Marketing — the Atlas · ch.05 · how brands grow
📣 Chapter 5 · Part I · The battle for the mind

How brands actually grow

In 2010 Byron Sharp published buying-panel data that embarrassed half the marketing canon. This chapter is the physics lesson — what decades of purchase records say brands can and cannot do.

Here's the whole chapter in one line: brands grow by being easy to think of and easy to buy, for as many category buyers as possible — mostly the ones who barely think of you at all. Everything below is that sentence, backed by fifty years of purchase records.

1Marketing gets a profiler

For its first century, marketing ran on anecdote. Award reels, gut feel, the highest-paid person's favorite campaign story. Everyone had a theory about why brands grew; nobody had run the numbers, because there were no numbers to run.

Then came consumer panels: tens of thousands of households recording every purchase, week after week, for years. Andrew Ehrenberg — a statistician, not an ad man — got hold of them in the 1950s and did the unglamorous thing: he fit distributions. Panel data is the profiler you finally ran on marketing's hot loop. For decades everyone argued about where the time went. Now there was a flame graph.

What the profiler showed was offensive in its boredom. Buying behavior follows regularities — the same mathematical shape (a model called the NBD-Dirichlet) fits category after category, country after country, decade after decade. Colas, banks, detergents, gasoline: different products, same curves. These are laws in the physicist's sense — not deep truths about the soul, just patterns that keep showing up no matter who's looking.

Ehrenberg published for forty years and adland mostly shrugged. Then in 2010 Byron Sharp of the Ehrenberg-Bass Institute compressed the whole program into How Brands Grow — same math, ruthless prose — and it landed on the CMO shelf. Its sequel with Jenni Romaniuk extended the evidence to emerging markets, services, and luxury. P&G and Unilever rebuilt their media plans around it. The shrug became doctrine.

"Repeat buying follows a law."
Andrew Ehrenberg · NBD-Dirichlet · statistics journals
the moveFit one distribution family to panel after panel — buying behaves like arrival noise, not persuasion.
the findingLoyalty metrics are predictable from market share alone.
receptionAdland shrugs — too boring to be true.
why it matteredLaws you can predict with beat stories you can only retell.
"Marketing's first physics textbook."
Byron Sharp · Ehrenberg-Bass Institute · 2010 (Part 2 with Jenni Romaniuk)
the moveSame math, ruthless prose, aimed straight at the CMO shelf.
the findingGrowth = penetration; loyalty mostly comes along for the ride.
receptionDoctrine at P&G and Unilever — media plans rebuilt for reach.
why it stuckIt made predictions, and the predictions kept coming true.

2Double jeopardy

The first law out of the panels is the one with the courtroom name. Small brands get hit twice: fewer buyers, and those buyers are slightly less loyal — that's the whole scandal of double jeopardy. Not less loyal because small brands are worse. Less loyal because they're small.

Plot every brand in a category — penetration on one axis, loyalty (say, purchases per buyer per year) on the other — and the dots don't scatter. They hug a single rising curve. The big brand gets more buyers and a little more frequency; the small brand gets fewer of both. Every category reproduces the same picture.

The uncomfortable translation: loyalty is mostly a computed column, not an input. It's derived from size. Marketing departments keep hiring people to turn a dial that isn't wired to anything — you can't hold penetration fixed and crank loyalty, any more than you can raise a cache's hit rate while refusing to give it more entries.

Below is the scatter every category keeps reproducing. Try to build the brand every founder pitches — tiny but beloved — and watch where the data lets you actually stand.

Interactive · the double-jeopardy line Drag any dot along the axis · tap the ghost
Every brand in cola hugs the same rising curve: more buyers ⇒ slightly more loyal buyers. Drag any dot horizontally and try to escape it.
📈
Evidence check. The pattern was first named by sociologist William McPhee in 1963 (he found it in radio audiences), and Ehrenberg then documented it in shopping baskets. It has since shown up in dozens of categories across dozens of countries — packaged goods, banks, gas stations, prescription drugs, B2B service contracts, even political donations. When a "law" survives that many attempted counterexamples, you stop treating exceptions as strategy and start treating them as measurement error.

3The heavy-buyer illusion

Every marketing plan eventually rediscovers the same seductive idea: find your best customers and get more out of them. The 80/20 rule says 20% of buyers drive 80% of revenue — so aim everything at the 20%. There are two problems, and the panels expose both.

First, Pareto is real but mild. In actual purchase data the top 20% of a brand's buyers deliver around 50–60% of volume — not 80%. Which means the bottom 80% — the people who buy you once or twice a year and couldn't pick your logo out of a lineup — deliver the other half of your revenue. The customers you never think about are half your business.

Second, heavy buyers are already maxed out. Someone buying you every week has nowhere to go. Worse, this year's heavies are partly this year's lucky draws — next year they regress to the mean and buy less, through no failure of your marketing. Building a growth plan on heavy buyers is optimizing a function that's already at its ceiling while ignoring the enormous flat tail where all the headroom lives.

Here's the tail. Hover the bars, then try both strategies and watch where three years of effort actually compounds.

Interactive · where growth lives Hover a bar · then run each strategy
squeeze heavies · not run recruit lights · not run
A year of buying for one brand: 42% of the category bought you zero times, and the bars fall off fast. Hover any bar for its share of people and of revenue.
📈
Evidence check. Across the categories Sharp and Romaniuk audited, the top-20% share of volume averaged in the mid-50s — a "60/20 rule" at best. And heavy-buyer panels re-measured a year later show the regression effect directly: the heaviest slice always lightens, the zero-buyer slice always produces new customers. The buyer base isn't a fixed org chart; it's a distribution that keeps resampling itself.

4Growth = penetration

Put the two previous laws together and the arithmetic of growth only has one direction left. If loyalty is chained to size, and the heavies are at their ceiling, then brands grow by recruiting more buyers — overwhelmingly light ones — not by squeezing more out of the buyers they already have.

The panels confirm it with brutal consistency. Compare growing brands to shrinking ones and nearly the entire difference is how many people bought them at all. Purchase frequency barely moves. When a brand doubles, it doesn't have buyers buying twice as often; it has roughly twice as many buyers, most of them occasional. Growth looks like a wider, shallower pool — never a deeper puddle.

Which reframes two decades of loyalty-industrial complex. A loyalty program mostly pays people for behavior that was already going to happen: your heaviest buyers sign up first, collect points on purchases they'd have made anyway, and the incremental purchases are a rounding error. That can still be rational — as a pricing instrument, a selective discount to your base. It just isn't where growth comes from, and the budget line should say so.

⚠️
Don't overcorrect. Retention still matters operationally. In a subscription business the leaky bucket is the whole P&L — churn compounds against you, and fixing a broken cancel-flow can be worth more than any campaign. The law isn't "ignore existing customers"; it's narrower and sharper: growth comes from new and light buyers, so don't book loyalty spend as growth spend. Plug leaks because leaks are expensive, not because plugging them will make you bigger.
"AAdvantage takes off."
American Airlines · 1981
the movePoints for repeat flying — the first mega loyalty program.
the beliefRetention is cheaper than acquisition, so loyalty = growth.
the effect"Loyalty" becomes marketing's favorite word for two decades.
the catchMostly rewarded flights that were going to happen anyway.
"Cap the frequency. Chase the reach."
P&G, Unilever & the biggest ad buyers · 2026
the moveFrequency caps, incremental-reach curves, light-buyer targeting.
the beliefGrowth comes from the next buyer, not the tenth purchase.
the effectLoyalty programs refiled under pricing, not growth.
the catchNone of it excuses a leaky product — churn still runs the P&L.

5You share your buyers

One more law, and it's the one that quietly kills the tribal fantasy. The duplication of purchase law: your customers buy competing brands roughly in proportion to those competitors' market shares. Coke's buyers also buy Pepsi — a lot of them. Pepsi's buyers also buy Coke — even more of them, because Coke is bigger. Nobody's base is a walled garden.

Put differently: "Coke buyers" are mostly "cola buyers who buy Coke a bit more often." The brand doesn't own a tribe; it owns a slightly larger slice of everyone's repertoire. People keep a handful of acceptable brands per category and rotate through them with the enthusiasm of someone picking a parking spot.

Programmer's version: your customer base isn't a private table. It's a view over the whole category, weighted by share. Query "our customers" and "their customers" and you get almost the same rows with different weights. Any strategy that assumes your rows are a different species — different values, different psychology, a different "tribe" — is doing analytics on a join artifact.

💡
This is why rival brands' buyer profiles look eerily identical in survey after survey — BMW and Mercedes buyers, Coke and Pepsi buyers, iPhone and Samsung buyers differ from each other far less than either differs from non-buyers of the category. The battle isn't for a special kind of person. It's for a bigger share of the same people's very divided attention.

6The two levers

So growth means recruiting light buyers you don't control, can't identify in advance, and who think about you almost never. What could possibly move that? The evidence keeps loading onto exactly two levers.

Mental availability: the probability you come to mind in a buying situation. This is Chapter 1's ladder, operationalized — not one rung in one list, but wiring to many situations. Thirst, road trip, guests coming over, 3pm slump: each cue is a separate retrieval path, and the brand that's linked to more of them gets "thought of" more often without anyone feeling persuaded.

Physical availability: the probability you can actually be bought when the situation fires. Distribution, shelf space, delivery coverage — and their digital descendants: search rank, app-store placement, being in stock, a checkout that doesn't fight back. Every step between impulse and receipt is physical availability, whether it happens in a store aisle or a tap target.

The crucial fact is that they multiply, not add. An occasion where you're remembered but not findable is lost. An occasion where you're stocked but unthought-of is lost. Nearly everything in marketing that measurably works — advertising, packaging, distribution deals, SEO — works because it loads one of these two levers. Slide them and watch the arithmetic.

Interactive · the two levers Each square is a buying occasion · you win only where both levers reach
40%
40%
both — you win it in mind only on shelf only neither
Mental 40% × physical 40% ⇒ you win ≈ 16% of buying occasions. You only win where both are true.

7Evidence check — the pushback

Every doctrine deserves its stress test, and this one has real boundary conditions. Here's where it bends.

Subscriptions. The laws were minted on repertoire categories — many small, low-stakes purchases. A subscription is one decision that then repeats by default, so churn math gets real weight: a lost subscriber is a lost annuity, and retention work has direct, computable value. Penetration still rules acquisition, but the leaky bucket is no longer a footnote — it's a term in the equation.

Luxury. When scarcity is the product, mass penetration can dilute what you're selling. Yet even here the doctrine half-survives: luxury houses advertise far beyond the people who will ever buy, because a Birkin only works if the millions who can't afford one know exactly what it is. The audience is broad even when the customer list is short.

Durables and tiny B2B universes. Cars and CRMs sell on decade-long cycles to buyers who enter the market rarely — panels thin out, and when your total addressable market is 94 procurement teams, "penetration" needs air quotes. But the logic survives translation: reach everyone in the universe, stay remembered across the long silent gap.

And notice what the critics concede. Reach beats precision-targeting surprisingly often, even in performance-marketing's own attribution data. Distinctiveness compounds where clever repositioning decays. Going dark hurts on a lag. The boundary conditions adjust the coefficients; they haven't overturned a law yet.

📈
The 95:5 rule. The LinkedIn B2B Institute ran the Ehrenberg-Bass playbook on business buying and landed on a blunt heuristic: at any moment, roughly 95% of your potential B2B buyers are not in the market — no active deal, no budget line, nothing this quarter. The instinct is to ignore them and chase the in-market 5%. The evidence says advertise to the 95% anyway: they're the ones whose memory you can still shape cheaply, and when they finally enter the market, the shortlist was written years earlier. Brand advertising is a cache-warming strategy for demand that doesn't exist yet.

8Monday morning

The satisfying thing about laws is that they compile down to a to-do list. If the panel data is right, Monday looks like this:

  • Maximize reach within the category. Talk to all category buyers, especially the light ones you're tempted to ignore — that's where the headroom is. Sophisticated targeting that quietly narrows your audience is a tax on growth dressed up as efficiency.
  • Never go dark. Memory decays on its own schedule, not your fiscal calendar. Continuous presence beats brilliant bursts separated by silence.
  • Build distinctive assets. Reach only compounds if people can tell it was you. Colors, shapes, sounds, characters — the next chapter is entirely about this.
  • Fight for physical availability. Every missing retailer, every page-two search rank, every extra checkout step is share you handed to whoever was easier to buy.
  • File loyalty schemes under pricing. Run them if the discount math works. Just stop booking them as growth.

One thread is left hanging. Mental availability means being retrieved from memory — but retrieved as what? Not as a positioning statement; nobody's brain stores your value proposition. What the mind actually keeps is embarrassingly concrete: a shade of orange, a bottle silhouette, three notes of a jingle, a lizard with an accent. Those assets are how reach becomes memory — and they're where we go next.

Check yourself

3 questions · instant feedback 0 / 3