Attractiveness and ability to win

A segment is a good first target only if it's worth having and you can win it. So score two sides and read them together.

Segment attractiveness asks what a segment is worth to anyone who wins it: its size and contribution, how steady it is through the year, how contested it is.

Ability to win asks whether you can serve it better than its alternatives: the share of enquiries that become a first package, marketing spend per new family, whether you can prove what these families value most, and whether you can reach them.

Uliana's favorite, exam families, scores high on size and money; that's one criterion on one side. A big segment can still be a poor first target if everyone fights over it, if its work piles into a few spring months, or if Klasna can't show why it's the better choice.

A quick test for any criterion: would it be the same number for a rival school? If yes, it's attractiveness; if it depends on Klasna, it's ability to win.

Before any points, write for every criterion which way is good. It sounds obvious; it's the step most often skipped, and the one that most often flips a result.

On Klasna's sheet:

• Contribution: higher is better.
• Share of the year's lessons in the spring peak: lower is better. Steady work through the year is cheaper to staff than a rush of tutors for three months.
• Contested (share of families who named a paid alternative at sign-up): lower is better. Each family is easier to win.
• Share of enquiries that become a first package: higher is better.
• Marketing spend per new family: lower is better.

Half the criteria point down. That's why the direction belongs on the sheet, right next to the name, in words: "lower is better, because…". Anyone who later reads or edits the sheet sees at once which way each arrow goes, and why.

To turn a percentage into points, write bands first, and write them before you look at the segments. Then nobody can stretch a band to suit a favorite.

Klasna's bands for "contested" (the share of a segment's families who named a paid alternative at sign-up; lower is better):

A segment at 48% is above 45% and not above 55%: 3 points. Because lower is better on this criterion, the points fall as the percentage rises.

Two habits keep bands honest: make the edges unambiguous ("up to 45%" includes 45%), and keep the same bands for every segment, even when a favorite lands one point short.

Weights say what matters most: shares that add up to 100%, each with a reason, set before the points. A weighted score multiplies each criterion's points by its weight and adds them up.

A quick sheet with three criteria: size 40%, steadiness 30%, contested 30%.

• Exam families (5, 1, 2): 0.4 × 5 + 0.3 × 1 + 0.3 × 2 = 2.9
• EU families (2, 5, 4): 0.8 + 1.5 + 1.2 = 3.5
• Adults paying for themselves (1, 4, 5): 0.4 + 1.2 + 1.5 = 3.1

A common slip is the plain average: for the adults, (1 + 4 + 5) ÷ 3 = 3.33. It ignores the weights, so size, the criterion weighted most, counts no more than the others.

Weights are judgement, not data. That's fine as long as they sit on the sheet where anyone can argue with them, each with its reason, and are fixed before anyone sees which segment they favor.

A single criterion scored in the wrong direction can choose the plan's lead segment on its own. Uliana's first sheet gave more points for more rivals on "contested".

With weights size 40%, steadiness 30%, contested 30%:

Same data, same weights, one reversed arrow, and the ranking flips: exam families move from second to first.

This is why the direction sits next to every criterion, in words, with its reason. It's also why a result that rests on one criterion deserves a second look: if a single line decides the winner, check that line twice before the plan relies on it.

After scoring, run a sensitivity check: move each weight by five or ten points and see whether the winner changes.

If it doesn't, the choice is robust: it doesn't hinge on the exact weights anyone picked. If it does, the scores are close, and the sheet should say so. Close scores mean the choice needs judgement, not more decimals. Adding a decimal place until one segment "clearly" wins is false precision: the weights themselves are judgements accurate to maybe ten points, so a difference of 0.05 means nothing.

What to write instead: "EU families and exam families are within 0.2; with steadiness at 25% exam families lead. We choose EU families because…", followed by the reason that tips it (capacity in spring, proof Klasna already has, a risk the team wants to avoid).

And never the reverse: adjusting weights until a favorite wins turns the sheet into decoration for a decision made in advance.

Each fact should enter the score once. "Lots of rivals proves the demand is real" sounds reasonable, and it double-counts.

Demand is already scored in size and contribution. "Contested" measures something else: how hard each family is to win, because it's comparing you with other paid options. If rivals also earned points, the same demand would be counted twice, and the difficulty of winning would disappear from the sheet.

So on "contested", fewer rivals earn more points. In this lesson, reversing that one criterion was enough to flip which segment led the plan.

The general check: for every criterion, ask what it measures that no other criterion already measures. If two criteria rise and fall together for the same reason, merge them or drop one.

An AI assistant can draft scoring criteria quickly. Each line then passes four checks before it goes on the sheet:

1. Can you measure it from your own data? Contribution per segment and the share of enquiries that become a first package can stay.
2. Is it about need and behavior, never a sensitive trait? A gender split per segment comes off.
3. Do you market to parents, not children? "How easily ads reach 15-year-olds directly" comes off: ads aimed at minors are off limits.
4. Does every number have a source? A market size with no source is a guess; it waits until checked.

And the data rule: ask the assistant about the method, keep the customer rows at home. Personal data goes into an outside tool only with a lawful basis.

How it fits together: score attractiveness and ability to win; set directions, bands and weights before the points; count each fact once; test the weights; check AI drafts line by line. The next lesson takes the same method to Dana's practice: a first niche her hours can fill.