THE MACHINERY OF THE RATIO NOT THE TOTAL

Why a bigger number tells you nothing, and the eye that learns to divide before it reacts

Two months, side by side, on the same report.

June: revenue one hundred and forty thousand. July: revenue one hundred and ninety thousand.

Everyone in the room can see that July was the better month. It is fifty thousand dollars better. The chart goes up and to the right and there is nothing to discuss.

Then someone asks what was spent to get it, and the room goes quiet, because in June the answer was twenty thousand and in July the answer was ninety.

June turned one dollar into seven. July turned one dollar into two.

The bigger month was the worse month, and it was worse by a factor of three, and every person looking at the report was looking directly at the number that could not tell them so.

PART ONE: A TOTAL IS AN ANSWER WITH THE QUESTION REMOVED

A total is a real fact. Revenue was one hundred and ninety thousand. That is true, it is not a lie, and it is almost useless on its own.

Because the question a total answers is: how much happened.

And the question that decides whether a business lives is: how much happened per unit of the thing that produced it.

Those are different questions, and the second one is the only one that carries any information about whether the thing can be done again, done more, or done at all.

A total has no denominator. It has been stripped of the very thing that would let you judge it. And once the denominator is gone, it cannot be recovered by staring harder at the numerator, no matter how big the numerator is or how many people are impressed by it.

The Denominator Is Always There, Whether or Not It Is Written Down

Here is the part that turns this from an accounting preference into a law.

Every total was produced by something. Spend, time, headcount, attention, inventory, risk. There is no such thing as an output that arrived without an input.

So the ratio exists whether or not anyone computes it. It is not created by the analyst. It is uncovered by the analyst. When a report shows only the total, the ratio has not been avoided. It has merely been left unread, and it is doing its work in the dark, and it is usually doing it against you.

The business that grows revenue while its ratio collapses is a business that is dying loudly. It has more customers than ever. It has more revenue than ever. It has more staff than ever. And every additional unit of the thing it is proud of costs more than it returns, so the faster it grows, the faster it fails, and every celebration is a report on the acceleration.

Between 1999 and 2000, a company called Pets.com went from almost nothing to real revenue, and it grew that revenue fast enough that the growth itself became the story. The commercials were famous. The sock puppet was on television during the Super Bowl. The total was moving. Underneath it, the ratio was fixed and it was known. The company was selling bags of dog food, heavy things, at prices below what it cost to buy and ship them. Every additional order was an additional loss. Every marketing dollar that worked, worked correctly and profitably in the only sense the total could measure: it produced a customer, and that customer bought, and revenue rose. The faster the total climbed, the faster the money left, because the ratio underneath the total was less than one and no amount of volume repairs a ratio less than one. Volume multiplies it. The company went from its public offering to liquidation in nine months. Nothing about that was a demand problem. People wanted the product. The business simply had a per-unit relationship that guaranteed that success would kill it, and the total was the number that made success look like success right up until the end.

PART TWO: THE FOUR RATIOS UNDER EVERY BUSINESS

There are only a handful of denominators that matter, and each one converts a total into a fact that can actually be acted on.

Per dollar spent. What came back for what went out. Revenue over spend. The single most brutal number in marketing, and the one most often left uncalculated because the total is right there and it feels like enough.

Per customer. What the average person is worth, and what the average person cost. A revenue number that rose because ten thousand more people bought is a different business from a revenue number that rose because the same people spent more, and the total cannot tell those two apart.

Per unit of time. Rate, not volume. Fifty thousand a month and fifty thousand a week are the same total and different companies.

Per person on the payroll. The output the organization produces divided by the people it took. This one moves quietly, and it is the one that tells you whether you built a machine or bought a crowd.

The Ratio Is the Only Thing That Predicts

A total tells you what happened. It is a receipt.

A ratio tells you what will happen if you do it again, which is the only reason anyone is looking at the report in the first place.

If a channel returned four dollars for one, then the next dollar has some claim on returning about four. Not a guarantee, but a basis. If a channel returned one hundred thousand dollars in revenue and nobody knows what it cost, the next dollar has no claim on anything at all, and the decision to spend it is not a decision. It is a mood.

This is why the person who reads ratios can act and the person who reads totals can only react. The ratio contains the future in it. The total has already spent it.

PART THREE: WHY THE EYE GOES TO THE TOTAL

The pull toward the total is not stupidity. It is structural, and it comes from three places at once.

The total is larger. One hundred and ninety thousand is a more impressive object than the number two point one, and the nervous system does not weigh those two things fairly. Magnitude reads as importance.

The total requires no work. It is already computed. The ratio requires that someone go and find the denominator, and the denominator usually lives in a different system, owned by a different person, updated on a different day.

The total is what gets asked for. Nobody in a status meeting asks what the return on spend was. They ask how much revenue came in. So that is what gets prepared, and what gets prepared is what gets seen, and what gets seen becomes what is real.

The Soviet planning apparatus ran on totals for seventy years and produced one of the most reliable demonstrations in economic history of what happens next. A nail factory measured on the number of nails produced flooded the country with tiny useless tacks, because the total was nails and tacks are nails. The planners, seeing the problem, switched the measure to the tonnage of nails produced. The factory then produced enormous nails, some reportedly the size of railway spikes, because the total was now weight and a heavy nail is heavy. Both times the factory hit the plan. Both times the plan was a total. And both times the thing the country actually needed, which was usable nails in the sizes that hold buildings together, was destroyed by the number that was chosen to protect it. Nobody in that story was lazy or corrupt. Everyone was optimizing exactly what they were measured on. The failure was upstream of every person involved, in the room where someone wrote down a total instead of a ratio.

PART FOUR: THE THREE WAYS A RATIO LIES

The ratio is the better instrument, and it is still an instrument, and instruments can be misread. Three ways, and each one has ended a company.

The Denominator Was Chosen Too Small

A campaign returns eight dollars for every one spent, and the room celebrates, and the denominator was the ad spend only. It did not include the discount that closed the sale, the shipping that was eaten, the support hours the new customers consumed, or the salary of the three people who ran the campaign.

The true denominator is everything that had to happen for the numerator to exist. Any cost that is real and left out of the bottom of the fraction reappears later as a mystery, as a business that is profitable on every campaign and unprofitable in the bank account.

The Ratio Is Fine and the Volume Is Trivial

A channel returns twelve to one. It is the best ratio in the business by a distance. It is also two thousand dollars a month, and it cannot absorb more, and nothing about it can be scaled.

An excellent ratio on a base of nothing is a curiosity, not a strategy. The ratio tells you the quality of the machine. The total tells you the size of the machine. Reading only the ratio produces the opposite error: a business that optimizes itself into a tiny, immaculate, irrelevant corner.

Which means the two numbers are not rivals. They are two coordinates of one point, and a decision made on either one alone is being made blind in one eye.

The Ratio Is an Average Over Things That Do Not Belong Together

The blended return on spend is three to one, and it is composed of one channel at eight and one channel at zero point four, and the good one is subsidizing the bad one in a way that is completely invisible in the average.

Every blended ratio hides its own composition. The instant it is disaggregated, the decision usually becomes obvious, and the reason it was not obvious before is that averaging is a form of deletion.

In the summer of 2018, a large consumer brand ran an experiment that has since been repeated by others with the same result. They turned off a portion of their branded search advertising, the ads that appear when someone types the company's own name into a search engine. The reported return on that spend had been excellent. It was one of the best-performing lines in the entire budget, and it had been renewed year after year on the strength of that number. When the ads went dark, revenue from that traffic barely moved. The people had been typing the company's name because they had already decided to buy. They clicked the ad because the ad was at the top. Had it not been there, they would have clicked the free link below it, and the sale would have happened anyway, at no cost. The ratio was not wrong arithmetically. Every click was real, every sale was real, and the division had been performed correctly. The ratio was wrong because the numerator contained revenue the spend did not cause. The company had been paying, for years, to purchase customers it already had.

PART FIVE: THE INCREMENT IS THE ONLY HONEST DENOMINATOR

That last case opens the deepest part of this.

The question a ratio is really asking is causal. Not what came back alongside this spend, but what came back because of it.

Almost every reported ratio in marketing answers the first question and is quietly presented as an answer to the second. The channel gets credit for every sale that passed through it, including the sales that would have happened if the channel had never existed.

Which means the honest denominator is not the spend. It is the spend, compared against the world where the spend did not happen.

That world is not visible, and it cannot be inferred from the dashboard, and it is the only thing that would settle the question. So it has to be created deliberately: hold a region back, turn a channel off for two weeks, split the audience and starve one half on purpose.

This is the one activity in marketing that feels like a waste of money and is not. Turning off spend that appears to be working, in order to see what is actually true, is how the difference between a real ratio and a flattering one gets discovered, and there is no other way to discover it.

PART SIX: WHAT CHANGES IN THE PERSON WHO SEES IT

Something happens when the eye learns to divide before it reacts, and it happens quietly.

The big number stops landing. Someone announces revenue and the mind does not feel anything yet, because the sentence is incomplete, and the incompleteness has become physically noticeable, the way a missing stair is noticeable.

Per what.

That question, arriving automatically, before the reaction, before the congratulation, is the whole of the skill. It is not a technique and it cannot be installed by agreeing with it. It arrives when enough total-shaped statements have been divided, out loud, in front of the consequences, that the division starts running on its own.

And then a strange thing follows. Calm.

Because the person who reads ratios is no longer at the mercy of the direction of the chart. A rising line does not produce elation and a falling one does not produce panic, since neither line is yet an event. Both of them are half a fact waiting for a denominator, and the denominator is the part that decides what to do.

The room celebrating the record month, and the one person in it who has gone quiet and is doing arithmetic, are not looking at different data. They are looking at different numbers of dimensions in the same data.

SYNTHESIS

THE TWO WAYS TO READ THE SAME MONTH

  TOTAL                          RATIO
  revenue $190,000               revenue / spend = 2.1
  up 36% over June               June was 7.0
  the best month ever            the worst month in a year
       │                              │
       ▼                              ▼
  spend more, it is working      stop, something broke


  A TOTAL HAS NO DENOMINATOR. IT WAS REMOVED, NOT ABSENT.


THE CHAIN

   every total was produced by something
            │
            ▼
   so the ratio exists whether or not it is computed
            │
            ▼
   uncomputed, it works in the dark, usually against you
            │
            ▼
   the business grows, the ratio collapses, and the
   growth is what kills it, because volume multiplies
   a ratio, it does not repair one

A total answers how much happened. A ratio answers how much happened per unit of what produced it, and only the second one contains the future.

The denominator is always there. Every output had an input. Choosing not to divide does not remove the relationship, it only removes your ability to see it, and a business whose ratio is below one grows itself to death while every report says it is winning.

The eye goes to the total for three reasons that have nothing to do with truth: the total is bigger, the total is already computed, and the total is what gets asked for in the room.

The ratio lies in three ways, and all three are repairable: the denominator was drawn too small, the ratio is excellent on a base of nothing, or the average has deleted the composition that would have made the decision obvious.

And underneath all of it sits the causal question that almost nobody pays to answer. Not what came back alongside the spend. What came back because of it. The only way to know is to turn it off and watch, which feels like burning money and is the cheapest information a business can buy.

The skill is one question, arriving before the reaction, every time, without effort.

Per what.

CITATIONS

Goodhart, C. (1975). Problems of monetary management: the U.K. experience. Papers in Monetary Economics, Reserve Bank of Australia. When a measure becomes a target it ceases to be a good measure. A total made into a target is the most common instance.

Blake, T., Nosko, C., & Tadelis, S. (2015). Consumer heterogeneity and paid search effectiveness: A large-scale field experiment. Econometrica, 83(1), 155-174. eBay turned off branded search advertising and found the measured return was close to zero: the traffic converted anyway. The reported ratio had been crediting the channel with sales it did not cause.

Lewis, R.A., & Rao, J.M. (2015). The unfavorable economics of measuring the returns to advertising. Quarterly Journal of Economics, 130(4), 1941-1973. The statistical power required to detect true advertising effects is far greater than practitioners assume, which is why observational ratios routinely mislead and why deliberate holdouts are necessary.

Nove, A. (1958). The Soviet Economy. Allen & Unwin. The nail factory measured in units and then in tonnage. A canonical demonstration that the chosen total, not the intention behind it, determines what an organization produces.

Kaplan, R.S., & Anderson, S.R. (2004). Time-driven activity-based costing. Harvard Business Review, 82(11), 131-138. Most reported per-unit economics omit real costs that live in other systems, which is how a business is profitable on every campaign and unprofitable in the bank.

THE MACHINERY OF THE COST OF A CUSTOMER. One denominator, taken all the way down: what a customer actually costs, and how the eye learns to feel when it is too high.
THE MACHINERY OF THE LIFETIME. The other half of the fraction. What a customer is worth over the whole relationship, and why a ratio computed on the first sale is a ratio computed on a fragment.
THE MACHINERY OF THE EFFICIENT CHANNEL. What happens when the same ratio is computed per channel instead of blended, and why the blend was hiding the decision.
THE MACHINERY OF UPSTREAM LEVERAGE. The general law underneath: a small correction at the origin instead of a large and expensive one at the end.