Why Your Best Customers Don't Stay Best

Why We Act · 8 October 2026

The heaviest fifth of buyers supply far less than 80 per cent of sales, and many of them are not heavy buyers a year later. Plan for the drift.

Key takeaways

Byron Sharp's How Brands Grow prints annual Pareto shares averaging 50, 51, 55 and 49 per cent, far below 80.
A peer-reviewed 2019 analysis by Daniel McCarthy and Russell Winer found the top fifth supplied about two thirds of sales on average.
The share depends on the window: Sure body spray's heaviest fifth supplied 42 per cent over three months and 53 per cent over twelve.
In American tomato sauce data, the heaviest 9 per cent of households fell from 43 to 34 per cent of a brand's volume in a year, while previous non-buyers supplied 14 per cent.
A reward budget is not a growth budget: much of the drift in heavy buyers is regression to the mean.

Your best customers are not a group of people. They are a position on a spreadsheet, and many of the people in it this year will not be in it next year. A loyalty budget should be built around that, and it cuts against the most quoted rule in marketing: that a fifth of your customers bring in 80 per cent of your sales.

The rule is right about the shape. It is wrong about the size, and more usefully, it is wrong about who the fifth are.

Who counts as a best customer, and for how long?

A best customer is usually defined after the fact. Someone looks back over a period, ranks buyers by what they spent, and draws a line under the top fifth. Everything that follows depends on how long that period was, and it is rarely written down.

Byron Sharp's How Brands Grow shows why it matters. One of its tables takes buyers of Sure body spray in Britain, a brand with 16 per cent of its category, and asks what share of the brand's sales came from its heaviest fifth. Over three months the answer was 42 per cent. Over twelve months it was 53 per cent. Across the brands in the category the averages were 39 and 50.

Same brand, same customers, same shops. Only the window changed.

The reason is simple once you see it. In a short window most buyers buy once, so the top fifth cannot pull far ahead of everyone else. Over a longer window the differences between people have time to show, and the heaviest buyers stretch away. A company that reports its Pareto share without saying over what period has reported half a number.

The window also changes who lands in the top fifth. A household that stocked up in the quarter you happened to measure looks like one of your best customers. Measure the next quarter and it may look like a lapsed one, and drop out of the best customers list without having changed anything about how it shops. Neither label describes the household very well. Both describe the timing.

This is the first reason to hold the label loosely. "Best customers" sounds like a type of person, the devoted regular who would follow the brand anywhere. In the data it is closer to a snapshot of who happened to buy most during the weeks someone chose to count.

None of that makes heavy buyers unimportant. They matter a great deal, and the shape the rule describes is real: a minority of buyers do account for a disproportionate share of sales in almost every category. The argument is about how disproportionate, and about whether the minority stays put.

Is it eighty, sixty or somewhere in between?

The 80/20 figure is a convention, not a measurement. Two lines of work have tried to measure it, and they do not land in the same place.

Sharp and his colleagues at the Ehrenberg-Bass Institute call their version the 60/20 law. In How Brands Grow he writes that the top fifth's share over a year is "usually not far over 50% and rarely anything like the proverbial 80%". His own tables bear that out: the averages printed beneath the law are 50, 51, 55 and 49 per cent, from British body spray, twenty-two American categories, Australia and South Africa.

A 2019 working paper by Sharp, Jenni Romaniuk and Charles Graham makes the same case, that the heaviest fifth generally supply not much more than half of sales. It was posted to the Social Science Research Network and, as far as we can find, has not been through peer review.

Daniel McCarthy and Russell Winer looked at the question in a peer-reviewed paper in Marketing Letters the same year. Across the datasets they examined, the top fifth of customers accounted on average for about two thirds of sales, and for less in subscription businesses, about 59 per cent. Their conclusion was that the 80/20 rule overstates concentration. Their average sits above the Ehrenberg-Bass figures.

Neither figure is 80.

Both sides agree on the main thing. Eighty is too high. Your best customers matter less to your total than the slogan claims, and your light customers matter more.

Your best customers are not a group of people. They are a position on a spreadsheet.

Where they differ is how much less, and the difference is not small: a little over half against two thirds. Part of it may be the data: Sharp's tables are packaged goods bought in supermarkets. Our reading is that both are right about their own data, that the peer-reviewed figure deserves more weight as a general claim, and that the honest answer for any one business is a range that depends on the category, the window and how customers are counted. A company with a few very large accounts will look more concentrated than a supermarket. A brand bought every week will look different from one bought twice a year.

The practical point survives either figure. At half or at two thirds, a third or more of sales comes from the other four fifths of buyers, most of whom buy rarely. Any plan that treats them as an afterthought is planning around a minority of the revenue.

Why do heavy buyers drift back toward average?

The second problem with the rule is that the fifth keeps changing. Sharp reproduces a table, from American panel data on a leading tomato sauce brand, that makes this visible. In the first year, the heaviest buyers, those who bought it five or more times, were 9 per cent of the households in the panel and bought 43 per cent of the brand's volume. In the second year, the same households bought 34 per cent. Meanwhile, households that had not bought the brand at all in the first year supplied 14 per cent of its volume in the second.

Sharp calls this buyer moderation, and the pattern runs in both directions. Heavy buyers get lighter. Light buyers and non-buyers get heavier.

Nobody needs to have changed their mind about tomato sauce.

People run out at different times, have guests one month and not the next, or go away on holiday.

It is one brand, in one category, in one country, over two years, and the book offers no other table for this law. Work by Romaniuk and a colleague, presented at a marketing conference in 2010, points the same way: about half of a brand's heavy buyers were still heavy buyers in the following period, and the group's share of sales fell by about 17 points. That paper was not published in a journal, so we treat its figures as indicative rather than settled.

The statistical name for most of this is regression to the mean. When you select people because they scored high on a measure that contains some luck, their next score will tend to be closer to average, because part of what put them at the top was the luck. Purchase records are full of that kind of luck.

So a group of "best customers" chosen this year is partly a group of genuinely heavy buyers and partly a group who had an unusually heavy year. The second part will drift back whatever the brand does. The first part may stay heavy for years.

A business cannot tell the two apart by looking at a single period. It can only tell them apart by following the same people over time.

What does that mean for a loyalty budget?

The usual programme rewards this year's best customers in the hope of keeping them. Two problems follow from the evidence above.

The first is that a reward given to a heavy buyer is mostly a reward for behaviour that was already happening. The second is that some of the drop in their buying next year was coming anyway, so a programme can look as if it is failing when it is only watching regression to the mean, or look as if it is succeeding when it is only taking credit for people whose year was always going to be good.

Sharp's argument, set out at length in How Brands Grow, is that growth comes mainly from reaching more buyers, including the many who buy rarely, rather than from deepening the loyalty of the few who buy often. The light buyers who supplied 14 per cent of next period's tomato sauce were not in anyone's loyalty segment the period before.

That does not mean ignoring the heavy end. Losing a genuinely heavy buyer to a bad experience is expensive, and service failures are worth fixing for their own sake. It means not confusing a reward budget with a growth budget, and not treating a ranking from one window as a fixed list of people to defend.

From the customer's side, the same evidence is reassuring. If a brand's emails treat you as one of its best customers this quarter and ignore you the next, nothing about you has changed. The brand's window has moved.

There is also a fairness point for the people writing the brief. A programme that showers rewards on a fifth of customers chosen in one window, and asks nothing of the rest, is spending most heavily on the people least likely to need persuading. Whatever the right budget split is, it should be decided knowing that, rather than inherited from a slogan.

How can you check this in your own data?

Every number above comes from someone else's customers. Your own records will answer the question for your business, and the check needs nothing more than a year or two of purchase history.

Pick a window and rank. Take one year, rank customers by spend, and mark the top fifth.
Measure their share. Note what share of the year's sales they supplied. That is your concentration figure, and it is only true for that window.
Follow the same people. In the next year, find what share of sales the same customers supplied, and how many of them are still in the top fifth.
Look at the newcomers. Find what share of next year's sales came from people who bought little or nothing the year before.
Shorten the window. Repeat the exercise with three months instead of a year, and see how much the first number moves.

If your heaviest buyers mostly hold their place, and their share holds from one year to the next, your category is unusual and worth understanding. If they drift the way the tomato sauce buyers did, you have learned which budget you were running. Plan for the drift.

Common questions

What is the 60/20 rule in marketing?

It is the Ehrenberg-Bass Institute's name for the finding that the heaviest fifth of a brand's buyers supply far less than 80 per cent of its sales. Byron Sharp's own tables average about half over a year. A peer-reviewed analysis by Daniel McCarthy and Russell Winer found an average of about two thirds.

Do heavy buyers stay heavy buyers?

Many do not. In the tomato sauce data Sharp reproduces, the heaviest buyers' share fell from 43 to 34 per cent of the brand's volume in the next year, while previous non-buyers supplied 14 per cent. Much of the change is regression to the mean.

Sources

Sharp, B. (2010). How Brands Grow: What Marketers Don't Know. Oxford University Press. Read in full by this publication, 2026-10-08. Ch. 4: Table 4.1 (United Kingdom body spray and deodorants, three- and twelve-month Pareto shares, Sure 42 and 53, average 39 and 50); Table 4.2 (22 United States categories, average 51); Table 4.3 (Australia 55, South Africa 49); Table 4.4 (a leading United States tomato sauce brand, Anschuetz 2002, IRI panel: heavy buyers 9 per cent of households, 43 then 34 per cent of volume; year-one non-buyers 0 then 14 per cent); p. 46 quotation.

Sharp, B., Romaniuk, J., & Graham, C. (2019). “Marketing's 60/20 Pareto law.” SSRN working paper 3498097. doi:10.2139/ssrn.3498097. Not peer reviewed as far as we can find. Abstract only.

McCarthy, D. M., & Winer, R. S. (2019). 'The Pareto rule in marketing revisited.' Marketing Letters, 30(2), 139-150. doi:10.1007/s11002-019-09490-y. Average top-fifth share about 0.67; subscription businesses about 0.59.

A note on scope: the heavy-buyer turnover figures come from a 2010 conference paper by Romaniuk and Wight, not published in a journal and treated as indicative. The five-step check is this publication's suggestion, not a method from any of the papers.