity attack before he moves on:

Amazon is driving up hate crimes against the LGBTQ community.

That’s a heck of an accusation, and what caused one of my good (gay, Amazon-working) friends to share Jeff’s article on facebook with a comment. Rather than argue this myself I’ll just quote his post:

“In an effort to somewhat balance this very slanted article, I’d like to say I’m personally very thankful/proud that Bezos gave $2.5 MILLION DOLLARS to help fund the campaign for marriage equality in WA – the first state to have this right passed by popular vote. Insinuating that a rise in LGBT hate crimes and gun violence is because of Amazon is really offensive (among other awful statements about Amazonians like me.) In other news, you are all invited to attend this Amazon-funded gala where 100% of proceeds benefit a local gay-straight alliance youth chorus: http://2014glamazongala.eventbrite.com/”

 (Just one thing: $100K to fight the increased taxes is self-serving in that it helps Amazon attract workers. $2.5M to fund marriage equality also helps attract workers – both the LGBTQ community and those who support the community (which includes most tech workers). Both help Amazon, but one amount is obviously MUCH higher than the other. Still say Amazon hates diversity Jeff?)

But Jeff has just touched the surface so far, “But what about at a deeper level?” he says…

Confrontational Culture

Next Jeff starts blaming Amazon for the “Seattle Freeze”. For those not aware, the “Seattle Freeze” is the belief that people in Seattle are really nice but they are also cliquish and you won’t get an invitation to their place for dinner. It’s hard to make good new friends here.

I moved to Seattle in 2009 and I found at least part of the stereotype to be true. I joined an ultimate Frisbee team to meet people and while I had a great time playing with them on the field, it didn’t turn into any off-field friendships (unlike similar situations where I’ve played in other cities). Over time I developed a theory on this:

Most people in Seattle back in 2009 had grown up here or been here a long time. They had a long-time friend group, and there wasn’t a strong need or desire to bring in new people who just moved to the city. They were friendly, but no, they didn’t need to invite you into a friend circle that was working very nicely thank you very much.

It took many years (and meeting my future wife), but I now have a very broad friend group in the city. What changed? New people. In the last 6 years Amazon has brought a TON of new people into the city. Unlike people who have long-standing friend groups, these people are looking to make new friends.

Amazon isn’t increasing the Seattle Freeze with their culture, they are dramatically reducing it with changing the mix of the city.

It’s hard to make friends later in life after you have left school and gone on to your second or third job. But the easiest way to have it happen is to be surrounded by people who have to build new friend groups at the same time you are. Thanks Amazon.

Amazon mistreats employees

Amazon works its employees hard. So does McKinsey. So does P&G. What all these firms have in common is that after you have spent some time there your personal brand improves. People know that you are able to work hard. And people know that you have been trained in the “Amazon” (or “McKinsey” or “P&G”) way of doing things. Amazon doesn’t hide it. No one takes a job at Amazon and then says “Wait. You want me to work hard?” (Okay. Maybe some people do, but they were hiring mistakes).

I don’t work at McKinsey anymore and I will never go back to that lifestyle, but I believe the choice to work that hard for four years of my life was definitely worth it. Can we give Amazon employees the same benefit of the doubt? And if they don’t want to work that hard, maybe they should quit and get another job? That’s the beauty of our capitalist system.

“But Ed, they are quitting. Jeff says the average employee retention is 1 year vs 4 years at Microsoft.”

Great! Even more benefit for Seattle.

If Amazon hired all these talented people and then kept them inside Amazon that would be good for the city. But even better for the city is attracting these talented people to the city, training them for a year, and then setting them loose. Wikipedia lists 19 “Notable” companies founded by ex-Amazon employees. And that doesn’t include all of the ex-Amazon employees filling the senior and middle-management ranks at technology companies across the city. You would be hard pressed to find a technology company in Seattle that doesn’t have someone from Amazon somewhere in the mix.

Philanthropy

According to Jeff Amazon gives 0.5% of sales to charity. My first reaction to that is that I can’t believe it’s that high. He must be making a mistake. Amazon is a retailer. Which means it operates under some pretty tight margins. It’s not uncommon for retailers to have 10% or less gross margins, from which they need to pay for their fixed costs (including logistics, real estate, etc.). On a good quarter Barnes and Nobel will make 2.5% profit (on a bad quarter it is -2.5%). If they were to donate 0.5% of 2.5% that would be 20% of their profits.

(I won’t repeat my earlier argument that management choosing to give away profit that is owed to the owners is not that different from stealing. It’s fine if individuals want to make donations but getting people to donate other people’s money is just an easy way out – especially it’s not your money being donated)

Jeff also praises Bill Gates. How soon we forget. Bill only became a philanthropist late in his career. For decades he led Microsoft that was often accused of being stingy. Amazon is still in the very early stages. Every dollar they donate is a dollar they aren’t investing in future growth and innovation.

Another quote from Jeff:

“Researching this piece I was most struck by a side note in Brad Stone’s Business Week expose, “New hires get a backpack with a power adapter, a laptop dock, and orientation materials. When they resign, they’re asked to hand in all that equipment—including the backpack.” If push comes to shove Seattle, expect Amazon to treat us with the same regard.”

What does that mean? If Seattle quits Amazon they will demand their laptop bag back? I’m sure there is supposed to be a metaphor in there somewhere, but I have no idea what that metaphor is…

(By the way: Giving your laptop and power cords back when you leave a company is pretty standard. I’ve never heard of a company not asking for them back. Giving them back in the laptop bag rather than handing them back a loose pile of cords seems like a polite thing to do. I’ll bet if you lost the bag for some reason Amazon isn’t charging you for it. Give me a break.)

Amazon is a bubble

Ah. We are finally at some meat. As I said early in this essay: If all this growth is only a bubble maybe we will be in trouble. Amazon goes bankrupt. Housing prices crash. Everyone moves away (on the upside: Seattle traffic will disappear and rent prices will come right down… Wait a minute: Aren’t those the two things Jeff wants to happen?). Really Jeff isn’t making a argument so much as throwing every Amazon complaint at his computer screen. “Amazon isn’t profitable” is just another complaint about the company.

(Want to know how to make Amazon even less profitable? Have the employees work less hard, have the taxes in Washington State go up and have Amazon give more to charity. Just sayin’.)

This post is long enough without a defense of Amazon’s secretive investment practices. But know that if Amazon capitalized and spread out the expense of all their marketing and R&D over 5 years instead of 1, their EBITDA would look a lot better. Their cash would not – but that is because they are taking every dollar they have available in cash and investing it in ROI positive opportunities (at least to the best of their abilities). It’s a very reasonable strategy that many could agree or disagree with (For example you could contrast it with Apple which holds onto a lot of its cash for opportunistic opportunities).

Investors hit Amazon recently, but that was driven more by the failure of their bet on the Fire Phone than it was on their basic strategy of investing in growth.

What Amazon Could Do Differently

Jeff ends with his recommendations to Amazon.

1. Advocate for higher taxes for people living in Washington
2. Recruit women aggressively from your competitors. Pay them a premium over men so you can get them and increase the diversity of Seattle
3. Make donations to local Seattle charities to help the poor (and artists)

#1 is just ridiculous for Amazon to consider. Why would they do that?

#2 is interesting, but would not increase the number of women in tech – it would just move them from wherever they are to working at Amazon. It would make it a lot harder for other tech companies in Seattle to have any women working for them. The women gain (since they are getting paid more), but I’m not sure how this really helps

#3 Is basically Jeff saying: “Please Mr. Bezos: Can you take your money and give it to my personal causes?” Maybe Jeff’s causes are more important than what Bezos cares about (like say the $2.5M he donated to marriage equality). But given his ‘logical’ arguments in this piece I would bet on Bezos’ choices for donations over his at this point.

This post was pretty far from my normal posts on marketing and understanding data, but I’ve been asked about this article multiple times this week. I thought it was important to get my critiques down on paper. I think the arguments Jeff made this week will happen again in a different context in the future. I like the idea of being able to send people back to this post every time someone complains about the rent going up.

Thanks for giving me the opportunity and structure Jeff.

And thanks for reading all the way to the end of this monster.

I will try and respond to all comments below.

(Note on conflicts of interest: Other than the fact I live in Seattle, I have no financial interest in Amazon (except maybe as part of an ETFs that I am unaware of). I am also not a property owner in Seattle, so I have no personal motivation to see property prices go up. In fact higher property prices would eventually mean higher rent that would come out of my pocket. I’m also married so the gender ratio of the city doesn’t affect me as directly as it does the singletons in the city)

This is Part II of my look at automating authenticity (and why it’s a bad idea). Part I talked about Twitter specifically (I recommend clicking the link and then coming back to this piece). In Part II I am going to apply the principles to customer communication in channels like email, direct mail and websites. To start with we will look at Target.

Why the Target story is BS

No talk of personalization is complete without talking about the Target story. For readers who have not heard the story, it goes something like this:

• A woman receives a flier from Target with stuff for pregnant women
• Her dad sees the ad and is furious as his daughter is still a teenager
• He complains to Target and the manager apologizes
• Later the man goes back to Target and apologizes himself. It turns out his daughter WAS pregnant and he didn’t know(in some versions of the story the daughter didn’t know either)
• The moral of the story is that Target’s personalization was so good they knew who was pregnant before the woman did(or at least the father)

I don’t know anyone who works for Target, but I know the story is complete BS. Here’s why:

1. The Story: A man looks at a flier from a store and then complains to the store about what they are advertising? Does that make any sense? Wouldn’t it be more likely he would see a flier for pregnant stuff for his daughter and just ignore it? Who complains to a store about ads for stuff you don’t want? Don’t we all get ads for stuff we don’t want all the time
2. The Targeting (sic): Assuming Target even had a program trying to target pregnant women, and even assuming it was superhumanly accurate, it would be wrong more than it would be right.

How could something be extraordinarily accurate and still wrong more than right? To understand why we can dive into a different type of pregnancy prediction: bad chromosomes.

There is a test that will tell you with 99% accuracy whether the embryo in your belly will have chromosome problems. 99% sounds pretty definitive. If the test tells you your kid is going to have issues, you better be expecting issues, right? Wrong.

What is more important than a test’s accuracy are the underlying probabilities in the population being tested. In this example, let’s say you are a 25 year old woman without risk factors. Before you take the test they know that the likelihood of someone like you having a baby with specific chromosome problems is about 1 in 5000. So let’s look at what happens after the test (Each cell is a person out of 50,000 people being tested)

(Obviously the decimals are just estimates. In reality people either fall into a box or they don’t. Why we can simplify and eliminate the decimals for this example)

Better?

What does this chart tell us?

Look at the column on the far right first. This is just the sum of the real population. As we said earlier 1 in 5000 25-year old women will have chromosome problems. So out of 50,000 women, we can expect 10 to have problems. Easy so far.

Now we run the test. Each test result gets its own column. We don’t know which group an individual woman will fall into (that’s why we are running the test!), but in our omnipotent world we can look at each group separately. This means looking at each row one at a time.

In the top row, for the 10 women who do have the problem: With 99% accuracy the test should basically get it right every time (once we rounded up in our second example). 99% accuracy seems to do a really good job. No one who has a bad chromosome is getting a test result saying everything is okay.

Now let’s look at the second row. Here there are 49,990 women (i.e., almost all the women). The test is 99% accurate here too, but this time the 1% failure rate leads to a lot of misdiagnosis. 1% of 49,990 is 499 women. 499 women is a pretty small number compared to 49,990 women (1% in fact), but a small percent of a really big number can still be significant. Still, if you have a healthy fetus the test will tell you that 99% of the time. Sounds pretty good.

But now let’s flip it. Instead of looking at rows from our omnipotent vantage point, let’s look at columns. Columns are what we can actually see with results. We don’t know if a woman has a bad chromosome or not, we just know what the test says. That’s what the columns tell us.

In the right column the test says ‘all clear’, ‘you are completely healthy’. And it turns out the test is right. For the 49,491 women it gives an all-clear to, they get it right basically 100% of the time (after our rounding. If you go back to the earlier table you can see they actually get it wrong 0.1/49,490.9 – or less than 0.0005% of the time). If the test says you are fine, you can pretty much sleep easy. You are fine.

In the left column we get a different story. 509 women will be told that they have a positive test result and that their baby has a bad chromosome. Those women will, at the very least, go through a lot of stress. Some may even choose to abort their baby. But almost all of those women will have perfectly healthy babies. In fact only 10/509 of these women will have babies with chromosome problems. That’s not the 0.0005% odds of the right column, but it’s still only a 0.2% chance of an issue.

So our test that was 99% accurate actually only got it right 0.2% of the time. Those stressed out women who are told that they are 99% likely to have a baby with a chromosome problem (because many doctors explain it that way and many people interpret 99% accurate that way) have a 99.8% chance of being completely fine.

It’s perplexing math.

Hopefully I’ve covered in in enough detail that you have internalized it. If not, please do ask questions in the comments below. I would be happy to expand further.

(For those who are wondering at the example: My wife and I went through this exact situation while she was pregnant. It was indeed a false positive. And I had to do the math myself to understand the ingoing odds.)

What does this have to do with Target and automated personalization?

Good question.

The first problem with automated personalization is being accurate. Developing a model that can predict what you are trying to predict using whatever data you have. If you ask people, “Are you pregnant”, the answer to that question is likely pretty good at predicting if someone is pregnant (or at least believes they are pregnant). Maybe it gets it right 95% of the time (5% of the time people lie on the question). I would go so far as to say it would be the single best predictor of whether someone is pregnant. Much better than their buying behavior (one of the conclusions often made from the Target story is that companies will be able to predict your needs better than you can yourself. They knew she was pregnant before she did! I don’t believe that and have certainly never seen it in reality – and I have helped dozens of companies develop models like Target claimed to have).

Now let’s say that Target managed to build an amazing model that predicted pregnancy. Let’s say they were 90% accurate – almost as good as just asking people. The issue is the same as what we saw with the chromosome tests: Only a small number of women are pregnant at any given time. Let’s try to estimate that number roughly. Let’s only look at women ages 14-64 (50 years). Let’s assume that the population is evenly spread out along that time. Let’s assume that some other Target model can predict the customer’s gender and age extremely accurately so they can assume that men and women under 14 or over 64 are never pregnant (so their accuracy is actually much better than 90%. It’s 90% accurate for 14-64 year old women and 100% accurate for everyone else). Let’s also say that women have an average of 2 children each during that time period and the model can pick up on things after 3 months of observing their buying – so about 6 months per pregnancy times two pregnancies equals 1 full year during those 50 years – or 2% of the women.

With those facts let’s build the same table we did for the chromosomes:

These odds are not as wild as the chromosome table because a 2% ingoing likelihood is much larger than a 1/5000 ingoing likelihood, but hopefully at this point you can see the same problem.

If the awesome 90% model says someone is not pregnant it is almost always right – 48510/49410 of the time (about 98.2% of the time). But if the model just assumed that women were NEVER pregnant it would still be right 98% of the time (since they are only pregnant one year in fifty).

If the model says the targeted woman is likely to be pregnant then it does much better. It’s right 100/509 of the time – about 25% of the time. That’s much better than the 2% a monkey would get by just guessing randomly. But see the problem? This 90% accurate model is still wrong 75% of the time when it says a woman is pregnant.

If you use information like that to show ads about pregnant stuff to women you will be showing those ads to non-pregnant woman 75% of the time.

But, and here’s the kicker: There is nothing wrong with that.

When a company puts an ad on TV for diapers there are far less than 25% of people watching that might need to buy those diapers. Ads touch the wrong people all the time. The goal of modeling is to try and get the ads to touch wrong people a little bit less – and that’s a great thing when it costs you money to touch someone who is not going to be interested.

THAT is why you do modeling and try and improve targeting.

Unfortunately marketers and data analysts have started drinking their own Koolaid and they seem to believe they can be 100% accurate (they can’t). And when you believe you are 100% accurate you start using the models to do things the models have no business doing.

Like automated personalization.

There is nothing wrong with sending an ad for pregnancy stuff to all of your mailing list. No matter what the Target story will infer no one is going to storm your building demanding an apology for getting a bad ad mailed to them.

If that mass mailing is ROI positive, then by all means use some modeling to limit who you send it to (here’s some free modeling for you: Don’t send it to single men. Don’t send it customers below the age of 20 or above the age of 40. Maybe only send it to married women. Yes younger and older women and single women get pregnant, but you aren’t trying for 100% you are just trying to improve your odds).

But here is a terrible idea:

Send a mailing to the ‘modeled women’ with a personalized message that says (in so many words), “We know you are pregnant.”

“Experts” will tell you not to do this because it’s creepy. I tell you not to do this because it’s wrong (and not morally wrong – just wrong wrong).

You don’t know she is pregnant. You only have a 25% shot of being right.

What is true of pregnancy is true of any low probability event. Not most of them. All of them.

And almost all ‘personalization’ relies of using automation to send unique messages based on models of low probability events.

Your personalized automated authenticity ends up falling into one of two buckets:

1. It’s not personalized. It’s generic. This is the bucket for 90% of the ‘personal tweets’ I received on a daily basis
2. It’s personalized wrong. This is the bucket that most companies fall into when they try and automate the personalization of their websites or communications

Hopefully I’ve driven home the point hard enough. All this said, there ARE ways you can “personalize” effectively. I will cover that in another post. Stay tuned.

Two things have happened to me a lot recently that have merged in my mind.

The first: Generally I get very positive comments on my Twitter ‘followback policy’, but every now and then someone complains that twitter should not be automated and that you need to have authentic relationships. How dare I automate my following? It’s immoral (if not evil).

The second: There has been a huge push for companies to provide more ‘personalized’ experiences. The thinking is that if a company can make a website or email more ‘personalized’ then users will have a better experience.

I am pro auto-following but anti auto-personalization.

This post is to help explain why.

It has nothing to do with morals and everything to do with capabilities.

What computers are good at

Computers are really good at doing what you tell them to do. If you can predict exactly what you need them to do, you can program them to do that. The more random or uncertain the situation the more computers falter. It is no surprise that computers mastered chess before Jeopardy before Go and it will be a few years before we ever see a computer write a novel worth reading.

Following people on twitter has two parts you need to do in order to do it well:

1. Figure out who you want to follow
2. Click a button to follow those people

The first part is somewhat hard. There is obviously nothing difficult about the second part.

When I find something that is “somewhat hard”, I believe the best method is to do it manually. As I do myself I start to internalize the “rules” I am unconsciously following to succeed at the task. I can rarely guess what those rules would be before I started, but sooner or later my activities become automatic. If I spend a little time thinking I put those “kind-of rules” into “strict rules” that a computer could understand.

At that point it’s time to automate the activities.

For my twitter following that means following people who tweet about specific things, who have specific terms in their profiles, who tweet at specific frequencies, who have low spam estimations, and other characteristics. Once I figure that out I don’t need to keep doing it myself. I can outsource it to either a computer (or in this case, another human, since twitter bans the use of computer automated following). Whether it is a computer or another human the principle is the same: I create detailed rules for someone/something else to follow to the letter.

Doing it myself at this point doesn’t make it more authentic or real or anything else. It just makes it more work.

The only advantage of doing it myself rather than outsourcing are the errors. My rules will not be perfect. If I do it myself I will presumably catch some of those ‘rule errors’ and prevent them from happening. But doing it myself can result in a different category of errors – where I fail to follow the rules at all and that results in bad final results. So whether I do it myself or use others there will always be errors. One of the concerns about self-driving cars is that they might get in accidents. They WILL get in accidents, but the real metric is if they get in more or less accidents than the alternative. In the case of following I think the automated method makes more errors. But an error in this case just means I follow someone I would rather not have followed. That might pollute my feed a little until I correct it (by unfollowing), but otherwise it does me no damage. So I am okay with the automation.

What computers do poorly

Another thing I automate with twitter is when my tweets are sent. I use BufferApp to spread out my content tweets about stuff I find when I am reading, and I use SocialOomph to spread out tweets about this blog (Re-tweets I usually do myself without any automation at all). But while I use tools to automate when tweets go live into the twitter-verse, I never automate the content of the tweets.

The reason I don’t is that computers are very very bad at this.

There are tools out there that will do that for you. They scan for content related to the content you have shared in the past or content your followers are sharing with each other, and then auto-send tweets for you on your behalf. I have very little confidence in a computer’s ability to guess what content I think is valuable enough to share with my followers, so I would never use a service like that. To me THAT is being unauthentic.

I believe the only thing worse than autotweeting content, is autotweeting personal ‘messages’.

I just took a break from writing this to check my Twitter notifications. In the last hour (60 minutes) I have received 5 auto tweets. For the record they are:

.@Ednever Thanks for supporting .@XXXXXX Did you get a $25 reward level t-shirt yet? Let´s connect. XXXXXXX@gmail.com

• Really? You want to “connect” with me? Do you really mean “I want to spam everyone I can with an offer (including you) and then throw on an ‘offer’ to connect so it doesn’t look spammy”? Cause if so, it didn’t work.

@Ednever welcome

• Uh. Welcome to you too. I guess. Am I supposed to do something now?

@Ednever Thanks very much for the follow, Edward. Looking forward to your Tweets!

• Mine and everyone else’s. You have failed to make me feel special.

Thanks for the love my #leaders @XXXXX @XXXXXXX @Ednever I always appreciate it.

• I know. I can tell from your feed. Always.

@Ednever Thanks for the follow!

@Ednever Thanks.

• Am I obligated to say “You are welcome”?

How do I know these are all auto-tweets? The easiest way is to just open the individuals twitter feed and see all the identical tweets they sent to other people. They aren’t really using twitter to share information as they are using twitter to repeatedly send the same message to everyone who interacts with them.

My DM feed is even worse. I can’t be sure that DMs are automated (since I can’t see what they are sending to everyone else), but I am confident enough that I have effectively stopped checking it.

Here’s the last random DM in my mailbox:

• Thanks so much for the follow I tweet about healthy living and network marketing. What do you like to tweet about?

What do I tweet about? Um. Maybe the best way to know that is to read my twitter feed. Or to read my bio. Or visit my website. Maybe the best way isn’t to automate a tweet that asks that question to the mailbox of everyone who follows you?

This is the ugly side of automation. It’s trying to fake a connection when none exists. There is no difference between a computer ‘clicking’ to follow you or a human clicking to follow you, but there is a big difference between auto-messaging everyone who follows you with a question that asks them to do work (while you automate your work) and sending a personalized message to someone who follows you based on reading what they have to say and making intelligent conversation.

It can be fun to have conversations with other humans. It’s insulting to have conversations with computers pretending to be humans.

In Part II I will talk about why it is so hard for a computer to do automated authenticity well and how it applies to email programs, websites, direct mail, and other things non-twitter.

In April 2014 I had the good fortune to travel with Warburg Pincus through three cities in India. We had the opportunity to meet with a dozen different up-and-coming India technology companies to talk about their products, their marketing and their business plans. My purpose there was to help with their various marketing challenges, but I am pretty sure I learned as much as I shared. Here is a summarize of some of my my key take-aways from the conversations.

1. Don’t undervalue the ecosystem

I don’t think we realize the huge advantages companies have here in America. To best demonstrate this let’s compare Fandango to BookMyShow. Both were founded in the late 90s to do the same thing: Bring movie ticket sales online. Today Fandango dominates that market across the country. BookMyShow dominates the same way in India, but the path there was a lot harder.

BookMyShow had to install their own printing kiosks (since the theatres didn’t have them). They sometimes even had to provide security so the kiosks were not destroyed and wireless so that they worked. With that expertise they tried to offer full events – like LiveNation. But to do that they needed turnstiles. There was no easy way to rent them, so they ended up owning hundreds of turnstiles – and a fleet of trucks to transport them to events across the country. And obviously they issued the tickets for the events. So today they are the Fandango-LiveNation-Ticketmaster of India. But because the market is so immature, they are nowhere near the size of even one of those companies in the US. Which leads to the second insight…

2. Being too early doesn’t mean you fail, but it means you need a lot of patience

BookMyShow has been waiting more than 15 years for the movie-ticket business to really take off in India. They have managed to stay in business during that time and now they are well positioned to take advantage of it. In fact it would be very difficult for anyone to come in and try to compete with them in the space they have carved out.

This ‘in early and wait’ is even more true in used car sales. In the US about 50% of car sales are new and 50% used. In India 90% are new – the growth in the market means there just aren’t enough cars out there to meet the growing demand. That will obviously change when the market matures, but that is still years away.

In mature markets there is almost always a winner-take-all company that acts as the ‘market maker’ for used cars (for a lot of reasons it makes sense to have all the listing in one place). In the US that is Autotrader. In established markets there is no single dominate player in new cars. In fact ‘portals’ for new cars tend not to make much money at all.

But in India the used car market isn’t big enough to support a company like Autotrader – yet. But it is very obvious it will be sometime in the future. You might think that means the smart money is on waiting until the market gets bigger and then launching a Autotrader-like copycat. But if you did that you would lose.

There are at least three strong car portals players in India right now. All of them are trying to position themselves to be the Autotrader when India is ready for Autotrader, but they need to keep their lights on in the meantime – which means spending some time (if not most time) on building a great new car portal.

It’s a waiting game. It’s likely only one of the three will ‘win’, but it’s pretty clear one of them will – it won’t be someone who jumps in when the timing is right.

3. It’s not all about copying – Local market conditions matter

These car portals are doing all sorts of things that US companies don’t need to. In at least one case, when a new car is listed, the company sends an inspector out to determine if the car is a quality used car. If it is they put their stamp of approval on it and offer discounted (or free) warrantees.

Even more extreme is OlaCabs – the India version of Uber. Which is a funny statement in itself since Uber has already launched in India in a few cities. But these guys are in over a dozen markets and growing much faster than the American ex-pats. They have also adjusted the model to local conditions.

Uber’s model in India is basically the same as the US.

OlaCabs does things differently:

• 95% of people pay cash, not credit card
• The drivers are required to ‘deposit’ the 20% revenue share in advance before they are given any fares. When they have earned the 80% of what they deposited, they need to head back to the office to make a new deposit
• 90% of fares are booked in advance (rather than in for immediate pick-up)
• A significant share of fares are booked over the phone instead of via the smartphone app

“Wait a minute!” you might say, “Phoning in to book a car in advance and then paying cash? Isn’t that less Uber and more just ‘Taxi’?” It’s true. The basic taxi system in most Indian cities is so poor, the “Uber of India” is disrupting the Taxi business and the pre-taxi business (rickshaws, limos, private drivers) at the same time.

4. You can break all the rules

It was easy to see parallels across the companies I met: They all had to build infrastructure around their businesses – which tended to get them involved with auxiliary businesses and horizontal expansion; They all tended to be entering markets far earlier than the market existed; They all had to figure out an “India” way of doing things.

And then there was Zomato.

Most Americans I’ve spoken to have never heard of Zomato, but I bet they will. Zomato could be described as the “Yelp of India”. Like a lot of the other tech companies trying to make things work there, they were forced to come up with an India way of doing things. Rather than just relying on consumer reviews, they created city teams that visited every restaurant in a city and took pictures. Most importantly they took pictures of the menus and transcribed the results – so they have a searchable menu database of every major market in the country. If you want Chicken Parmesan in Mumbai, they can tell you where you can get it.

After being successful in this space their next step was clear: No Indian consumer internet company has ever been successful outside of India, but it should be child’s play for them to expand horizontally and become the OpenTable and GrubHub of India.

They decided not to. They went international.

And it worked.

They are now in 13 countries and on track to be profitable in all of them. They are taking on Yelp (and some would say winning) in London and Germany. They are coming to Canada soon to see if they can take on Yelp closer to home.

Did I mention they generate more revenue per pageview than Yelp does. A lot more.

Which shows you that as soon as you develop ‘rules’ by generalizing across a bunch of companies, along comes a company that breaks all those rules. Turns out there aren’t rules at all – just examples that predict the past pretty well, but don’t work so well in predicting the future.

A few weeks ago I posted my take on the History of Marketing. I was apparently one of my more popular posts. In addition to the social media sharing, I received an email asking a question about this paragraph:

First, there is a difference between proof and Proof. Significant results are still wrong 5% of the time (and likely not important 50% of the time). When you are running Big-Data sized tests and trying to backward-infer results, this 5% gets really really important.

Here was the question:

“Significant results are still wrong 5%”  – is that really the same as there being a 5% chance of wrongly rejecting the null hypothesis? I’d like more on that please and, especially, the subsequent point about 50% irrelevance which I think is even more important for people to understand.

Rather than answer it via email, I thought I would expand upon it here for general consumption.

1. Yes. When I say “Significant Results” are wrong 5% of the time, I mean that the very definition of “significance” is that it is wrong 5%of the time. You wrongly reject the null hypothesis 5% of the time. Always.
2. The 50% comment: What I meant here is that even if your result is “proven” to be different from what you are testing it against, it says nothing about the magnitude. What often happens in practice is that your A/B test gives results like this:

A- 10% C/R
B- 11% C/R
Significance test says it is significant with a more than 95% certainty.

The analyst who ran the test takes the result to leadership and says the new landing page will be 10% better (and it’s significant!)

The new page is implemented, but a few weeks later they find C/R hasn’t gone up by 10%. In fact it’s gone down. How is that possible? (This happens ALL THE TIME by the way!)

Here’s what happened:
The ‘real’ result from B wasn’t a 10% improvement, it as a 1% improvement. But the randomness produced a 10% result. The 95% error bars around the result verified it the difference was more than a 0% improvement, but no one asked, “What is the chance this is only a 1% improvement?”. Turns out those odds are a lot higher than 5%.

But the good news is that it is at least a positive impact. So why didn’t the results after implementation show a 1% improvement? Two reasons:

1. There is more randomness after the fact. Your C/R is going to fluctuate all the time and it’s very unlikely you will be able to notice if your ‘real’ C/R shifts from 10.00% to 10.01% after a week (or a month)
2. Things change. Just because your new landing page is better than the old landing page last week, doesn’t mean it will be better next week. We know this intuitively, which is why we run A/B tests simultaneously rather than sequentially any time we can, but somehow we just think that those simultaneous tests will always turn into impact in a future sequential time period.

I’ll have another post at some point (likely more than one) that digs into this issue further. It’s a big issue that becomes more important as Big Data analytic techniques become more common (although it’s still a problem with traditional techniques the way they are often used)

If you have more questions on any of this feel free to comment below or email me (or even better sign-up for emails when I update the ‘book’ and just reply to the welcome email. It goes directly to my personal account).