Total revenue fell 6% while paying-customer count is flat — how does account concentration change your read?

## Start from the identity Total revenue is the number of paying accounts multiplied by average revenue per account. If the account count is flat and the total fell 6%, then average revenue per account fell about 6%. That is arithmetic, and it is where many candidates stop — concluding that customers are spending less. The interesting question is whether 'customers' means all of them or two of them. ## Concentration is the normal case, not the exception In most B2B and marketplace revenue books, spend is heavily concentrated: a small number of accounts contribute a large share of the total. An advertising business where the top ten advertisers are a quarter of revenue is ordinary. Under that shape, the average is a fragile summary — it is dominated by the top of the distribution, and it can move several percent because one relationship changed while every other account behaved identically to last month. This has a direct consequence for diagnosis: an aggregate percentage change is not evidence about the typical customer until you have checked how concentrated the metric is. ## The three checks that settle it **1. Measure the concentration.** Compute the share of revenue held by the top 1, top 10 and top 100 accounts in the earlier period. This tells you how much a single account can move the total. If the top account alone is 5% of revenue, a 6% drop is within reach of one churn. **2. Look at the distribution of per-account change, not just the total.** For accounts present in both periods, compute each one's change. If the median account is flat and only a couple of large accounts are deeply negative, the aggregate delta is an account story. If the median account is down 5%, something broad is happening and the whales are incidental. **3. Recompute the delta without the top accounts.** This is the decisive test. Take the same period comparison with the largest few accounts removed from both sides, and see how much of the 6% survives. If the remaining book is flat, one or two accounts explain everything. If 4 of the 6 points survive, the concentration explains only a third and the broad investigation continues. Report the number both ways — total change and change excluding the top accounts — because the pair is far more informative than either alone. Note what these checks are not. Deciding whether an unusually large value should be trimmed, winsorised, or replaced by a robust estimator is a separate discipline with its own rules. Here the large accounts are real revenue from real customers; the point is not to remove them from the truth, but to find out whether the aggregate movement is about them or about everyone. ## Why the answer changes what happens next The two diagnoses route to different owners and different actions. - **One large account churned or cut spend.** This is a commercial event. The right next steps are with the account team: why did they leave, was it price, a competitor, a budget cycle, a service failure, is it recoverable, and are similar accounts at risk. Engineering has nothing to fix. Announcing this as a product regression burns credibility and engineering time. - **The typical account is down.** Now a few percent across thousands of accounts points at something systemic — pricing, a product change, a competitor, a market shift, a seasonal effect. Concentration is irrelevant and the aggregate number was telling the truth. A subtle third case: the whale left *and* the median account is down slightly. Then the headline number overstates the broad problem, and reporting only the total hides a real, smaller systemic issue behind a dramatic single-account story. Quantify both parts rather than choosing a single narrative. ## The monitoring consequence A concentrated total is intrinsically volatile: its week-to-week variation is driven by the timing of a few large accounts' spend, invoicing and campaigns, not by the behaviour of the population. Any alerting threshold on that total must be wide enough to absorb ordinary whale behaviour, which makes it insensitive to broad regressions. The usual resolution is not one metric but a small set reported together: the total, the count of active paying accounts, revenue excluding the top N accounts, and the top-10 revenue share. The first is what the business earns; the others tell you, on sight, whether a move is about the population or about a relationship. ## How to communicate it The most persuasive artefact is two numbers side by side: revenue change including all accounts, and revenue change excluding the top ten. When the first is -6% and the second is -0.3%, the conversation ends quickly and correctly — and the follow-up is a phone call, not an incident channel.