Bellow is an chart of AOV for one brand that we worked with.
Its distribution of all different orders customers made.
The calculation is correct, the ‘AOV’ customer is fictional.
There was no smooth hill around the average order value. There were two hills with an empty valley between them — and the average sat in that valley, describing a renewal amount almost nobody paid.
01 · One number swallowed two businesses
Same average, opposite risk
The company sold different products through different offers. Those purchases produced customers with very different renewal values — yet the dashboard blended them into one AOV and one LTV. Averaging the groups didn't simplify reality. It erased the most important thing about it.
One is a healthy population worth $200 each. The other is a large group worth $100 propped up by a small group worth $600. Same average — but one scales predictably, and the other depends on a small, unstable group continuing to behave unusually well. AOV and LTV cannot tell you which business you own.
About 20% of eligible lower-tier customers reached a second renewal; only about 5% of higher-tier customers did. These figures are observational, not proof that price caused the gap — but they prove the blended customer isn't enough to understand the business.
02 · The average forces together what should stay separate
When the middle is real — and when it's a mirage
An average answers one narrow question: if we spread the total evenly, how much would each order receive? That's useful for accounting. It's dangerous when treated as a description of a real customer.
An AOV of $67 is a fair summary. It sits close to the crowd. A blended $100 describes neither.
The blended average suppresses the operating questions: which product did each buy, which funnel acquired them, what did acquisition cost, how often did they renew, how concentrated is revenue inside each group?
03 · Why renewal makes the mistake expensive
The error compounds when it reaches an LTV calculation
AOV looks backward at completed orders. LTV is used to justify future spending — so an inaccurate model of the customer becomes far more dangerous once it drives the acquisition budget.
Higher-value subscribers inflate blended LTV → the company raises allowable acquisition cost for everyone → marketing spends as if every buyer might behave like the blend → months later the renewals mature and the value never appears. Nothing was wrong with the arithmetic — the company applied the economics of one population to another.
04 · The $180 coaching customer
Too high for the front end, too low for the back
The same mistake becomes extreme in information and coaching businesses with a wide offer ladder. Most customers buy the book or course; a few enter the mastermind; one or two buy the $25,000 service. Blend it all and the dashboard reports an AOV of $180.
The book buyer didn't spend $180. The course buyer didn't. The mastermind buyer certainly didn't. The average lies in both directions at once — too high to describe the front end, too low to describe the back.
Treat $180 as an ordinary buyer and you overspend acquiring $27 customers. Optimize the $180 average and you may ignore the tiny group responsible for most of the profit. The questions that matter — how many reach the high-ticket offer, which entry products produce them, what happens if three whales disappear — stay hidden.
05 · High variance makes revenue experiments painfully slow
You need four times the data to be twice as sure
When one customer is worth $27 and another $25,000, revenue per customer has an enormous standard deviation — and the wider the spread, the less stable the mean. The uncertainty around a sample mean shrinks only with the square root of sample size.
To cut uncertainty in half you need roughly 4× the observations; to cut it to a third, roughly 9×. With a huge standard deviation, an AOV test may need an enormous sample before a real lift can be told apart from whale randomness.
B's revenue per visitor jumps and its AOV "wins." But the whale could have landed in either group; re-run the test and the winner reverses. The experiment measured where a rare customer happened to land, not whether the experience improved.
06 · The histogram reveals the business model
Four different businesses. One identical AOV.
AOV can't show whether revenue is stable, segmented, or whale-dependent. Neither can LTV. Even the mean and median together only tell you that something deserves investigation. The histogram shows you what.
One narrow hill: a homogeneous population. One hill with a long right tail: rare large purchases pulling the mean up. Two hills: distinct products, offers, or tiers. A near-empty chart with a few distant values: severe whale dependence. They need different acquisition limits, forecasts, experiments, and retention strategies — yet all four produce the same AOV.
Some information only becomes visible when you stop reducing it to a number.
07 · Is the business doomed?
The business may be healthy. The model in the dashboard is what's doomed.
A whale-funded business can be excellent — high-ticket services carry extraordinary margins, premium subscribers can be immensely valuable. The danger is operating that business as though its revenue were evenly distributed: spending against blended LTV, forecasting from blended AOV, and treating rare high-value customers as ordinary, repeatable outcomes.
If the histogram shows two populations, manage two populations.
If it shows a whale tail, measure the whales.
Don't distribute their value across everyone else and call the result a customer.
08 · What to report instead
Never report AOV alone
When the mean sits far above the median, investigate what pulls it up. If the histogram has two humps, split them and report count, acquisition cost, order value, retention, and LTV for each group.
For a whale business, add concentration:
· Share of revenue from the top 1% of customers
· Share of revenue from the top 10 customers
· Number of high-ticket buyers
· Front-end → high-ticket conversion rate
For experiments & subscriptions:
· Use purchase rate to test whether a lander converts
· Keep revenue secondary until the sample survives the whales
· Measure renewal milestones separately by product, offer, or tier
· Don't blend populations first and ask questions later
The average is answering a question you didn't ask
AOV tells you what every order would be worth if total revenue were spread evenly. LTV tells you what every customer would be worth if total customer value were spread evenly. Neither number ever promised to describe a real person. We simply started treating them as if they did.
Before trusting an average, compare it with the median. Before spending against it, look at the histogram.
Because sometimes the average customer does not exist — and sometimes your entire strategy is built around them.
You pay $1 today, and that gets you the full tool for a month, running next to whatever you use now so you can compare them on your own numbers.
At the end of the month, if Bratrax has earned the switch, you keep it at a flat $79 a month, locked for life, as one of the first 100 founding members.
If it hasn't, you cancel in one click and you're out a dollar. The doors close [set close date], or when the 100th seat is taken.
— Brat
P.P.S. Not sure Bratrax fits your stack? See exactly what you get, and how it compares to the tools you already know → bratrax.com
