Qiwi Partner: Recruitment via Paid Advertising — From Test to System

Qiwi Partner is a Polish platform connecting drivers to taxi services and couriers to delivery companies. It operates in Warsaw and other Polish cities — Katowice, Poznan, and others. The target audience is specific: predominantly male migrants looking for stable work with clear conditions and real income. That’s why ads were run simultaneously in three languages — Ukrainian, Russian, and English.

The business is seasonal. Peak season runs September–November; low season is winter. This directly shaped the strategy: during peak season, budgets grew to 80–100k PLN per month. In winter — they were cut in half.

The task was not simply to drive traffic to the website. It was to find people who would actually show up for work.

The client’s requirement was clear:

“Our KPI is +500 new active clients by the end of December — meaning those who stayed on and worked with us.”

Not leads in a spreadsheet — but people who showed up for their first shift.

To achieve this, the team had to solve several non-trivial challenges simultaneously:

• Work with three language audiences without losing relevance;
• Scale advertising during peak season without a sharp rise in CPL;
• Correctly allocate roles across three channels — Meta, Google, and TikTok.

Each channel had its own logic. Finding it — was the core task at the start.

Recruitment is not e-commerce, where someone sees a product and buys. Here, between the first contact and showing up to work, there are several steps, doubts, and trust-building. So the strategy was built not around a single channel, but around a system of three.

📌Meta Ads

The launch began with two campaigns — for drivers and couriers. From the very start it became clear: these are different audiences with different behaviors, so campaigns were immediately separated and optimized independently.

Simultaneously, the team tested formats, messages, and audiences. What worked — was scaled; what generated no leads — was turned off within the same week. Over time, they added English-language creatives, a separate car rental campaign, and a B2B segment test for experienced drivers working as entrepreneurs. Budget was redistributed weekly based on results and seasonal demand.

📌Google Ads

Google operated on a different logic: we don’t search for the audience — they search for us. Someone already typed “courier job in Warsaw” — the goal was simply to appear first in the results. They launched Performance Max campaigns separately for drivers and couriers, later adding English-language search campaigns and a standalone car rental campaign. CPL here was higher than on Meta and TikTok, but lead quality was better. That’s why the Google budget was eventually brought up to match Meta’s, and scaling continued.

📌TikTok

TikTok was used as a hypothesis — cheaper reach to a younger audience that’s looking for work but hasn’t yet moved to Google search.

TikTok produced the lowest CPL but couldn’t stably generate high volume. A strong supplementary channel — but not the primary one.

Once the strategy across all three channels was validated, the client signaled a push to scale:

“Our busy season is kicking off and we’ve decided to significantly increase budgets across all channels. For September, that’s 50,000 PLN.”

This marked the turning point from testing to systematic volume-driven work.

Every week — a separate report by channel: what worked, what was turned off, what’s planned. Three simultaneous channels meant transparency was especially critical: the client needed to understand where every unit of budget was going and why.
All changes — went through approval. Nothing was launched without confirmation.

• Lead quality across channels. Google produced more expensive leads, but a higher activation rate. Meta — cheaper and more volume, but lower quality. So channels weren’t compared purely on CPL — they were evaluated on activation rate, and budget was distributed accordingly.
• Scaling hurts efficiency. The bigger the budget — the higher the CPL. This was addressed through regular campaign structure reviews, turning off underperforming combinations, and reallocating to what was working.
• Three language audiences simultaneously. Ukrainian, Russian, English — each responded differently. Separate creatives for each language, not just translations of the same text.
• Analytics technical issues. UTM tags from leads were being lost due to a plugin — it was impossible to tell which campaign was generating quality applications. The client’s developers were brought in and data transfer was fixed.

A creative here isn’t just an image. It’s the first point of contact with someone looking for work in a foreign country. So the approach to creatives was just as systematic as the approach to channels. At launch, static banners ran with a clear offer: job type, conditions, income level. No fluff — the person needs to understand in a second that this is for them. Video was tested in parallel. And that’s where one of the project’s key insights emerged: couriers and drivers respond to fundamentally different formats. Couriers responded better to native video — simple, unstaged, shot like regular content. Drivers — the opposite: professional video published on a branded page. The same hypothesis, two opposite results. Creatives were tested separately for each language audience — Ukrainian, Russian, English. Not just translations, but separate messaging for each group.

June — 76 leads. Two channels, a test budget, many unanswered questions. Eight months later — over 6,500 leads from three channels.

📌Meta Ads

• 4,639 leads over 8 months
• Best CPL for couriers: 12.92 PLN
• 1,020 leads in November

📌Google Ads

• 1,646 conversions over 6 months
• Highest activation rate among all channels
• Record month: November — 586 conversions at CPL 52 PLN

📌TikTok

• Lowest CPL across all channels — 3.41 EUR for couriers
• Up to 238 leads per month at minimal spend
• Operated as a stable supplementary channel for three months