THE MACHINERY OF NEW VERSUS RETURNING

Two businesses hiding inside one revenue line, and why only one of them compounds

A shop does twenty thousand dollars a week, every week, for a year. The line on the chart is flat and boring and everyone stops looking at it.

Underneath that flat line, two completely different companies are possible.

In the first, four hundred people come every week. The same four hundred. They know the staff, they order without looking at the menu, and they have brought their friends. Marketing spend is near zero, because the customers are the marketing.

In the second, four hundred different people come every week. Nobody has ever been there before. Nobody comes back. Every single one of them had to be found and paid for, and next week another four hundred strangers have to be found and paid for, forever, or the number collapses to nothing.

Same revenue. Same chart. Same flat line.

One of those businesses is an asset that is quietly compounding. The other is a treadmill with a bill attached, and it is one bad month of ad performance away from having no business at all.

The revenue line cannot tell them apart. It never could.

PART ONE: REVENUE IS A SUM OF TWO THINGS THAT BEHAVE NOTHING ALIKE

Every dollar that comes in this month came from one of exactly two places.

Someone who had never bought before. Or someone who had.

That is not a categorization scheme. It is an actual partition of the money, complete and non-overlapping, and the two halves obey different laws.

New revenue costs money to produce. It requires reach, attention, persuasion, and the overcoming of doubt. It is bought.

Returning revenue costs almost nothing to produce. The reach already happened. The attention is already given. The doubt was settled by the product itself, the last time. It is earned, once, and then it arrives.

The Two Halves Have Opposite Failure Modes

A business running mostly on new revenue is fragile in a specific way: it fails the moment acquisition gets harder. And acquisition always gets harder, because the cheapest customers get bought first, the platform raises its prices, and the competitor bids against you. A business like this has no floor. Turn off the spend and the revenue goes to approximately zero, on a delay of about a month.

A business running mostly on returning revenue is fragile in the opposite way: it fails slowly, invisibly, if it stops adding anyone. Every customer base leaks. People move, they age out, they change, they die. A business with a strong base and no new blood looks completely healthy for two years and then discovers that it has been shrinking the entire time.

Which means the question is never which half is better. The question is what the ratio between them is, whether that ratio is the one you chose, and whether you know it.

PART TWO: THE FLYWHEEL AND THE TREADMILL

Here is the structural difference, and it is the whole reason this distinction matters more than almost any other split in a business.

Returning revenue compounds. New revenue does not.

A customer who comes back is not one sale. They are a sale, plus a probability of another sale, plus the probability of a referral, plus a reduced need for the next marketing dollar. The value of that one relationship keeps producing after the transaction that started it has been forgotten.

A customer who does not come back is one sale, and the cost of acquiring them is a fixed toll that must be paid again, in full, for the next one.

So the same revenue number, produced by returning customers, is worth several times what it is worth produced by strangers. Not emotionally. Structurally. It has a lower cost, a lower variance, and a compounding tail attached to it.

Amazon Prime is the most expensive thing the company ever built for the purpose of not needing to buy its customers again. The economics of it look insane on first inspection. Free shipping on everything, absorbed by the company, on goods with thin margins. Every analysis of the first-order numbers says the program loses money on shipping. It was never about shipping. A Prime member buys more often, across more categories, and stops comparison shopping, because the cost of checking somewhere else exceeds the friction of just ordering. The customer stops being a person who must be won on every purchase and becomes a person who arrives by default. The company chose to pay a large recurring cost in order to convert new revenue into returning revenue, permanently, and the returning revenue compounds in a way the shipping bill does not. Two decades later, Prime members spend multiples of what non-members spend, and the acquisition cost for each of those additional purchases is zero. The shipping was never the product. The habit was the product.

PART THREE: HOW THE MIX HIDES ITSELF

The mix is invisible for a set of very ordinary reasons, and each one is worth naming, because the invisibility is the entire problem.

The revenue line is reported as one number. Nobody is hiding anything. The sum is simply computed, and a sum destroys the composition of what it summed.

Growth from new customers and growth from returning customers look identical on a chart. Both are a line going up. The chart has no dimension in which the difference could appear.

The two halves are usually owned by different people. Marketing owns acquisition. Operations or product owns retention. Each one reports on their own half and neither reports the ratio, because the ratio belongs to nobody.

And most of all, new customers are more exciting. A new customer is a win, an event, a notification. A returning customer is a Tuesday. The nervous system reacts to the first and does not register the second, which is exactly backwards from how they should be valued.

The Growth That Is Actually Erosion

Put those together and you get the most common form of business self-deception there is.

Revenue is up eight percent. Everyone is pleased. Underneath, new customer revenue is up forty percent because spend was up sixty, and returning customer revenue is down twelve because the product got worse and nobody noticed, and the two movements sum to a pleasant eight.

The company is celebrating the number that is concealing its own decay. And the decay is in the half that compounds, which means it will keep compounding downward, while the half that was propping up the total is the half that has to be purchased again next month at a higher price.

There is no version of that report where anybody is lying. Every number in it is correct.

Groupon in 2011 was, by revenue, one of the fastest growing companies in the history of commerce. Revenue climbed from tens of millions to well over a billion in about two years, and it went public on the strength of that curve. The curve was real. It was also almost entirely new. The merchants who ran a deal mostly did not run a second one, because a coupon flood of one-time bargain hunters did not produce customers, it produced a crowd. And the customers who redeemed a deal mostly did not come back to the merchant at full price, for the same reason: they had been acquired by a discount, and what a discount acquires is a person who wants a discount. Both sides of the marketplace were being bought fresh, over and over. So the machine had to keep buying, and the cost of buying kept rising, and the growth that had made it famous was structurally incapable of turning into anything that stayed. The revenue was real every single quarter. The business underneath it was a treadmill that had been mistaken for a flywheel, and the difference did not show up in the revenue line until the spending had to stop.

PART FOUR: THE THREE NUMBERS THAT MAKE THE MIX VISIBLE

Once you want to see it, three numbers do almost all of the work, and none of them are exotic.

The share of revenue that is returning. One fraction. Returning revenue over total revenue, this month, next to the same fraction last month and the same month last year. This one number tells you which of the two businesses you are actually running, and it takes about an hour to compute for the first time and about a minute after that.

The repeat rate of a cohort. Of the people who bought for the first time in January, what fraction bought again by March. Then the same for February’s group, and March’s. This is the closest thing a business has to a health readout, because it is the only number that measures whether the product earns a second visit, with acquisition held constant.

The direction of the cohort curves. Not the level, the direction. Is the group that arrived this month coming back more often, or less often, than the group that arrived six months ago. That is the derivative, and it moves long before revenue does. A product that is quietly getting worse shows up here first, and it shows up here while there is still time.

PART FIVE: WHAT THE RATIO IS TELLING YOU TO DO

The mix is not just a diagnostic. It prescribes.

When the returning share is high and the total is flat, the machine works and nobody knows about it. The product earns the second visit. The constraint is at the top, not the middle, and money spent on reach will compound instead of leaking. This is the one condition under which aggressive acquisition spend is obviously correct, and it is the condition under which businesses most often refuse to spend, because everything feels comfortable.

When the returning share is low and the total is rising, stop. The rise is being purchased and the purchase is not building anything. Every additional dollar of acquisition is pouring water into a bucket whose hole has not been found. The instinct in this state is always to spend more, because spending more does raise the number, and that is precisely the trap.

When the returning share is falling while the total holds, the product broke and the spend is hiding it. This is the most dangerous state a business can be in, because it looks like stability. It is the state in which the report says nothing is wrong for two consecutive quarters while the asset underneath quietly dissolves.

The order is not a preference. Fix the hole, then pour. A leaking bucket filled faster is not a fuller bucket. It is a wetter floor and a larger bill.

PART SIX: WHAT CHANGES IN THE PERSON WHO SEES IT

The revenue number stops being one thing.

Someone says the month was up, and the mind no longer completes the sentence. It waits. Up on which half. And that waiting is not skepticism or negativity, it is the simple fact that the sentence, as spoken, does not yet contain enough to know what happened.

Then a second thing happens, and it is stranger.

The excitement about new customers goes quiet, and something like respect appears for the returning ones. Not sentiment. Arithmetic. The person who came back did not have to be found, did not have to be persuaded, did not cost anything, and brought a friend, and did all of it without generating a single notification.

The loudest revenue in the business is the revenue that was most expensive to get. The revenue that compounds arrives silently and gets no credit at all.

Once that inversion is seen, it cannot be unseen, and it changes where a person looks first when the report comes in.

SYNTHESIS

ONE REVENUE LINE. TWO BUSINESSES.

  $20,000 / week                    $20,000 / week
  400 people, the same 400          400 people, all strangers
       │                                 │
  cost to produce: ~0               cost to produce: full price
  compounds: yes                    compounds: no
  turn off the ads: fine            turn off the ads: gone in 30 days
       │                                 │
       ▼                                 ▼
   AN ASSET                          A TREADMILL


THE CHAIN

   every dollar came from a first-time buyer or a returning one
            │
            ▼
   the two halves have opposite costs and opposite risks
            │
            ▼
   the revenue line sums them, and a sum deletes composition
            │
            ▼
   so a business can grow its total while the half that
   compounds erodes, and every report says it is winning


   THE ORDER IS FIXED

   returning share LOW  + total RISING   ──►  stop. the bucket leaks.
   returning share HIGH + total FLAT     ──►  spend. it will compound.
   returning share FALLING + total FLAT  ──►  the product broke.
                                              the spend is hiding it.

Revenue is not one thing. It is the sum of money from people who had never bought and money from people who had, and those two halves cost different amounts, carry different risk, and behave in opposite ways when conditions change.

Returning revenue compounds. It was paid for once and it keeps arriving, and it brings other people with it. New revenue is bought at full price, every time, in an auction whose price only goes up.

The mix hides because a sum destroys composition, because both kinds of growth draw the same line on the same chart, because the two halves are owned by different departments, and because a new customer feels like an event while a returning one feels like a Tuesday.

Three numbers make it visible: the returning share of revenue, the repeat rate of each arriving cohort, and the direction those cohort curves are moving. The third one moves first, which makes it the only early warning a business gets.

And the order of operations never changes. Find out whether the bucket holds water before deciding how fast to pour.

CITATIONS

Reichheld, F.F., & Sasser, W.E. (1990). Zero defections: quality comes to services. Harvard Business Review, 68(5), 105-111. The original quantification of retention economics: small changes in the fraction of customers who return produce disproportionate changes in profit, because the returning customer carries no acquisition cost.

Fader, P.S., & Hardie, B.G.S. (2009). Probability models for customer-base analysis. Journal of Interactive Marketing, 23(1), 61-69. Cohort-based repeat purchase modelling. The health of a customer base is legible in the repeat behaviour of arriving cohorts long before it is legible in aggregate revenue.

Ehrenberg, A.S.C. (1988). Repeat-Buying: Facts, Theory and Applications. Oxford University Press. Repeat purchase patterns are highly regular and highly predictive, and aggregate sales figures routinely conceal them.

Gupta, S., Lehmann, D.R., & Stuart, J.A. (2004). Valuing customers. Journal of Marketing Research, 41(1), 7-18. The value of a company is closely approximated by the value of its customer relationships, which means the composition of revenue, not its size, is what is being valued.

Blattberg, R.C., & Neslin, S.A. (1990). Sales Promotion: Concepts, Methods, and Strategies. Prentice Hall. Customers acquired through discounting exhibit systematically lower repeat rates. What a discount acquires is a person who responds to discounts.

THE MACHINERY OF THE LIFETIME. What a returning customer is actually worth, taken all the way out to the end of the relationship.
THE MACHINERY OF THE COST OF A CUSTOMER. The price of the other half. What it costs to buy a stranger, and how the eye learns to feel when it is too high.
THE MACHINERY OF THE RATIO NOT THE TOTAL. The general form of this error. A total is an answer with the question removed.
THE MACHINERY OF THE EFFICIENT CHANNEL. Where the new half comes from, and why the blended average was hiding the decision.