Most companies know their margin. Almost none know where it leaks.

A benchmark is a position, not a verdict.

A benchmark does not tell you whether you are good or bad. It tells you where you sit on a distribution, and how far that is from where the recoverable margin usually hides. Two businesses with the same reported profit can sit at opposite ends of every dimension below: one earns its margin cleanly across the book, the other earns a large margin on a few accounts and gives most of it back on a long tail it cannot see.

This framework is deliberately built from operating models rather than from a survey. A survey tells you what finance leaders believe about their costs. A model tells you what the costs actually are once operational effort is attributed down to the customer and the order. That is the standard the first four dimensions are held to, and it is also why three of them still show no figure: we produce those numbers one engagement at a time, and we have not aggregated them into something we would publish. The fifth dimension, costing maturity, is different by design. It asks whether the mechanics exist at all, which is a question an organisation can answer about itself, so a structured self-assessment is the right instrument for it. That is the one figure on this page with a stated sample size behind it, and even that sample is small and self-selected.

Five numbers a headline margin cannot show you.

Each dimension answers one question and is measured one way. Where we have a figure we publish it with its source. Where we do not, the row says so rather than filling the gap with a plausible number.

The widest gap, not the worst number, is where you start.

Placing yourself is simple, and it works even on the rows where we publish nothing. Put your own number in the last column for each dimension. On costing maturity you can compare directly: the median of the 19 organisations we assessed is 33 out of 100 and 89% of them scored 50 or below, so a score in the forties is common rather than comfortable. On unprofitable-customer share you can compare against the 20 to 40% we see in our models. On the remaining three the comparison you can make today is with yourself, and if you cannot produce the number at all, that is the finding. The dimension nobody has ever measured is usually the one carrying the most recoverable margin, precisely because nobody has looked.

A short illustration, with illustrative figures. Suppose a distributor looks unremarkable on four dimensions but finds that a large share of its customers contribute negatively once cost-to-serve is attributed. Cumulative profit climbs to well above the reported total on the profitable core, then a long tail of small, frequent orders drags it back down. The gap between the peak and the reported total is the recoverable margin. The answer is almost never to drop those customers. It is to re-price, consolidate orders and match effort to value.

Dated, sourced, and honest about the sample.

One figure here carries a stated sample. Costing maturity comes from The State of Costing, H1 2026, which reports 19 organisations from outside our own team who completed the Profit Check with a business email address between 7 May and 18 June 2026. Median 33 out of 100, mean 36, and 89% scoring 50 or below. It is a small, self-selected group, most likely skewed toward organisations that already suspect a costing problem, so we read it as an early signal and not as a market census.

Unprofitable-customer share is different. The 20 to 40% is what we observe in the operating models we build, attributing operational cost down to individual customers, orders and resource groups. It is a practitioner observation across our engagement base, not a study with a published sample, and we label it that way rather than dressing it up as research.

The other three dimensions we also measure inside those models, one engagement at a time, but we have not aggregated them into figures we are prepared to publish. Until we have, those rows say not yet reportable. We would rather hand you an empty cell than a number we cannot show the working for. As the base grows we will fill them in, with the sample and the date attached. Last reviewed 26 July 2026.