Case: reposition Klasna Club

When customers leave an offer saying "something free does this", the first question is not "how do we compete with free?" but "which shelf have we put ourselves on?". The frame of reference from the first lesson of this module works in reverse too: an offer filed on the wrong shelf is compared with the wrong alternatives.

Klasna Club is a monthly subscription for adults with live speaking clubs and an AI practice partner. Its page headline is "Your AI English partner, 24/7", the picture is a phone with a chat, and the live clubs are described in the last section. Every loud signal points to the chatbot shelf. On that shelf, free chatbots are the benchmark, so a paid Club looks like a chatbot that costs 450 UAH a month.

That is the diagnosis: the page sells the part of the offer with the cheapest neighbours and hides the part members value. Before deciding anything about the Club, check whether the shelf is the problem.

An exit question asks people why they are leaving. Read its answers by what each one asks you to fix, not just by size.

Since the July launch, 110 of 590 Club joiners have cancelled, and 48 answered the exit question:

• "A free AI chatbot does the same": 21, or 21 ÷ 48 = 43.75%. A shelf problem: they judge the Club as a chatbot.
• "The club times don't fit my evenings": 11, or 22.92%. An offer problem: the live part exists but doesn't reach them.
• "I stopped learning English": 9. Neither shelf nor offer; no marketing change will win these back.
• Other: 7.

So the biggest group points at perception, the second at the offer. A fix that only changes the headline would miss the 11; a fix that only adds times would leave the 21 comparing the Club with free chatbots.

Remember who answers: 48 of 110 who left, people willing to say why. It is a strong signal, not a census.

Exit answers say why people leave. To see what keeps people, look at those who stayed and compare groups.

The team took a sample of 40 members who joined in July and split them by how they started:

• 16 went to two or more live clubs in their first month; 13 of them were still members three months later: 13 ÷ 16 = 81.25%.
• 24 went to fewer; 8 of them were still members: 8 ÷ 24 = 33.33%.

Always compare shares, not counts. "13 stayed versus 8 stayed" hides that the groups have different sizes; 81.25% against 33.33% shows the gap honestly.

The gap is large, and it fits the exit answers: people who leave say the AI alone isn't worth paying for, and people who stay are the ones speaking with other people. What the comparison can't tell you yet is why the gap exists. That is the next card's question.

81.25% against 33.33% is tempting to read as "live clubs keep people", and Uliana does: "push everyone into clubs". Slow down.

The comparison is a correlation: two things that go together. There are at least two stories that fit it:

• the live clubs make people stay; or
• motivated people both go to clubs and stay, and would have stayed anyway.

With 40 people and no comparison group, the data can't tell these apart. Making clubs compulsory on the strength of it could annoy members and change nothing.

That doesn't make the sample useless. The gap is large and points the same way as the exit answers, so it is worth acting on carefully: as a hypothesis to test, not a law to enforce. A cheap test would nudge one group of new members into a live club early and not the other, then compare who stays. Only that would separate cause from coincidence.

Leo's instinct is common: "we can't beat free, so let's close it before it costs more". It is right about one thing: on the chatbot shelf, a paid Club can't win. It is wrong to stop there.

Put the evidence together. The biggest group of leavers compares the Club with free chatbots, which is a shelf problem caused by Klasna's own page. The second group can't make the club times, which is a fixable offer problem. And members who speak with people early stay far more often. None of this says the offer is unwanted; it says the wrong part of it is being sold.

So the decision is not "close or keep pushing" but "move it to a shelf where it can win, fix what stops people using the valued part, and test". Closing stays a real option, with a condition set in advance: if the repositioned Club still fails the test, close it. That is also what Leo asks for: what would make you close it after all?

Repositioning means moving an offer to a different shelf in customers' heads. There are two levers, and most real cases need both.

• The perception lever changes how the offer is understood: headline, picture, the frame you name. For the Club: the shelf "live speaking practice in small groups", with a headline and picture about speaking with people. On that shelf the Club is compared with conversation tutors and language meetups, not free chatbots.
• The offer lever changes the offer so it lives up to the new shelf. For the Club: club times in the evenings that fit working adults, and a first live club booked in each new member's first week, so newcomers meet the part worth paying for.

The AI partner stays, described honestly: practice between sessions, not a replacement for people. Two moves fail: boasting about the chatbot ("the smartest chatbot in Ukraine") keeps the Club on the old shelf and can't be proven, and hiding the live clubs throws away what members value.

Before rebuilding pages and timetables, test whether the new position works. A cheap test needs four things:

1. People who agreed to hear from you. Adult learners on Klasna's email or messenger list who consented to messages. Not a bought list of phone numbers: those people never agreed to marketing messages.
2. Two versions. Half get the current invite, half the live-first one.
3. One metric with a threshold set in advance, and a fixed period, such as four weeks. Decide what result counts as success before you look, so the data can't be read to fit a hope. Also say how many people you need before a difference is worth trusting; the statistics course covers how.
4. Honest copy. No fake scarcity ("only 3 places left" when there are many), no new price claims, and the AI described as what it is.

The test is small on purpose: it answers one question, whether the new shelf brings people to the valued part, before the team spends on everything else.

The test needs one metric, and it should measure the thing the new position is about. The Club's new shelf is live speaking practice, and the evidence says members who speak with people early stay. So the metric is: who joins and books a first live club within 14 days of the invite.

Weaker choices are easy to collect but answer a different question. Email opens show the subject line worked. Page views and likes show curiosity. Sign-ups alone count people who may never reach a live club, which is exactly the pattern that ends in "a free chatbot does this". Each can rise while the position fails.

One metric fixed in advance also keeps the decision honest: if the live-first invite beats the threshold, roll it out; if not, Leo's condition applies.

How it fits together: diagnose the shelf from the page and exit answers, read the sample as a hint, move the Club with both levers, and test with one metric. The next module asks how people actually find Klasna and buy, and what each channel really costs.