Case: name Klasna's segments

Write each segment as a rule you can run on your data, then give it a plain name. The rule makes the segment checkable and repeatable; the name makes it memorable.

Klasna's five rules split the 160 accounts in the export exactly, with no account in two groups and none left over:

40 + 24 + 54 + 16 + 26 = 160. When rules split a file exactly, every number in the row adds up and nobody is counted twice. A name such as "the go-getters" with no rule behind it can't be checked, and slowly drifts to mean whatever the speaker wants.

Once a segment is written as a rule, finding it is a filter, and anyone can repeat the count. Recall: a segment is a group that needs, buys and behaves alike.

The rule for Ukrainian families whose goal isn't the exam: payer = parent, country = ua and goal other than national_exam. In Klasna's export of 160 accounts it returns 54 accounts (33.75%), the largest single group. They're younger learners, aged about 10 to 14, with goals school_grades (41) and coding_hobby (13).

Their behavior: about 4.22 lessons a month for 7.26 active months on average, and 37 of 54 (68.52%) still active in September 2026. Their main influencer is different from the exam groups' too: friends and family, not school teachers.

A plain name for the plan follows from the rule and the behavior: "younger-learner families: about a lesson a week, most of the school year". The name is for Leo; the rule is what keeps it honest.

Measure each segment on staying, not only on size or pace; a group that's quiet in ads can be one of the most loyal.

The rule country = eu returns 26 accounts, Klasna's EU diaspora families. They take only about 4.21 lessons a month, but stay 9.58 active months on average. Still active in September 2026 (last_lesson = 2026-09): 18 of 26 = 18 ÷ 26 = 69.23%, the highest of any group. Compare the spring crammers: 30%.

Why it matters for the plan: retention changes what a group is worth and what it needs. A family that stays most of the year values steady scheduling and continuity with the Ukrainian program; it's reached through word of mouth (half of them first heard from friends or family) rather than through urgent seasonal ads.

A plain name: "diaspora families: slow and steady, the most loyal group". Whether they come first is a later choice; making their loyalty visible is this row's job.

When an interview note seems to fit a group, check it against the same rules that define the segments. Impressions drift; rules don't.

Lesya's note: her son's math teacher mentioned Klasna in January; they started in February, three lessons a week; she wants to see every week whether it's working; they'll stop after the exam. By the rules: goal = the national exam, first lesson in February (2026-01 or later) → spring exam crammers.

The rest of the note agrees with the group's numbers, which is a good sign the rule captures something real: crammers average 9.71 lessons a month (three a week is about twelve), 28 of 40 are gone by May or June, and 16 of 40 (40%) first heard of Klasna from a school teacher, exactly like Lesya.

She isn't a steady exam family (they started by December), not a younger-learner family (goal isn't the exam), and not diaspora (they live in Ukraine).

Every dataset describes one population, and a finding holds only inside it. Before you generalize, say what the file covers and what it leaves out.

Klasna's export is a sample of 160 lesson accounts whose first lesson fell between September 2025 and June 2026. In it, all 10 Klasna Club members are adults, all among the 16 adults paying for themselves. That's a true statement about lesson accounts.

It isn't a statement about the Club. Since its July 2026 launch, most of the Club's members take no lessons at all, so they never appear in this file. "The Club is for adults", "half the Club is adult" or "the Club has 10 members" would all stretch the sample beyond what it holds.

The same habit protects every row of the plan: write the population next to the number ("of 160 lesson accounts"), and when a question is about a group the file doesn't cover, find the data that does.

An export counts what people did; an interview tells you why, under which constraints and where. Use the notes for exactly what the data can't hold.

Arsen's note (family in Poland, daughter 12) adds three things the export can't show:

• A constraint: lessons have to be after five in the evening, their time. That shapes the timetable and the offer.
• The exact place word of mouth happens: a Ukrainian parents' chat in their town. The export only says friend_or_family; the note names a place you could reach honestly.
• The reason behind the goal: they may go back one day, and then she'll need Ukrainian math and the national exam.

What the note doesn't add: the number of lessons or the country, which the export already has.

And its limit: one family's words are a lead to check, not a fact about all 26 diaspora families. A few more conversations, or a question added to the sign-up form, turn the lead into evidence.

The segments row has one job: make every real group visible and honest. Ranking them, and choosing whom to serve first, is a separate decision with its own inputs, and it comes in the next row.

Mixing the two distorts both. Leaving a group out because it's small decides the target before anyone has compared attractiveness or Klasna's ability to win. Klasna's adults paying for themselves are only 16 of 160 accounts, yet they're clearly different from every family segment: one person uses, pays and decides; they find Klasna by searching; and all 10 Club members in the export are among them. That difference earns a line.

A complete row for each segment: a plain name, the rule, the size, need and behavior with numbers, who uses, pays and influences, and one line of guardrails (situation and need, never sensitive traits; age to describe, never to aim ads at minors). Plus what three interviews can't prove.

How it fits together: rules find segments, numbers size them, notes explain them, the population limits them, and the row names them all. Next module: whom Klasna serves first, and why the biggest group isn't automatically the answer.