Case: forty companies, who is our ideal customer?

When someone says the answer is obvious, treat it as a claim to check, one layer at a time. Leo's claim is that retail is Teams' best segment because it brought the most money. The data: forty companies with first contact from January 2024 to June 2026, twelve of them won.

Start where the claim starts, with revenue. For a won deal, monthly revenue = lessons a month × price per lesson. Retail: 120 × 680 + 150 × 650 + 100 × 700 = 249,100 UAH a month. All twelve won deals together: 484,500. So retail brought 249,100 / 484,500 = 51.41% of revenue.

On revenue, Leo is right, and it's worth saying so plainly. Checking a claim is not the same as trying to refute it. But revenue is only the first layer: the next ones take off delivery cost and then the cost to serve each client. Lost deals answer a different question altogether: not how much, but why.

After revenue come two cost layers. First the delivery cost: 430 UAH to the tutor and 31 for video and AI per lesson. Retail paid 650 to 700 after volume discounts, everyone else 780 or 800. Even after this layer retail still leads: 78,530 UAH a month of contribution against 60,420 for IT services. That's why the trap is easy to fall into: two layers in, the obvious answer still looks right.

Then the cost to serve. Retail's service runs 170 to 210 UAH per lesson; IT services' 30 to 45. Per retail client:

• 680 − 461 − 190 = 29 UAH a lesson × 120 lessons = 3,480;
• 650 − 461 − 210 = −21 × 150 = −3,150;
• 700 − 461 − 170 = 69 × 100 = 6,900.

Together: 7,230 UAH a month of profit from 249,100 of revenue. One of the three, the chain at 650 a lesson, lost money every month it was a client. The layer that changes the answer is the one nobody sees on an invoice.

Do the same for every segment with wins and rank them by profit a month after service cost:

• IT services: 60,420 − 6,650 = 53,770;
• creative studios: 20,340 − 3,040 = 17,300;
• professional services: 18,645 − 2,575 = 16,070;
• retail: 78,530 − 71,300 = 7,230.

The segment that leads on revenue and on contribution comes last on profit. The revenue ranking (retail, IT services, studios, professional services) is the one Leo had in mind; it answers "where does money come in?", not "where does winning more clients pay?".

Two cautions. Don't let size bias you either way: professional services and design studios are small segments, but they keep most of what they earn. And a monthly total depends on how many clients a segment has; per lesson, IT services keep 298.72 UAH and retail 19.54, which says even more clearly where each extra lesson pays.

One comparison makes the case in a single line: put each segment's share of revenue next to its share of profit. All twelve won deals leave 94,370 UAH of profit a month. Retail's 7,230 is 7,230 / 94,370 = 7.66% of it, against 51.41% of revenue. IT services, with 29.6% of revenue, bring 56.98% of profit.

A gap this large between the two shares is the signature of a segment that is expensive to serve or heavily discounted. It doesn't mean the revenue was fake: retail really paid 249,100 a month. It means most of that money went straight back out in delivery and service, so winning more clients like them adds work faster than it adds profit.

The same check is worth running on any "biggest" claim: the biggest client, the biggest channel, the biggest region. If its share of profit is far below its share of revenue, the next question is why, and in Teams' case the answer was the cost to serve plus volume discounts.

Profit is one criterion of six. A strong conclusion should hold on the others too, and here it does:

• Win rate: IT services and creative studios won half their deals (4 of 8, 3 of 6); retail 37.5% (3 of 8); professional services 2 of 6.
• Speed: the median retail win took 120 days; IT services about 45, studios 26.
• Staying: none of the three retail clients is still with Teams. All three bought English as a perk.

When independent criteria point the same way, the conclusion is more trustworthy than any single number. If retail had led on win rate and speed while losing on profit, the memo would have to weigh a real trade-off. Instead, retail is slower to win, less likely to win, less profitable once won and quicker to leave.

That is also how to answer "but they're the biggest names we've had": size is real, and it is the only criterion where retail comes first.

Lost deals answer "why", and the answer decides what goes on the disqualifier list. Logistics: four deals, no wins, and Veloria is logistics. Do you stop?

Read the reasons first: shift schedules twice, no budget owner once, an incumbent renewed once. None of these is "logistics companies don't buy English". Each is a condition you can screen for, or design around: a schedule built around shifts, a known approval path, a pilot that gives procurement something to compare. Banning the industry on four deals would throw away Veloria together with the bad bets.

Other segments show the same method. Manufacturing: perk cut twice, gone silent twice, shifts once. Hospitality: staff turnover all three times, which really is a pattern worth screening for when the people who need English change every season.

So logistics is not in the profile, and not out either: it's a bet outside the profile, taken knowingly with conditions that answer the reasons it was lost before.

A disqualifier needs evidence from the deals, not a hunch. Teams' history supports three:

• English wanted only as a staff perk. All three retail wins churned, and perk buyers were lost again in retail and manufacturing.
• Nobody can name the budget owner or how training is approved. Three deals faded exactly that way.
• The staff who need English turn over every season. All three hospitality deals were lost to turnover.

Two tempting candidates fail the evidence test. Small size is not a problem: the design studios, with 25–40 staff, are among Teams' best clients by profit per lesson and speed. And a quick decision is good news, not a warning sign: studios decided in a median of 26 days and stayed.

Each disqualifier should be checkable on a first call: "What do your people use English for?", "Who approved training last time?", "How long do people usually stay in these roles?" A disqualifier you can't check early saves no time.

The write-up turns the analysis into choices. Drop retail as a perk for store staff, not retail as a word: a retail head office with a real work need, like Orchid's buyers and import staff, is scored on its own facts, at the list price. Say how sure you are: twelve wins is thin, so mark the confidence of each line and set a review date.

For a busy reader like Leo, structure the memo as decision first, then the numbers that carry it, then what you don't know yet. Show revenue and profit side by side: hiding revenue looks like hiding something, while showing both is what makes the case (51.41% of revenue, 7.66% of profit).

How it fits together: this module built a profile from your own deals, turned it into a score using profit after service, and split time by tiers; this case tested all three on forty real companies. Next module moves to the deals themselves: a pipeline whose stages mean something.