How these numbers are made
Written so that a figure you disagree with can be argued with, rather than only believed or ignored.
Read this first.
195 of 195 cohorts currently held are synthetic placeholders. They are generated from a visible formula to exercise the engine and the interface. They are not measured benchmarks and no decision should rest on them. Whether a real UAE peer corpus may lawfully be built is an open question, not a scheduling one.
What a cohort is
A cohort is one metric, one jurisdiction, and one cut of the population. Most measures are cut by trade rather than by company size, because trade explains more of the variation than size does: a twelve-person MEP contractor and a forty-person MEP contractor have more in common on gross margin than a twelve-person contractor and a twelve-person software company.
Professional fees are the exception and are cut the other way — by revenue band, across every trade. An audit fee tracks the work a set of books makes, and size drives that far harder than sector does: a clinic and a trading company at AED 8m are quoted within a few thousand dirhams of each other, while the same clinic at AED 400k is quoted a fifth of it. Cutting fees by trade as well would split a thin population into cells that say nothing, and would imply a sector effect the data does not carry. The engine will read the exact size cell or refuse; it never answers a large company from a smaller company’s cell.
Every figure travels with the number of businesses behind it and the accounting period it covers. There is no code path that renders one without the other, because the type that carries a benchmark does not offer a shape without them.
The floor, and why it is where it is
No figure is published from fewer than 30 businesses. Below that a median starts to describe the businesses inside it rather than the trade around them — a competitor who sits in a small cohort and knows their own figure can narrow the rest.
Three other floors exist and none of them is this one:
- 3 — statistical offices publishing a fixed table nobody can submit to.
- 5 — a private report a client reads about themselves. Right for that use, and not inherited into this one.
- 10 — the recommendation for a public tool where the audience chooses the cohort, can sit inside it and can submit to it.
This tool holds 30, which is stricter than that recommendation and is a deliberate starting position. Lowering a floor once a corpus is real is cheap. Explaining a published cohort that was too small is not.
A cohort also fails to publish when one business accounts for more than 50% of its total for a metric. A cell can clear a headcount floor and still be one company wearing a disguise.
Of 195 cohorts held today, 172 clear the floor.
Why the answer is sometimes UNKNOWN
UNKNOWN is a result, not an error. It is what you get when the cohort is below the floor, when no cohort is held for your trade and metric, or when you did not supply the figure. In every one of those cases the alternative would be a number with nothing behind it, and a number with nothing behind it is indistinguishable from a real one once it is on the page.
Why your position is a band and not a point
You select ranges, never exact figures. A range is all a percentile ever needed, and an exact revenue figure typed into a stranger’s website is a disclosure nobody owes us. The cost is precision, and the interface pays it in the open: your marker is a capsule as wide as the range you picked, not a line pretending to know more.
How cohorts change
Recomputation is quarterly and scheduled. It never runs on write. If a published median moved the moment a submission landed, anyone watching the before and after could solve for the figure that had just been added — which is the opposite of what a floor is for. The same rule governs deletions: a cohort that would drop below the floor is withdrawn entirely rather than recomputed, because recomputing publishes the difference.
The measures
- Gross profit margin
- Am I charging enough for what I deliver?
(Revenue − Cost of goods sold) ÷ Revenue - Net profit margin
- What actually stays in the business?
Net profit after all costs ÷ Revenue - Revenue per employee
- Am I carrying more people than the work supports?
Revenue ÷ Full-time-equivalent headcount - Current ratio
- Can I cover what I owe over the next twelve months?
Current assets ÷ Current liabilities - Owner pay as a share of revenue
- Am I paying myself a normal amount?
Total owner compensation ÷ Revenue - How long your customers take to pay
- Is everyone in my trade waiting this long, or is it just me?
Trade receivables ÷ Revenue × 365 - What you are owed that has gone quiet
- How much of my invoiced work has stopped moving?
Receivables aged over 90 days ÷ Total trade receivables - Payroll as a share of revenue
- Is the team the size this revenue supports?
Total employment cost including owner pay ÷ Revenue - Rent and premises as a share of revenue
- Am I paying for space the revenue does not need?
Rent, service charge and utilities ÷ Revenue - Marketing as a share of revenue
- Am I spending what my trade spends to win work?
Marketing and advertising spend ÷ Revenue - Annual audit fee
- Is my auditor charging me what auditors charge?
Fee quoted or invoiced for the statutory audit of one financial year - Monthly bookkeeping fee
- Am I paying a normal price to keep the books?
Recurring monthly fee for bookkeeping and management accounts - Annual trade licence renewal
- Is my renewal in line with what others pay?
Total paid to renew the trade licence for one year, including fees and agent charges
What this tool will not tell you
It states what your numbers are and what the population’s numbers are. It does not tell you what to do about them and it does not value your business. It is not accounting, tax, audit or investment advice, and it is not a substitute for an engagement with a qualified accountant.
