A dashboard where a brand tracks its cashback campaigns in banks on its own: sales, customers, budget spend and payback. Before, all of it arrived as Excel files from an account manager.
The company runs brands' cashback campaigns in banks as performance advertising: the brand pays for a completed purchase, not for impressions.
The case is built on a real campaign: 25% cashback on Uvelka rice.
Results arrive in Excel: a weekly snapshot and a final report of 7 sheets.
There is no single picture: channels, segments, receipts and finance are spread across files.
The brand learns about overspend after the fact: the ₽3 million limit ran out on 27 January, and ₽86,299 went over it.
CAC (Customer Acquisition Cost), ROMI (Return on Marketing Investment) and ROAS (Return on Ad Spend) are calculated by hand; every report is a manager's work.
Brand manager, the main user: keeps the campaign under control and reports to management.
Marketing analyst: slices, exports, details.
Brand finance: investment and payback.
The brand's agency: limited access.
Self-service: answers without asking a manager.
A transparent budget and alerts on the limit.
Proof of effect: new customers and payback.
Less manual reporting and repeat sales of campaigns.
Hypothesis 1. A brand manager needs answers to three questions rather than a set of charts: how much has been spent, who bought, did it pay off. We build the sections around the questions, not around Excel sheets.
Metric: time to answer a question from management; the share of brands that open the dashboard on their own.
Hypothesis 2. Overspend can be prevented by showing the spend rate and the forecast of when the budget runs out, not just the total.
Metric: the share of campaigns that end without going over the limit.
Hypothesis 3. A brand will stop asking a manager for reports if it can check every metric by its formula and export the data in a familiar form.
Metric: the number of report requests to managers before and after launch.
Hypothesis 4. Receipt data — new customers, sales by retail chain and SKU — is the main argument for a repeat campaign.
Metric: the share of brands that request a new campaign from the dashboard.
Hypothesis 5. An honest negative ROMI next to the annual ROI via CLV earns more trust than picking convenient metrics.
Metric: we check trust qualitatively: questions and disputes about the finance report.
One long dashboard. Seven report sheets don't fit on one page without losing legibility. The overview answers the main question in 10 seconds, and mini-slices lead to the detailed sections.
Moving the Excel sheets over one to one. The screens are built around the brand's questions; the structure of the original files is kept only in the exports of the Reports section.
Only convenient metrics. We could have shown ROAS and hidden the negative ROMI for the period. We showed both, with the annual ROI via CLV next to them.
Open customer data. The customer ID is hashed: the brand sees the receipt and the chain, but not personal data.
An alert only at 100% of the limit. It reports overspend but leaves no time to act. We set thresholds at 80%, 95% and 100%.
Took apart two reports: 7 sheets, 94,748 receipts.
Built a metric tree: volume → efficiency → structure → transactions.
Grouped the sections around the brand's questions: “how much has been spent?”, “who bought?”, “did it pay off?”.
Designed 11 screens on shadcn design system components.
Built a working coded prototype as the base for development.
01
The brand's home screen: every cashback campaign and how it performs, on one page
Checks hypotheses 1 and 4
02
The campaign's main screen answers “how did the campaign go?” in 10 seconds
Checks hypotheses 1, 2 and 3
03
How the campaign developed over time and when the budget will run out
Checks hypotheses 2 and 3
04
Comparing partner banks and offers: where cashback works best
Understand what each bank and targeting segment contributes, to reallocate the budget in the next campaign.
Checks hypothesis 4
05
Who bought during the campaign and whether they will come back: proof of new customers
Show management how many new customers the campaign brought, and assess their loyalty.
Checks hypothesis 4
06
Where and what people buy: sell-out by retail chain and SKU from fiscal receipts
Get data for trade marketing: which chains and SKUs drive sales, and when people buy.
Checks hypothesis 4
07
A report for the finance team: how much went in, what came out and whether the campaign paid off
Justify the budget to finance and management, and compare costs with results.
Checks hypotheses 3 and 5
08
Down to a single transaction: 94,748 fiscal receipts with cashback
Check the data at the purchase level, look into a disputed case, and reconcile accruals with the brand's own records.
Checks hypothesis 3
09
Self-service instead of emails with Excel: an archive, an export builder and scheduled delivery
Get data in a familiar format for internal reports without waiting for a manager.
Checks hypotheses 2 and 3
The dashboard has launched. There is no quantitative measurement yet, so the impact is shown on the campaign's own data: how the alerts would have worked had the brand had the dashboard from day one. The ₽3 million limit ran out on 27 January.
The weekly Excel from a manager has been replaced by self-service: 11 screens, exports in the format of the original reports and scheduled sending.
Hypothesis 2: the share of campaigns that end without going over the limit.
Hypothesis 3: the number of report requests to managers before and after launch.
Hypothesis 4: the share of brands that request a new campaign from the dashboard.
Showing the dashboard to brands brought no feedback, so the first improvement is to collect it inside the product.
— A “Was this section useful?” rating and a short survey after a report export.
— Product analytics: which sections are opened, what is exported, where people drop off.
— An action right in the alert: “Request a higher limit” or “Pause the offers”.
— Comparison with the brand's past campaigns, so the decision on a repeat campaign rests on the trend.
— A weekly email digest for management, who don't open the dashboard.
— Hints on the CAC, ROAS and ROMI metrics on first visit.