Goals by step, tools by goal

A campaign calendar (exam ads from February, a restart push in September, posts whenever someone has an hour) is a list of activities. It says what the team will do, not what each activity is for, so nobody can tell afterwards which parts worked.

Start from the path to a purchase instead, and give each step one goal you can check. For exam families:

• parents of 10–17-year-olds know what Klasna is for;
• they believe the weekly progress note is real;
• they book a trial;
• they buy a second package.

A good goal names the people, the step and a change you could observe. "Raise brand awareness" or "do more marketing" fail: they name no step and no change, so any activity can claim to serve them.

Once each step has its goal, every later choice gets a test: which goal does this ad, post or email serve, and how will we know it did?

With a goal per step, pick the tool for each goal from four families:

• Paid: ads and paid creator posts. They buy attention, fast.
• Owned: your site, the email list, chats people joined. They carry proof and repeat purchases.
• Earned: reviews, a teacher's recommendation, a mention you didn't pay for. They come from customers who are satisfied.
• Shared: things people pass on, like a parent forwarding Klasna's exam checklist.

For exam families: parents of teens hear what Klasna is for through a labelled paid post by a creator whose audience is parents; they believe the note is real from a sample note on Klasna's own site; searching parents book a trial through search ads on exam-prep queries; families buy a second package thanks to progress notes and an email to parents who opted in; satisfied families tell others when asked for a review after a finished package.

Picking the tool after the goal keeps the mix from turning into "more ads" for every problem.

"Parents know what Klasna is for" sounds hard to count. The fix: give each goal one number you already collect.

• Know what it's for: new enquiries that mention the progress note in the sign-up question.
• Believe: parents who open the sample note and then book a trial.
• Book a trial: trials booked from search visits.
• Buy again: second packages within 90 days, the measure the last module used to compare channels.

Then each tool is judged on its own step first, and on paying families in the end. A creator post won't sell packages by itself, so judging it on sales would always call it a failure. But if parents who saw it don't enquire, or never mention what it said, it didn't do its job either.

This also prevents the opposite mistake: crediting every sale to whichever tool happened to be last. The number at each step shows where the path works and where it breaks.

Some steps of the path happen all year: someone searches for a maths tutor in November too, and packages end every week. The tools that serve them should be always-on: search ads and a clear page for parents searching for an exam tutor, the site, reviews requested whenever a package ends, a regular newsletter.

Other moments are peaks, when many families decide at once: the exam season and the September restart. That is what campaigns are for: a two-week burst of creator posts before the restart, a spring push in parents' community newsletters, a September email to last spring's families who opted in.

The rule of thumb: if the need can arise on any day, keep the tool on; if decisions cluster in a few weeks, add a campaign there.

The two layers work together. The always-on base catches people whenever they look and keeps proof available; the campaign adds weight where many decisions happen at once. A plan with only campaigns goes dark between peaks, and a plan with no campaigns misses the moments that matter most.

Leo's version is simpler: "put everything into September; that's when people buy." Two things are wrong with it.

Part of September buying would happen anyway. About 30% of spring families come back in September on their own. A budget that piles into September takes credit for families who were returning regardless, and the plan overstates what the money did. (Measuring that properly is the subject of a later module.)

Going dark costs the rest of the year. If search goes quiet in October, the parent who looks then finds someone else. Reviews stop being collected, the newsletter stops, and proof goes stale before the next peak.

So the answer to Leo is not "spread everything evenly" either. It is: keep the always-on base running all year, and put the campaign on top where decisions cluster, with its own goal for the steps it serves. September deserves a campaign; it doesn't deserve the whole budget.

Every message passes a few rules before it runs.

• Parents are the audience, teens the users. That is Klasna's house rule, and as of October 2026 the law points the same way: Ukraine's advertising law (Art. 20) limits advertising aimed at children, EU consumer law bans ads that directly urge children to buy or to get their parents to buy, and EU platform rules stop profiling-based ads to users the platform knows are minors. A creator video for 15-year-olds saying "Get ahead of your class, use code TOP" runs into all three.
• Paid posts are labelled as paid.
• Reviews come from real families, quoted with their permission.
• Urgency only if true. "Only 3 places left" on a page where places aren't limited misleads, and a discount needs a real previous price; the Pricing course covers those rules.
• Marketing emails go to parents who opted in, with an unsubscribe link.

These are guardrails as of October 2026: check them against current sources before each campaign.

The four families of tools are defined by how the message got there, not by how it looks. A creator post that Klasna paid for is paid media, however natural it sounds.

The same holds for anything that ties the creator to Klasna: money, free lessons, a gift, a family member on the team. If there is a tie, the post is paid or sponsored and must be clearly marked as such.

Why it matters: viewers weigh an honest recommendation differently from an ad. A post that hides its payment borrows the trust of earned media without having earned it. That misleads viewers, and advertising rules treat hidden advertising as a breach. It also poisons the real earned media around it, because once people learn some recommendations were paid, they discount all of them.

Earned media is what happens without payment or ties: a parent's review, a teacher's recommendation, a mention nobody arranged. You can ask for it (a review request after a finished package), but you can't buy it.

Ostap asks an AI tool for twenty ad variants a day. That's fine for drafts: variety is cheap now. What is not cheap is a variant that breaks a rule, so a person checks each one before it runs:

• Can we prove every claim? "Best results in Ukraine" fails.
• Is it aimed at parents, not at teens?
• Is it in Ukrainian for families in Ukraine?

A variant that fails any check doesn't run, however well it tests. Testing tells you what gets clicks, not what is true or allowed; an ad that wins a test by promising too much is the one that breaks trust fastest.

How it fits together: set one checkable goal per step, pick the tool for each goal from the four families, judge it by its step's number, keep an always-on base with campaigns where decisions cluster, and pass every message through the same rules. The next lesson chooses between media options when the cheapest views reach the wrong people.