The Negative Binomial Distribution in Marketing: The Customer Demand Curve Explained

Definition

The negative binomial distribution (NBD) is a probability distribution that models how often individual customers buy a brand over a given period. In marketing it produces a characteristic curve, a tall head of infrequent light buyers and a long tail of frequent heavy buyers, that is remarkably consistent across categories. Patrick Gilbert calls it the Customer Demand Curve because the technical name has never once made anyone lean forward in their chair.

The Name Nobody Uses

There is a concept from marketing science that explains more about how brands actually grow than almost anything else, and almost nobody talks about it. Its real name is the negative binomial distribution, and that term has never once made someone lean forward in their chair. In Never Always, Never Never, Patrick Gilbert calls it the Customer Demand Curve, because that is what it really is: a picture of how your customers are spread out by how often they buy.

The marketing scientist Andrew Ehrenberg spent decades documenting this pattern across hundreds of product categories. The finding was almost boring in its consistency: plot any brand's customers by purchase frequency and you get the same shape nearly every time. A tall head of people who buy once or twice, and a long, thin tail of people who buy a lot.

What the Curve Actually Shows

Consider Coca-Cola. More than half of Coke customers buy between one and two cans per year. As purchase frequency increases, the number of customers drops off toward a long tail. A tiny handful of people buy Coke dozens of times a year, but they are the extreme outliers.

According to the data, the average Coca-Cola buyer purchases about 12 times a year, roughly once a month. That sounds perfectly reasonable. But the average Coke buyer is not typical. Roughly 60% of Coke customers buy six or fewer times per year. That average of 12 is dragged upward by the small number of heavy buyers on the tail.

Why the Average Lies

Averages are misleading, and they are often the reason marketing strategies fail. The place where it is most dangerous is calculating and forecasting customer lifetime value.

Imagine a subscription business selling virtual yoga classes for back pain. Most subscribers cancel after a few months once they feel better. A few stay for years. If the company calculates an average retention of, say, 24 months, that number is pulled way up by the long-tail loyalists, even though the vast majority only needed three months. Plan your acquisition spend around that inflated average and you will overpay for every single customer.

This is the exact mistake many brands made during the digital arbitrage era. Blue Apron told investors before its 2017 IPO that it was fine to acquire customers at $463 each, because loyal customers were worth more than $900 over three years. But more than 60% of customers cancelled after the first month, and only 18% lasted a full year. It was never feasible to spend $463 per customer. Blue Apron's stock lost 99% of its value within two years.

If you rely on averages without understanding the curve underneath them, this is the trap you fall into.

Patrick Gilbert, Never Always, Never Never

Enjoying this? Never Always, Never Never goes much deeper into the mental models and decision frameworks that shape how we think.

Get the Book on AmazonLearn more about the book

Advertising as Probability Management

If most of your customers are light buyers, and you cannot trust the average to model their value, why market to them at all? Because they are the ones your advertising can actually move.

For the average Coke drinker, there is roughly a 1-in-300 chance they will buy a Coke on any given day. Over a year, that adds up to about one purchase. Coca-Cola spends billions on advertising not to make everyone crave Coke daily, but to nudge that probability from 1-in-300 to slightly more than 1-in-300. Push it to 2-in-300 and you double revenue, and most consumers would never notice the change in their own behavior.

Heavy buyers are already buying at or near their ceiling. They are less responsive to advertising and less likely to change behavior. Light buyers are the opposite, and research from the Ehrenberg-Bass Institute confirms they account for 40 to 50 percent of a brand's total revenue across hundreds of categories.

How Growth Reshapes the Curve

Here is the part most people miss. Reaching more light buyers does not just add revenue, it changes the shape of the entire curve.

Say marketing helps you grow your customer base by 50%. You do not simply stack more one-time buyers on the left side of the chart. The whole curve shifts up and to the right. The head gets taller because you are acquiring more light buyers, but the tail also gets longer. You are rolling the dice more times, and a small percentage of those new customers will eventually become heavy buyers. The absolute number of heavy buyers grows, even though they remain a small proportion of the total.

Loyalty Is a Habit, Not a Conviction

We tend to think of loyalty as an emotional commitment to brands we have carefully chosen. That is not what the data shows. Most repeat buying is habitual: you grab the same coffee creamer every couple of weeks without being able to name the brand, because it is there, it is familiar, and thinking about creamer is not worth the brainpower.

This is why the double jeopardy law holds and why loyalty tracks so closely with market share. If loyalty is mostly a habit, then building it is not about making people feel something. It is about mental availability and physical availability: being easy to notice, easy to remember, and easy to buy, so that when the rare moment of decision arrives, choosing you is the path of least resistance.

• Read the distribution before you trust any average, especially for retention and lifetime value
• Model acquisition spend from the curve, not a blended average that the tail has inflated
• Aim advertising at reach across light buyers, the group whose behavior you can actually influence
• Grow the base to grow the tail, rather than trying to squeeze more from existing heavy buyers
• Build mental availability and physical availability so habitual choice defaults to you

Key People & Works

Practical Applications

• Stop planning around the average customer, and look at the shape of the purchase frequency curve instead
• Model customer lifetime value from the distribution, not a single blended average, so you don't overspend to acquire customers
• Treat advertising as a probability nudge across a broad base of light buyers rather than a conversion trigger for heavy ones
• Grow the whole curve by acquiring more light buyers, which lifts the number of heavy buyers over time
• Budget for reach that captures occasional buyers rather than frequency aimed at people already buying at their ceiling