Scoring segments and accounts: profit, not revenue

Account scoring turns the profile's criteria into a number, so companies can be ranked instead of argued about. Marta has twenty enquiries and time for five conversations: a score tells her which five to look at first.

The first step is the direction of each criterion. For some, more is better: more staff who use English at work, more profit per lesson. For others, less is better: a shorter sales cycle, a lower service cost per lesson, a smaller share of booked lessons nobody showed up to.

It sounds obvious, but skipping it is a classic error. If "sales cycle" is scored like "staff who need English", a company that takes 150 days to decide quietly collects top points, and the ranking rewards exactly what slows Teams down. Write the direction next to each criterion before anyone gives a single point.

After the direction comes the scale: written bands that turn a fact into points. For Teams' sales cycle: under 30 days, 5 points; 30–60, 4; 60–90, 3; 90–120, 2; longer, 1. A company that took 75 days scores 3, whoever does the scoring.

Without written bands, everyone's "4" means something different. One person gives a 4 to a company they like; another gives a 3 to the same facts. Then the score reflects moods, not accounts, and nobody can explain a number when Marta questions it.

Bands also have to be fixed before scoring. If you write them while looking at a favourite account, the scale bends to fit it: the band edge moves to 80 days just so that company gets a 4. Written in advance, the bands let anyone recheck any score, which is what makes the sheet trustworthy.

Criteria rarely count the same. Weights say what matters most, as shares that add up to 100%. Teams' current choice: profit per lesson 30%, a work need for English 25%, win rate 20%, speed 15%, size of the English team 10%.

The score is the sum of weight × points. A company with 5 on profit, 5 on need, 4 on win rate, 3 on speed and 2 on size scores 0.30×5 + 0.25×5 + 0.20×4 + 0.15×3 + 0.10×2 = 4.2 out of 5.

Two shortcuts give wrong answers. A plain average of the points treats size like profit, so a big company with weak economics climbs the ranking. A plain sum (13 points, say) isn't on the 1–5 scale at all.

Weights are a judgement, so they must be written down with a reason and tested later. Leo would put deal size at 50%. That's allowed, as long as the reason is visible and the team checks after a quarter whether high-scoring accounts actually won and stayed.

The money criterion is profit per lesson after service cost, not revenue. Revenue is what a client pays; what Klasna keeps is smaller, and it shrinks again by how much work the client needs.

The chain at Teams' list price: 800 UAH a lesson, minus 430 to the tutor and 31 for video and AI, leaves 339 UAH of contribution. Then the cost to serve comes off: coordinator time for scheduling and reports, plus lessons paid to tutors when learners don't show. An IT services client paying 800 with 40 UAH of service per lesson scores on 800 − 430 − 31 − 40 = 299 UAH.

The usual slips are in the subtraction: stopping at 339 forgets service, subtracting only service gives 760, and adding service instead of subtracting it gives 369. Writing the formula once, price − 430 − 31 − service, keeps every account comparable.

Big revenue can hide a loss, which is why revenue can't be the money criterion. Take one past retail client of Teams: 150 lessons a month, the biggest client Teams ever had. After a volume discount it paid 650 UAH a lesson, and serving it (rescheduling shift workers, reports per store, tutors paid for no-shows) cost 210 UAH a lesson.

The chain: 650 − 430 − 31 = 189 UAH of contribution, minus 210 of service = −21 UAH on every lesson. At 150 lessons a month, that's 150 × 21 = 3,150 UAH lost every month it stayed.

On a revenue view this client was the star: 97,500 UAH a month. On a profit view it was the worst account Teams had. Both numbers are facts; only one belongs in the score. That's the message for anyone who asks why a famous name ranks low: show the whole chain side by side, revenue, contribution, service cost, profit, and let them see where the money goes.

Score bottom-up: from facts about each company and each past deal, then add the accounts up into segments. Not top-down from "retail is a huge market", which says how big a segment is, not how Teams does in it.

For money, add up the money: a segment's profit per lesson is its total profit divided by its total lessons. An average of the accounts' per-lesson figures lets one tiny, great client hide big losing ones. Example: a small account with 10 lessons at 300 UAH profit each and a big one with 150 lessons at −21. Average of averages: (300 − 21) / 2 = 139.5 UAH, which looks healthy. Summed money: (3,000 − 3,150) / 160 = −0.94 UAH a lesson: the segment loses money.

A segment with no wins, like logistics, gets what you know and "unknown" for the rest: no wins in four tries is a low win-rate score, not a verdict on the industry, and there is no profit to score yet.

It is tempting to rank segments by monthly revenue, on the theory that bigger clients usually earn more. In B2B training they often don't. Big clients tend to come with volume discounts, which cut the price per lesson, and with higher service costs: many locations, shift schedules, more no-shows, separate reports. Both squeeze profit from opposite ends.

Teams' own history shows it. Retail brought the largest share of revenue of any segment, yet the retail clients left little or nothing after service, and the biggest of them lost 21 UAH a lesson. IT services brought less revenue and far more profit per lesson.

So the score uses profit per lesson after service cost, summed bottom-up from the accounts. Revenue still matters as context: it shows size and how much depends on one client. It just isn't the measure of whether winning more such clients would help.

Account scoring follows a fixed order, and the order is what keeps it honest:

1. Choose the criteria from the ideal customer profile.
2. Set the direction of each: is more better or worse?
3. Write the bands of the scale for each criterion.
4. Set weights that add up to 100% and note why.
5. Score each account from its data.
6. Add the accounts up into segments and rank them.
7. Each quarter, compare scores with real wins and losses.

Direction and bands come before any points, or the scale bends to fit the account you like. Weights are set and justified before scoring, for the same reason. Accounts are scored from data, and segments are built from accounts, never the other way round. The quarterly check closes the loop: if high-scoring accounts don't win or don't stay, the weights or bands are wrong and get adjusted. Marta's request fits this design: she wants the sheet, not just the number, so that if she disagrees with a 3 she can see which band it came from.

A high score means look first, not sign first. The score ranks companies by how well they match the profile; it doesn't replace the checks. A top-scoring company with no budget owner still fails a disqualifier from the profile, and a month spent chasing it is still a month lost.

So use the score to order Marta's five conversations, then let each conversation test what the score can't see: who owns the budget, how approval works, whether the need is real today. When the facts from a call contradict a score, update the row; the bands make it clear which point changes and why.

How it fits together: scoring sets each criterion's direction, writes the bands, weights the criteria openly, scores money as profit per lesson after service, builds segments bottom-up and is rechecked against what happened. Next lesson: once accounts are ranked, how much effort each one deserves, which is what account tiers decide.