How I doubled conversion and take rate on microloans
I reworked the microloans page on the Alfa-Bank website. Application conversion grew 2×, and loan Take Rate by roughly 2×.
Context
Alfa-Money is one of the bank's key products for acquiring microloan customers. A customer fills out an online application and gets a microloan on their card within 5 minutes.
How the section looked before
There were no critical problems on the page. It had run quietly for years — bringing in applications, never causing a flood of complaints. Metrics held steady, and the bounce rate sat in the green.
So where was the problem?
I'd long sensed the page was dated and that I was leaving conversion on the table. I went through session recordings, ran user interviews, and found:
The page itself had piled up a huge amount of legacy code, and every new UX improvement had to clear a wall of technical constraints that often got badly in the way.
Defining the task
Quarterly goal: lift CR from 10% and TR from 9%. Beyond the product task — pilot Alfa-Bank's new UI on this page.
There was a quarterly goal — grow CR and TR. I was free to experiment and spend whatever resources I needed on it. As the designer, I also got a special brief: pilot Alfa-Bank's new UI on this surface and, more broadly, figure out what the microloans section should become. I experimented a lot and tested different hypotheses. Here are the ones that mattered most — the ones that actually moved conversion.
In interviews I found that, on an emotional level, users are wary of the very words “microloan,” “loan,” and any mention of an MFO.
How I solved it
- 1.
Removed the blocker words everywhere and replaced some with softer wording. For example: “loan” → “money,” “approval” → “a decision in 1 minute,” and so on.
- 2.
Ran an A/B test with two variants to lower user anxiety.
This variant produced no statistically significant change in conversion — likely due to banner blindness.
I placed variant 2 right under the form — that gave a lift of about +2 pp. I kept it.
If I add a CTA button at the top of the page, more users will reach the application
A/B test: a variant with the button and one without
Variant A won: application CR higher by
7 pp
If I move the benefits and the form up, I'll raise application conversion ⠀⠀⠀⠀⠀⠀⠀
A/B test: trim the benefits copy and move it up with the form
Variant A won here too. Application CR grew by
3 pp
The most common reasons people take a microloan:
Putting the insight to work
I assumed that showing users quick offers matched to their task would raise application CR — because they wouldn't have to “translate” their situation into loan parameters. Instead of an A/B test, I went straight to the target segment to gauge interest.
Lands on the page
Sees a block of ready-made offers for real-life situations
Recognizes their situation and taps “Get”
Smooth scroll to the application form
Fields are pre-filled for the scenario: purpose, amount and term
Adjusts the fields if needed
Enters contact details
Submits the application
The bet on quick offers didn't pay off on take rate: the demand is there — 27% CTR — but it hasn't converted into approved loans yet. My hypothesis why: the set of life situations was too narrow, and the pre-fill was based on the segment median rather than each user's own profile. I'll validate this with a full A/B test — in a 2026 follow-up.
Post-analysis and takeaways
I went through several dozen hypotheses in Q3-Q4 2025 — only the ones that genuinely moved the metrics made it into this case. My main takeaway: targeted UX fixes give a steady but small gain. The real jumps happened when I stopped speaking the bank's language and started speaking the user's. Not everything landed, and that's fine. Some hypotheses showed no statistically significant effect, some showed demand without moving the metric. But every test, even a failed one, pointed to where to dig next.
Doubling CR was the cumulative effect of a series of tests across Q3–Q4. The page shows the three most telling ones; together, dozens of small improvements and a shift to the user's language took it from 10% to 20%.
What I'd still like to do
Push the quick offers further: expand the set of life situations and validate them with a full A/B test, not just within the target segment.
Make pre-fill smarter — pull the amount and term from each user's own profile and history rather than the segment median.