You are taking over marketing at a reporting service for small retail chains. At the first meeting, the marketer shows rising conversions; Zhanar, the head of sales, says payments are down; and Talgat, the owner, asks you to double annual revenue. The team proposes switching ad channels, buying a new system and letting AI publish offers. There is no money for all three at once.
Your task is a plan the team can start working on and use to make the next decision. All the numbers and circumstances in the case are made up. The service economics come from the marketing P&L, and we add the following month to them for diagnosis. You will not need access to any ad account.
The brief: two reports count different things
A client pays 60,000 ₸ for a monthly package. Preparing the report and supporting the client costs 20,000 ₸ for each company served. In the first closed month, 180 companies received and paid for the service: ten new and 170 existing. Advertising took 200,000 ₸, and other fixed costs were 6,800,000 ₸.
Suppose that in the following closed month 178 packages were delivered and paid for: eight new and 170 existing. Price, variable cost per package, advertising and fixed costs did not change. There are no refunds or discounts, amounts exclude indirect taxes, and the result is before income tax.
The ad summary shows 30 form submissions and 10 first payments for the first month, and 50 forms and 8 first payments for the second. Someone added these events together and called them conversions: 40 and 58. Payment records have been reconciled with finance, but some acquisition sources have not been recovered yet. Nobody has lined up the full lead-to-purchase windows by channel.
The current team can serve up to 182 companies a month. That is a condition of the case: in real life, Yernar, who runs delivery, would confirm that figure. When extra work has to be paid for, how much cash is on hand and what annual revenue is, we do not know.
Prepare your decision before reading the analysis
Write one or two pages for Talgat. The first part covers what happened to revenue and to the month's result. Then come the first checks: who reconciles which records, what result they hand to the team and what that result lets you do.
The plan must include the budget for the next check, delivery constraints and one AI process. For each stage, name the condition for moving on and a reason to stop. Leave what you do not know as questions to the people responsible for it.
Separately, work through the proposal to spend another 120,000 ₸ on acquisition. How many additional paid and delivered monthly packages would pay for that advertising? Would a scenario with six additional companies fit into current capacity? Future renewals cannot be treated as confirmed.
This is written work for self-assessment. The site does not grade your document or award a CMO qualification. The short quiz reinforces a few principles; the quality of the plan shows in its calculations, assumptions and actions.
Analysis: separate data collection from business results
What grew in the summary is the sum of two different events. The same client can submit a form and then pay, so the sum is not a count of buyers. Going from 40 to 58 is 45%, but calling it sales growth would be a mistake: there were two fewer paid and delivered packages.
In this case, the financial register confirms how much service was delivered. The ad summary helps you find where definitions diverge, but its events by source still have to be reconciled. Switching off a channel just because the totals do not match means deciding before the diagnosis.
Every check has someone doing it. The marketer describes the events and the summary rules, Bekzat, the developer, links the approved records from the website and the CRM, Zhanar clarifies what happened to the leads, Meruert confirms payments and costs, and Yernar confirms how much service was delivered. Everyone works with the same period, clear identifiers and the same statuses.
When you audit digital analytics, you separate the business operation from its record and from whether that record reached the report. If the operation happened and the event was lost, it is the measurement that needs fixing. If the client could not pay, something else needs checking. In our case, it is confirmed that fewer packages were served and paid for, and the cause is still open.
Calculate both monthly results
After variable delivery costs, each package leaves 60,000 - 20,000 = 40,000 ₸. That contribution pays for advertising, fixed costs and the company's result. In a P&L calculation, each cost sits on one line: the marketer's salary is already in the fixed 6,800,000 ₸ and is not deducted a second time.
| Metric | First month | Following month |
|---|---|---|
| Fulfilled and paid packages | 180 | 178 |
| Revenue | 10,800,000 ₸ | 10,680,000 ₸ |
| Variable delivery costs | 3,600,000 ₸ | 3,560,000 ₸ |
| Contribution before advertising and fixed expenses | 7,200,000 ₸ | 7,120,000 ₸ |
| Advertising | 200,000 ₸ | 200,000 ₸ |
| Other fixed expenses | 6,800,000 ₸ | 6,800,000 ₸ |
| Result before income tax | 200,000 ₸ | 120,000 ₸ |
Revenue barely moved: down 120,000 ₸, about 1%. The result fell by 80,000 ₸, or 40%: fixed costs stayed the same, and the company lost the contribution of two packages. The calculation explains why the result shrank, but not why there are fewer clients.
An extra 120,000 ₸ of advertising pays for itself with three more packages: 3 × 40,000 = 120,000 ₸. That covers only the advertising itself at first-month contribution, with no new costs for a launch or for the team's work. And you still have to tell three purchases the ad account credited to advertising from three purchases the advertising actually added.
With 178 current clients, four places are free. A scenario with six new companies makes 184, two over capacity. In unit economics, a change in delivery cost goes into the model before you promise extra volume. And even when income and costs for the period balance, that does not answer whether the cash will last until clients pay.
Sequence the 90 days around conditions for moving on
| Period | Work and owner | What lets you move on |
|---|---|---|
| Days 1-15 | Bekzat and Meruert reconcile operations; Zhanar and Yernar pin down losses and constraints | Shared definitions, an agreed calculation and a list of discrepancies, each with an owner |
| Days 16-45 | The owner of each problem found fixes it; the marketer tries one AI process | A repeat check confirms the fix, and the process evaluation counts quality and all costs |
| Days 46-90 | The marketer and Meruert run a justified growth test | Mature data, an uncertainty estimate and a decision that accounts for cash and capacity |
The periods show an order, not a rigid schedule. On day 46, it may turn out that the agreed purchase window for new leads has not closed yet. Then the report will carry an incomplete observation, and the decision to raise the budget will wait. Work on independent problems can run in parallel.
Every loss you find gets checked again after the fix. Suppose some accepted leads were never assigned to a manager. Bekzat confirms the fix with the routing of new leads and the assignment log. Future sales are not booked as results at the moment of the fix.
Talgat will get three separate messages: what is confirmed, what is fixed and what is still under investigation. A new system is justified when you have found a constraint in the current work and know what it costs to remove. A platform's name does not provide that justification.
Bound the AI assignment
A good pilot is a draft offer based on the approved service catalogue. The model gets current terms, permitted wording and a list of missing information as input. It prepares material for an editor and has no right to change prices or budgets, or to publish the offer.
The editor checks every promise against the catalogue. If the text changes, the new version is the one that gets approved. The process has a person who can stop publication and return preparation to the previous way of working.
The cost of a finished result with AI includes preparing the inputs, every attempt, review, corrections and maintenance. Compare at equal quality, and keep tool invoices separate from estimates of staff time. A freed-up hour is time for another task, not money in the bank.
Defend the conclusion and apply it to your project
A strong answer ties each problem to a check and to the person responsible for it. From the calculation, Talgat sees that the monthly result fell to 120,000 ₸; from the plan, which data are needed before the next spend. Doubling annual revenue stays a goal, but a long-term strategy needs an annual baseline, seasonality, renewals, demand and delivery resources. Two months are not enough for that.
Now prepare the same kind of memo for your own project. One confirmed problem, a calculation of its size in money and the next check are enough. Name the source of every input, the unknowns, the person responsible and the document the decision will rest on. If you have no live project, use the case above and keep its assumptions.
The memo is ready if a colleague can recalculate it and carry it out. If an action reads "improve marketing", it has neither an area of work nor a sign that the work is done. If the budget leans on future renewals, put them in a scenario of their own and explain what it rests on.
Finally, change one premise: in both months 180 packages were delivered and paid for, ten of them new each month, and only the number of recorded first-payment events went down. With everything else unchanged, both months' result is 200,000 ₸. Then the first task becomes restoring measurement and reconciling the missed operations. A plan to rescue falling sales no longer fits this case, although checking the data and assigning an owner is still needed.