Everyone who became a client in the same month is followed through their whole life with us, month by month. This report answers one question: how much is a client worth, and how long does it take them to become worth it?
Take everyone who first paid us in June 2026 — 212 people. That group is a cohort, and it gets one row. The row then tracks what those same 212 people go on to spend, for as long as they keep spending.
This is the single thing people get wrong. The columns are months of life, not calendar months. M1 means "the cohort's own first month", which is a different calendar month for every row. That's what makes the rows comparable — you're always lining up like with like.
| Cohort | Jan 2026 | Feb 2026 | Mar 2026 | Apr 2026 | May 2026 |
|---|---|---|---|---|---|
| Jan 2026 row | M1 | M2 | M3 | M4 | M5 |
| Mar 2026 row | — | — | M1 | M2 | M3 |
Both rows have an M1. They just happen in different calendar months. Comparing M1 to M1 tells you whether March's clients started stronger than January's.
Every number in the grid is the same thing: our 25% share, divided by the number of clients in the cohort. Two steps, both worth understanding.
When a client pays for therapy, most of that money goes to the therapist. We keep 25%. So a cohort that generated £28,500 of bookings shows up here as £7,125. This report never shows the gross figure — it shows what the business actually keeps.
That 25% is then divided by the cohort's headcount, so a 212-client month and a 46-client month can be compared fairly. Every cell is "per one client".
Same data, two questions. The tab you want depends on what you're asking.
Each cell is just that month's earnings. The numbers fall away as clients finish or lapse. Good for spotting a month where retention held up unusually well.
Each cell is the running total. This is the one to read if you're asking "what has this cohort been worth per client by now?" It only ever goes up.
Four different cell treatments, four different meanings. None of them are decoration.
Shading is relative to the biggest cell on screen. It's there so you can see the M1 column dominating without reading a single number.
From this month on, the cohort has earned back what it cost to win — ads plus placement shifts, divided by new clients. Only cohorts from March 2026 have a known cost, so only those can turn green.
Not zero. The cohort simply hasn't lived through this month. A June 2026 cohort has no M5 because May 2026 wasn't in its future.
The month happened and brought in no revenue. This is a real, meaningful zero — worth noticing when it appears early in a cohort's life.
The last two columns hold each cohort's cumulative value at six months old and twelve months old. Sometimes that's a fact. Sometimes it's a forecast. The report always tells you which.
The cohort has completed that month, so this is real money, counted. January 2026's clients were worth £123.62 each by their sixth month.
The cohort hasn't got there yet. This is where it's heading, based on how comparable cohorts grew from the same age. It will be replaced by the real figure in time.
It is arithmetic on past cohorts, not a prediction model and not AI. Three steps:
1. Find the cohorts that genuinely reached the target month. 2. For each one, work out how much it grew from the age our young cohort is now, to the target. 3. Take the middle value of those growth figures and apply it.
Middle value, not average. One cohort that behaved strangely can drag an average a long way. Taking the middle of the six keeps a single odd month from setting the forecast for everyone.
The report names them in its own footnote, so the number is always auditable. It uses the six most recent cohorts that reached the target and had at least 20 clients — the size floor matters because a cohort of four people produces per-client figures that swing wildly.
Recent cohorts only, on purpose. If retention improves, a forecast built on three years of history would be anchored to how we used to perform. A rolling window follows the improvement instead.
A cohort with one month behind it hasn't shown you anything yet — forecasts from that little data swing by tens of pounds. So the columns stay empty until a cohort has two complete months. July and June 2026 show nothing at all; May 2026 is the youngest cohort with a forecast.
This one looks odd until you know why. February 2026's sixth month is the month we're in right now — half-counted. So the grid shows £98.15 of part-month revenue, while the M6 column forecasts ~£101.48. The column is refusing to call a part-month figure "final". Once the month closes, it becomes the real number.
Every row's last live cell is the current month, part-way through. It rises until month end — never read it as final.
Early-2025 cohorts have 5–30 clients. One client booking a long course moves the whole per-client figure. Weight your reading by the Clients column.
Everything past year one is pooled into one cell. You can't see whether it arrived in month 14 or month 30.
The M12 forecasts currently lean on four cohorts from early 2025, when we were a much smaller business. Useful for ranking cohorts against each other; not a number to plan cash against.