Sketch Comedy with Kevin MacDonald

This is definitely NOT a Marketing post.

Back in March I had the opportunity to take a sketch-writing workshop with Kevin MacDonald (of Kids in the Hall fame). We spent the day learning his techniques, then went up on stage with a random group to do a single improv sketch. We spent the rest of the day workshopping that scene to turn it into a performance-ready sketch. That evening we performed the sketch to a paying audience.

You won’t learn any marketing from the following video:

Media Mention: The Street

A few weeks ago The Street wrote a trend piece on how children are increasingly becoming responsible for their parents financial well-being when it has traditionally been the reverse. As part of the article I was quoted based on a survey we had done at A Place For Mom on how prepared families were for senior housing needs.

“Through our work with families, we find that it’s quite common for adult children to provide financially for their aging parents, but it’s not often clear if families had expected or planned to do so,” said Ed Nevraumont, chief marketing officer with A Place for Mom, a senior living referral service that co-commissioned the survey.

Here is the full article:

Personalization is BS [the video]

Last month I had the honor of speaking at TechTO, a monthly tech event in Toronto Canada. Each speech is only 5 minutes followed by 5 minutes of questions. I decided to do a short speech on how Personalization isn’t all it’s cracked up to be. I thought it went very well. The Q&A afterwards was also valuable.

Here is the full 10 minute video:

Media Mention: Top 10 CMOs to watch in 2015

Wealth Engine recently released their Top 10 CMOs to watch in 2015.

It’s unclear how the list was made. My guess is the SEO team at WealthEngine thought they could get some social-juice/link-juice by promoting some CMOs that have large social media followings/blogs. Well it worked. Here is their link:

Interesting that they built my profile without speaking to me. Itwas just pulled from this blog I believe (with some errors like saying my “Marketing Is Easy” book is complete, when I think that is only about 10% true)

Media Mention: Time Magazine

It has been a couple of weeks since I have posted here. The new baby has made it hard enough. Then last week I spent the week in India talking to 15 different tech CEOs there. I was last in India in April last year (Here is the post I made based on what I learned on that trip). This was the return engagement. I was hoping to write up my summary of the trip on my flight home, but I ended up sleeping and watching the Hunger Games instead.

Expect the India Part 2 update in the next couple of weeks.

In the meantime, yesterday I did an interview with (part of the Time group). The impetus was APFM’s release of a price index for senior housing. There are lots of senior housing price index’s out there, but all of them are built on self-reported data from the communities. But APFM is the only one in the industry that knows what people actually pay.

Each method has its advantages. The self-reported data allows indexes to track changes in the asking price for care, our index allows tracking of changes in what people actually pay. If discount levels change over time, our index will catch that (the traditional indexes will not). If the average care requirements of seniors moving-in are going up  (or going down) over time that would not change list prices (traditional indexes) but it would affect our index (because actual prices paid would be going up).

Again, each method has its own advantages, but until now there has only been one method to get any data at all. I’m excited we’ve pulled together the unique data set we have to expose what people are actually paying across the country. It’s our attempt at bringing a little more transparency to the industry.

Phil Moeller at was a pleasure to talk to. What started as a discussion just about the index went much broader. We talked about the challenges people have looking for care. How it is much more expensive than many people think; How important it is to start thinking about it early; How the housing bust in 2008 is now causing prices in senior housing to increase as supply has not kept up with the recent spike in demand; How those increases in prices are causing people to wait longer, driving down demand, and increasing the average age of the seniors who move.

His full write-up on our interview can be found here:

EDIT 3/26/2015: Senior Housing News has also picked up the story. Their version is here:



Virality is BS

Everyone wants their product to go viral. Why wouldn’t you? It’s like the holy grail of marketing.

Too bad it’s BS.

If virality was a “thing” that could be reproduced on demand it would be the only marketing that would exist. Why would you do anything else?

And yet I constantly get marketing services companies pitching me on their ability to make things go ‘viral’. When I advise CMOs at other companies it is one of the top questions asked. My own team regularly asked me what we are doing to get our product to go viral.

Before I demolish the idea of virality completely, I will say there are some basic things you should do. You should make it easy for people to talk about your product. You might want to have a basic referral award system. Ideally you want to make your product visible when it is used.

These are all smart business practices. They should help on the margin, but none of them are going to make or break your business. The rest of this post will explain why virality doesn’t work. A future post will explain why doing some ‘good enough’ things might still be a good idea.


Why Virality Doesn’t Work: It’s all about the funnel

I own my understanding of this concept to Tony Wright. Tony is a friend of mine and the founder of multiple start-ups including RescueTime. He is the best person I personally know at making things go viral. The best example of his experience is CubeDuel. CubeDuel was “Hot or Not for LinkedIn”. When you signed up for CubeDuel it pulled two people from your LinkedIn network and asked you who you would rather work with in the future. If you rated 100 people you could find out your own scores. But when you got to 100 it would almost always say, “Not enough people have rated you yet. Share with friends to get yourself rated.”

It had easy tools for sharing. It was fun to play (lots of pictures buzzing in and out as you made selections. Super fast UI). It had sharing incentives as well as incentives to engage more (“Rank 500 people to see the ratings of your network.” “Rate 1000 people to see the top employees at your company.”  “Rate 2000 people to see the top people in your city.” Etc.)

Tony was responsible for most of those features (the basic product was created by Adam Dopplet, co-founder of UrbanSpoon and Dwellable).

CubeDuel was so effective in the early days that it broke LinkedIn. Rumors spread that LinkedIn blocked it because it was too good and they wanted to create the product in-house, but the reality was they just spiked the use of the LinkedIn API so fast that it couldn’t keep up. It was fixed in a day or two.

But “breaking LinkedIn” made a great story. The media picked up on it. It was talked about all over as the ultimate viral product. Its growth was off the chart.

Only Tony shared the numbers with me. It actually wasn’t viral at all.

To understand that, we need a definition of viral.


How do we measure Virality?

The idea of a viral product is one in which your existing customers do your marketing for you. If every one of your customers gets two friends to use your product and each of those gets two of their friends and you can keep up that rate indefinitely you will soon have everyone on the planet using your product. It sounds pretty good.

In fact, it doesn’t even need to be that good for it to be infinitely good. As long as each customer gets >1 friends to use the product, you will eventually get everyone on the planet using it. The number of additional customers each customer acquires for you is called the “virality index”. If your virality index is >1.0 you have a viral product. If it’s less than 1.0 (but greater than zero) you have a product with a word-of-mouth channel that augments (but never replaces) your existing marketing.

On the third episode of “How to Start a Start-up” podcast (an excellent podcast for those who listen to those things), the founder of HomeJoy talks about how important small changes are. She uses the example of how if you can make small changes to get your virality index from 0.98 to 1.02 you have just made your product go viral. While that fact is true on its face, I hope to show you that your initial product does not have anything close to a 0.98 virality index.

The way to do that is to think about funnels.

Let’s focus on sharing your product through one particular channel: Email referrals (this same method applies to every channel, but this particular one has a long history of quantifiable metrics which makes the example more powerful).

Your product looks like this:

  1. You ask your users to share your product with friends
  2. Your users email their friends about the product
  3. Those friends read the email and click through to your ‘sales pitch’
  4. They read your pitch and register for your product

Voila. Virality.

If the numbers are high enough.

Let’s go through some benchmark numbers. We will be very optimistic.

  1. Let’s say you get 100% of your users to share your product with their friends (well done! Sounds like an amazing product!)
  2. Let’s say each of them shares your product with 100 of their friends they think would like your product
  3. Each of those friends receive an email. A very good email open rate is 20%. Let’s say because it’s coming from friend we get double that. 40% open rates.
  4. These potential customers read the email and some decide to click through to your full sales pitch. A great click-through-rate (CTR) is 5%. We double that again to 10%.
  5. They land on your pitch page. A great sign-up rate for a free product is about 10% (that would be ridiculously high for a paid product). Let’s say you get 20%.

That funnel looks ridiculously implausible. But let’s see what it’s virality  index is.

100% send rate x 100 contacts x 20% open rate x 10% CTR x 20% conversion rate = 0.4 Virality index

This means that every customer you sign-up will get an additional 0.4 customers for you (And those customers will get you an additional 0.4, who will get you an additional 0.4, etc.). It’s a pretty great scenario – when all is said and done for every customer you acquire with traditional channels you pick up an additional 0.66 customers. That should let you spend more on traditional marketing, but it sure isn’t going to be ‘viral’.

And remember we made up ridiculously unrealistic funnel metrics. More realistic funnel metrics would look something like this:

10% send rate x 50 contacts x 10% open rate x 5% CTR x 10% conversion rate = 0.0025 Virality index

Every thousand customers is getting you an additional 2.5 customers. Ouch.



So why does anything go viral?

I just proved it’s impossible for something to go viral. And yet Instagram, Farmville and Gangnam Style all exist. Have I just claimed they are impossible?

Kind of.

I’m claiming that trying to be the next Instagram or Farmville is a lot like trying to get hit by lightning a couple of times. Many people have been hit by lightning twice (in the same day or even seven times), but it’s pretty random. The same is true for going viral. Every metric will vary within a bell curve. Sometimes a metric will be very far to the right on a bell curve. When that happens we call it viral and start trying to rationalize why it happened.

You DO have influence on many of those metrics. You can make it easy to send out the email to someone’s entire address book; You can provide incentives to send it to both customer and potential customers; You can create compelling emails and landing pages that have high conversion rates. All of those things will make your product ‘more viral’, but in no world can it increase the expected virality function anywhere close to 1.0. For that you need luck.

Or specific types of products that can skip the funnel altogether.


Viral Products

A good way to identify a viral product is to see if it skips part of the funnel. If something about the product lets it get a 100% rate in a part of the funnel where normal (or great) products have a 20% rate, that’s a big difference. If a product can do that a few different parts of the funnel, it has a reasonable chance of actually going viral.

Let’s look at Farmville.

Farmville was the first in a series of games that took advantage of the Facebook platform to go viral. Let’s look at what their funnel might have looked like.

  1. 100%of users shared Farmville with their friends (built into the game)
  2. Initially they shared it with all of their facebook friends. Let’s say the average number of friends was 400 people
  3. At the time the feed was a feed of everything  happening in your network (Facebook wasn’t screening for the best stuff), so we can assume, sooner or later, 100% of your friends saw your Farmville update. In fact we can go further. Most friends would see your update multiple times, giving them many chances to respond. It was kind of like you were emailing your entire address book every day over and over and almost every email was getting opened. Let’s estimate that each person in your network saw your update an average of five times (this may be too low)
  4. CTR might be a little higher than average at the beginning until people understood what this was. Let’s say it was 5%.
  5. Let’s say sign-up to play Farmville was a standard 10%

What is Farmville’s virality coefficient?

100% x 400 x 5 x 5% x 10% = 10


Every user of Farmville adds ten additional users of Farmville!

Remember a great virality number is anything over 1.0. Farmville should have blown up (it did for a while), and then it should have kept blowing up until everyone who ever logged into Facebook was playing (And I know that is not true as I have never played a game of Farmville). At a 10x coefficient, assuming each ‘hop’ from one set of users to the next took a week, it would only take 9 weeks to have a billion users of Farmville. In 10 weeks therewould be more users of Farmville than there are people on the planet.

Assuming the funnel is right and Farmville is not 100% penetrated into Facebook users, what’s wrong with our math? We left out one factor: Inoculation.


Inoculating Viruses

There are always people who are immune to every virus. The same is true of viral products. No amount of awareness and exposure was going to get me to play Farmville. And as Farmville expanded the percent of people who were not interested became a higher and higher percentage of the people their “virality” was targeting. Its growth slowed down, stopped and eventually went into reverse. (But it sure was a good run while it lasted!)

Farmville had a very high inoculation rate. Even though it’s vitality ratio was off the chart, it’s inoculation rate served to slow down it’s exponential growth and hurt it’s (very) long term prospects.

It’s a function of the universe that every viral product will eventually hit an inoculated population, but some hit it much faster than others. Farmville’s inoculation index was particularly high. Not only was a large population immune, but many people who used the product dropped off. While some people became very addicted and played a lot (and shared a lot), the friends of the addicted group quickly became immune to the virus. And Farmville died.

Not every viral product looks like Farmville.

Consider Instagram.

Instagram had similar characteristics to Farmville. Every time a user took a cool photo with an Instagram filter and shared it with friends it was an advertisement for the product. The Instagram filters were shared with user’s friends over and over again. But unlike Farmville seeing different cool photos one after another from different friends was less likely to make you annoyed and more likely to make you interested in figuring out how YOU could make cool photos like that. The virality effect actually went UP over time. And unlike Farmville it was not a small number of power users who stuck with the product. Instagram had a very high customer retention rate, so even if the virality effect slowed down over time, net user growth stayed positive due to low levels of churn.



Which brings us back to CubeDuel. According to Tony, CubeDuel had a very high virality index of about 0.7. That was enough that every time they signed up a user from a traditional marketing channel they picked up another 2.25 additional users through viral spreading. That is pretty awesome. But it only works as long as you continue to prime the pumps with new users. You can’t just sit back and let the virality take over.

The next piece of good news was that CubeDuel was getting lots of media coverage (remember “The Product That Broke LinkedIn”?). Coverage begets coverage.  After the story of breaking LinkedIn there were a series of stories about how viral the product was (Little did the reporters know it wasn’t actually viral at all!).

Tony and Adam saw the writing on the wall. They knew the product would take off as the 0.7 virality function did its work to build on the press coverage, but they also knew that as the press coverage died down eventually the traffic will dive down too. Their product was great and fun to use, but it wasn’t fun to use every day. After playing with the product for a while, users were ready to move on.

Tony and Adam sold CubeDuel to investors who were more optimistic they could continue to make the product go viral.

Today? redirects to

So irrational exuperence in virality is at least getting a little more traffic to a respected charity.

Unless that is the plan for your business, focus on more traditional marketing channels and don’t count on virality to make your business plan.

Guest Post: Branded thieves and non-branded Samaritans


David Nefs

Since launching this blog a year ago I’ve received a number of requests to guest post. I’ve turned them all down. The goal of this site has been to share my (I think) unique take on marketing and analytics and I wasn’t interested in diluting that. Since I’m not an academic, professional writer or consultant, this has meant I just publish whenever I find the time. The exercise I think has been a success. I’ve gone from not having a Twitter presence to being a Top-5 CMO on the platform. And I’ve had a number of interesting connections (and at least two job offers) from people who have found the blog.

Today I am finally branching out into Guest posts. David Nefs (@davidnefs) reached out to me with some questions on incrementality and he shared some of his own findings. I thought the findings were interesting enough (and well written enough) that I asked him if he was interested in having them published on the blog. He was, and this is that post.

David is an internet and marketing enthusiast and pursues both these interests working for HomeAway, the world’s leading vacation rental marketplace. Prior to HomeAway, David was a management consultant for Bain & Company and began his career working for Groupon. David loves travel, having toured extensively through Europe, Asia and South America, and his hobbies include running and struggling to learn Mandarin. Currently he lives in London, the home of his beloved West Ham United FC, of which he is a long suffering fan. David graduated from the University of Cambridge with a 1st class degree in Economics and an MPhil degree in Industry and Management.

A common fear of many executives is that spending on branded terms is just throwing money down Google’s drain… ‘why are we spending money on traffic we would have got anyway?!’ you’ll hear them exclaim.

This risk isn’t imagined. A couple of years back a study by eBay concluded brand terms were largely ineffective. For companies that can’t commission a 26 page research paper AdWords’ Paid & Organic report sheds some light on this dilemma. By showing CTRs for queries where organic results were displayed both with and without paid ads, we can get a view of the impact of PPC on SEO.

How it works

The Paid & Organic report in AdWords creates the output below for all queries where your site appears in Google showing you performance when you have Paid only, Organic only or Paid & Organic results showing for a given query.

By classifying each query as ‘brand’ or ‘non-brand’ and then analysing how CTRs change dependent on the results displayed we can create a controlled experiment on the impact of PPC ads on SEO and form a view of the incrementality of PPC visits.

Incrementality chart

Snapshot of AdWords’ Paid & Organic report

Note: It is often necessary (especially for non-brand queries) to control for average SEO position in this analysis as this can heavily influence Organic CTR.

Brand Queries

Results shown Organic CTR Paid CTR Total CTR Av SEO pos Av Ad pos
Organic only 55% 55% 1.03
Paid & Organic 36% 53% 89% 1.01 1.02
Delta -19ppts +53ppts +34ppts

Paid and Organic report results for brand queries

In the above example for brand queries we can see that when the organic result is shown independently it has a 55% CTR and when it is shown alongside a paid result CTR falls to 36%, evidence of PPC ‘cannibalising’ SEO. However, when both paid and organic results are shown there is a lift in total CTR to 89% meaning PPC generates an incremental 34 visits for every 100 brand queries and is not completely cannibalistic. To quantify, 34 incremental visits are created for every 53 paid visits meaning the paid traffic is 64% incremental.

For the 10+ brands I have looked at the incrementally of branded paid search varies between 16% and 73% and is around 50% on average. The practical implication of this for most brands is to continue to spend on brand terms as they generate incremental traffic cheaply enough to be (often highly) profitable .

Non-brand Queries

Results shown Organic CTR Paid CTR Total CTR Av SEO pos Av Ad pos
Organic only 4.6% 4.6% 7.67
Paid & Organic 6.3% 11.9% 18.2% 7.11 3.87
Delta +1.7ppts +11.9ppts +13.6ppts

Paid and Organic report results for non-brand queries

So, how does this picture look for non-brand queries? Here we see that organic CTR actually increases in the presence of a paid ad and 13.6 incremental visits are created for every 11.9 paid visits meaning the paid traffic is 114% incremental! In other words, paid is generating organic traffic you would otherwise not have realized. Often lip service is paid to the ‘branding’ effect of PPC and how taking up more of the SERP real estate improves organic CTR but it is fascinating to see this bear out in hard data.

For the brands I have analysed the incrementality of non-brand paid traffic varies between 93% and 134% and is around 115% on average, which means your non-brand spend may well be a little more profitable than it first appears.

Comments welcome.


Two week ago I wrote about the importance of Marginal Value when thinking about marketing. Last week I expanded on the concept by looking at the impact of fixed and variable costs when you are calculating your true marginal costs of incremental marketing. Today we will flip the story from costs to revenue and try to measure incrementality.

Incrementality is fundamentally tied with attribution (in fact I cover it briefly in my book chapter on Attribution). Once you attribute a customer (or revenue) to a channel you need to realize that you may not have got it exactly right. If you got it 100% right then it is 100% incremental. It is possible in theory, but it rarely happens in practice. The next step is estimating how incremental you believe the revenue is.

I like to think of incrementality as being a percentage. If a channel is 100% incremental then you know that every customer you get from that channel is an additional customer to your business. One way of thinking about it is if you turned that marketing channel off, you would lose all of those customers.

If a channel is 50% incremental, it means half the customers you are getting from that channel you would have got anyway from a different channel. If you turned the channel off you would only lose 50% of the customers you are claiming to that channel. When a channel has <100% incrementality we say it is cannibalizing other channels (i.e., when it grows, part of its growth is incremental and part of it is stealing – cannibalizing – other channels)

If a channel is 200% incremental, it means that for every customer that channel generates, it generates an additional customer in a different channel. If you turned a 200% incremental channel off you would lose all of that channel’s customers, but you would also lose the same number again in other channels. When a channel has >100% incrementality it is synergistic (i.e., when it grows it also grows the other marketing channels around it).

Different marketing channels have fundamentally different incrementality rates.

An example of a very incremental channel is Brand TV advertising. A sophisticated business will estimate the customers generated from brand advertising. They usually look at when the TV spot airs and correlate it with website traffic  that spikes at the same time. They create some sort of decay curve on that spike so that they still give the TV spot credit for those web customers hours (or even days) later. But even with these methods, the attribution fundamentally under-values the TV advertising. Let’s make a short list of the impact of that TV spot that you are not giving it credit for:

  • Web traffic that comes to the website weeks or months later (GoDaddy has a Superbowl ad. They see a traffic boost for years as their recognition goes from zero to the stratosphere)
  •  Non-web customers that you can’t tie back to TV
  • Improved CTR on your paid search ads (Which drives up quality score, and decreases your CPC)
  • Better SERP results for the SEO (as Google sees more type-in traffic and increases your Panda score)
  • Business Development opportunities – people reaching out to you
  • Higher conversion rate on recruitment (and people coming to you)
  • Improved B2B relationships (or leverage in negotiations)

I’m sure you could think of other effects of TV. I’ve rarely seen a company try and model all these effects. Instead they measure a few of them and then accept that there are significant synergies that they cannot fully measure. Those synergies are >100% incrementality.



The reverse of television advertising is coupon sites. Have you ever got to  the end of a checkout and seen a field: “Enter a Coupon Code”. What?!? I can get a discount?!?

The first thing I do is go back to Google and search for “Company Name + coupon” or +discount or +code. Every now and then you find a coupon site. You click on the coupon site, get a code and enter into your checkout. You have just given that coupon company attribution for the sale. But it is very clear you would have purchased anyway without the coupon. The coupon site just cannibalized the original channel that brought you to the site.


Every marketing channel is cannibalistic sometimes. Even a super-synergistic channel like television will have cannibalization. Some people who see the ad and act on it (and you measure them – say with a phone number on the TV spot), would have come to you anyway, through one of your other marketing channels. In the case of TV, the synergistic effects far outweigh the cannibalization effect.

Every channel is synergistic sometimes. Someone may be shopping on the coupon site. They see your company but forget to copy the coupon. The next day they come to your site through Direct Type-In and convert. You give the credit to DTI, but the real driver was the coupon site. The coupon site was synergistic in that particular case, but overall it is still a cannibalistic channel.


When Cannibals Matter

If you only had one marketing channel, then you don’t need to worry about cannibalization. If all your marketing is free, you also don’t need to worry about cannibalization. Cannibalization matters in three cases:

1. When a marketing channel isn’t profitable, but you think it is

Coupon sites are a very good example of this. Let’s say the coupon site is only 20% incremental (i.e., 80% of those customers you would have got anyway if you didn’t have a presence on the coupon site). Let’s say you are paying the coupon site $20 for every $100 they generate. And let’s say the coupon is a 10% off coupon. (And your business is 100% fixed costs – so 100% margin on that $100).

Spending $20 to make $90 ($100 – 10%) sounds great. Until you look at the incrementality. If the coupon site is only 20% incremental, then $80 out of the $100 is revenue you would have got anyway, even if the coupon site did not exist. So you are really spending $20 to make $20. But then you are giving a $10 discount. So you are spending $20 to make $10. You can’t make that up in volume…

These situations happen all the time. (It’s a related story to the selection effect in Loyalty Programs I wrote about a month ago)

2. When a more expensive channel cannibalizes a cheaper channel

Imagine Amazon had 100% market share of all book sales online. All of the customers were coming to Amazon through Google Paid Search. Amazon was profitable, but they were giving away dollars to Google (Let’s say $0.50 of every $1 they make in profit). Then Amazon introduces an email program. The program is 0% incremental. It is completely cannibalistic. All it does is steal sales from the paid search channel and give it to the email channel.

But that’s okay.

Every $1 sale that moves from paid search to email is saving Amazon $0.50. It doesn’t matter that it’s cannibalistic, as long as the cheaper channel is cannibalizing from the more expensive channel


3. When You aren’t spending enough due to attribution

Using the last example, once Amazon launches email they may start seeing Paid Search as less effective. There are two scenarios on how email could be cannibalizing paid search.

In the simple scenario once email launches, people stop searching and come direct to Amazon. In that case #2 applies.

A more complicated story is that people still search, but then Amazon hits them with an email. When they buy Amazon attributes the sale to Email.

In both scenarios email has cannibalized paid search, but in fundamentally different ways. In the first example, email did all the work. If search did not exist it wouldn’t matter one iota. In the second scenario it’s hard to say which channel did the most work. But Amazon is just choosing to attribute the sale to email. When that happens the rational thing for the paid search manager to do is to reduce Amazon’s Paid Search spending (since Amazon had to pay for that click, but paid search got no credit for the sale)

If it turns out that the customer needed both channels to convert, you need to be careful when the cheaper channel cannibalizes credit from the more expensive channel.


Next Week: I will put attribution to the side for week or two while we dive into how customer life time value fits into all this.

Fixed and Variable Costs

Last week I shared some thoughts on marginal unit economics. It is THE fundamental thinking needed to do effective marketing. In order to understand what your marginal unit costs are you need to understand what part of your costs are fixed and what parts are variable.

It’s a tougher problem than the textbooks make out to be.

Let’s use a restaurant as the example.

Some restaurant costs are obviously fixed:

  • Rent
  • Owners’ time
  • Manager cost
  • Furniture leasing costs
  • Menu costs

The most obvious variable cost is the cost of goods for the food eaten by the marginal customer. That cost could become fixed if the food was about to go bad, but let’s assume the restaurant is better managed than that.

The biggest question is the cost of the servers and cooks. The first server and cook are pretty fixed. You need someone there to make and serve the food even if no customer walks in the door (unless you have the manager or owner do it). In theory any staff beyond the first (for a specific role) is variable depending on how many customers you expect to have.

Basically you have a step-change function. The first two staff can handle x number of customers, but once you get to a specific point you need more staff. There is a fuzzy zone where you  could have one server who is scrambling or you could have two servers who are relaxing. That fuzzy zone decreases in size the bigger the restaurant but likely never really goes away.

After you have staffed your night, then the marginal staff cost for any given customer is zero.

So are staff costs fixed or variable?

It depends.

If you are talking about bringing in one additional table of customers it is almost definitely fixed. If you are talking about bringing in 1000 additional tables of customers they are definitely variable.


“In the short term all costs are fixed. In the long term all costs are variable.”

We saw a similar dynamic at A Place For Mom. We would source leads for our senior living advisors (SLAs) who would work with families to help them find senior housing. Overall we knew that SLAs could handle about xx leads per week. If they got more than that on a consistent basis then we saw our overall conversion rate start to slowly decrease. However, we also knew that the SLAs could handle almost twice the normal volume in a given week without any loss of overall performance.

So on a given week we had no marginal operations costs from sending an additional lead. But for planning purposes, the operations costs of handling leads were fully variable.

I believe staffing at restaurants is similar.

It was a complicated problem for a bunch of Harvard/Wharton/Stanford/McKinsey/Bain/BCG MBAs to solve at A Place For Mom. One can’t expect an owner-chef at a restaurant to figure out the math.

But without the math it is impossible to make rational marketing decisions.


What are your marginal costs?

Simplify it.

You know your costs that are definitely variable (COGS). You have a bunch more costs that are short-term fixed, long-term variable. Figure out their marginal cost to serve an additional customer. Add those two numbers together. You now have a range on what your marginal costs are.

For restaurants I have spoken to this works out to 30-60% of their revenue (30% is the cost of the food +30% for the cost of the staff if the staff cost were 100% variable). It means that if the restaurant rebates more than 70% they will be losing money no matter what they do (with some exceptions coming later). If the restaurant rebates 40-70% they may or may not be in trouble depending on how many customers they give that big a rebate to.

Less than 30% they will be making money on the margin – even if the total margin of the company is only 5%.

And that is the key insight.

I have spoken to many businesses who say, “My total margin at the end of the year is only 10%, so if I give a 10% discount I make nothing.”

That is true if you gave a 10% discount to every single customer. But it is not true at all if you give a 10% discount to incremental customers.


Let’s see it in a chart:

Metric Base Business
Customers 10,000
Revenue per Customer $100
Total Revenue $1,000,000
Variable cost per customer $50
Total Variable Costs $500,000
Fixed Costs $400,000
Total Costs $900,000
Total Profit $1,000,000 – $900,000 = $100,000 (10%)


Now what happens if he gets an additional 1000 customers at a 10% discount:

Metric Base Business Extra Customers TOTAL
Customers 10,000 1,000 11,000
Revenue per Customer $100 $90 $99.09
Total Revenue $1,000,000 $90,000 $1,090,000
Variable cost per customer $50 $50 $50
Total Variable Costs $500,000 $50,000 $550,000
Fixed Costs $400,000 n/a $400,000
Total Costs $900,000 $50,000 $950,000
Total Profit $100,000 $40,000 $1,090,000 – $950,000 = $140,000 (12.8%)


See what happened there? Even though we gave a 10% discount (our ENTIRE MARGIN!) to these new customers, we ended up growing our total margin from 10% to 12.8%. It’s magic.

As long as you can acquire customers for less than the cost to service those customers your business’s margin dollars will increase, regardless of the percentages. If you acquire customers at a higher marginal margin your overall margin percent will increase – even if those marginal customers have a lower total margin than your existing customers (in the above example, you could discount 45% on those incremental customers and your 10%

It’s a little mind-warping, but there is no more important concept to understand in marketing a business. And it’s a way of thinking that even many Fortune 500 CMOs don’t seem to fully internalize (in fact CFOs tend to get it more often than CMOs in my experience)

One (very very important) caveat: This only works if those new customers are truly incremental. If they aren’t it’s called cannibalization and the whole house of cards can come tumbling down.

Up next: Cannibals!

Marginal Value

When I taught my first MBA course I knew that the students would not remember everything I taught them. I also knew that whatever I taught first (and returned to throughout the course) would be internalized more than anything else. So I thought long and hard about what the first thing should be.

I landed on Marginal Value.

It’s a relatively simple concept: Instead of looking at averages, you look at the costs and benefits of an incremental decision. If the benefits outweigh the costs then that action is worth taking. If they don’t then you should not take that action.

As simple as it is, it is ignored by many many marketers.

Consider two very different businesses:

A retailer sells physical goods. Say they sell only one product: Product A. They sell it for $100 and it costs them $70 to buy it. Their margin is $30. If they discount Product A more than $30 (30%) they will be losing money on each sale. Now $30/sale may not be enough to keep them in business (they need to pay for their staff, the real estate rent, marketing, overhead like accounting, etc.), but if they sell enough they will be fine. But no amount of sales at $69 will keep them in business. In fact, the more they sell the more they lose. It’s the classic joke of “Sure we lose money on every sale, but we will make it up in volume.” More volume only works if your marginal benefits are higher than your marginal costs.

But a discount less than $30 could work.

If they sell 1000 units at $100, that’s the same profit as selling 2000 units at $85. If a 15% discount more than doubles their sales, then they should absolutely do that. But if they discount 29%, then they need to sell 30,000 units. That seems unlikely…

A lot of retail is like this – or worse. The marginal margin on products after their cost is very low. It doesn’t leave much room for discounting. And when you do discount, the lift you need to see if so high that it is unlikely to be the right choice. Do we really thing they will double their sales with a 15% discount? (and many products have much lower margins than 30%).


The second extreme example is an airline.

Airlines have huge fixed costs. They need to pay fees to the airports; they need to pay for their planes; they need to pay for the maintenance on their planes. In theory each flight is a variable cost: If they don’t make the flight, they don’t pay for the pilot or the stewards or the fuel (and likely a lot of maintenance as well). In practice they are committed to a certain flight schedule (and given the huge fixed costs of owning the planes it makes a ton of sense to keep them in the air as much as possible).

Now that we assume the flight is taking off, the marginal costs of seating a passenger is really really low. A little more fuel required to lift the passenger and their luggage. The free soda or peanuts provided. It takes a little longer to board which may increase the costs for the gate staff? (That last one is really reaching). I would be surprised if the marginal cost of an additional traveler is more than $20 on a $200 flight.

This means that an airline could give much larger discounts than a retailer and still make the business work. Let’s say a flight is half-full with 100 people at $200/seat. If the airline were to discount the seats by 15% to double the sales (same effect as we showed for the retailer) they would go from making $18,000 after marginal costs ($200 x 100 seats = $20,000 revenue – $2000 costs) to $30,000 ($170 x 200 seats = $34,000 – $4000 costs). That’s a lot better than the break even scenario of the retailer for the same discount.

Airlines obviously know this, which is why their pricing schemes are so much more complicated than retailers (imagine if Old Navy T-shirt prices changes as often as airline tickets?).

Airlines and retailers are an extreme examples. Most businesses sit in-between the two. But most businesses don’t automatically think about marginal economics and so make lots of mistakes.

Retailers discount far more than they should.

Restaurants don’t discount nearly as much as they should.

Next: Fixed and Variable Costs