Turning a leftover bank tool into a focused rate service

Mortgage rate work moved from Sber to Domclick, but the bank's tool came along unchanged. It carried around twenty features this team never used. A wrong rate means the company loses money or loses the customer, so every change was checked by hand. I rebuilt the tool around the one job managers actually do.

Role: Sole product designer, guided by my design lead

Faster core tasks: 50%+
Matrix errors: -76%
Interviews and gemba sessions: 10 + 3

The rebuilt sub-product page: basic parameters, participants, special conditions, discounts and divisions as labelled sections — Visuals rebuilt as a coded prototype for this case study. Structure and flows are as shipped.

Problem

The real job was the day the central bank moves

The core job is simple. Find the product, find the parameter, change the rate correctly. The old tool made each step slow. A 14-column table repeated the product in every row and made you work out the status from dates. The rate matrix was 90 separate fields, edited one at a time. Every mistake had to be caught by the head of the department in a manual final check.

I designed next to the people who used it. Ten interviews showed the team relied on one core function out of the whole tool. The gemba sessions showed when that function hurts most. When the central bank changes its key rate, managers have to update rates across many products by hand, fast, with no tool tracking the change for them. Nobody mentioned this in interviews. It became the center of the redesign.

Key decisions

Cut what nobody uses

Removing around twenty features was the riskiest call, because a wrong cut leaves people without something they need. Usage evidence from ten people made it safe, and nobody pushed back.

The legacy bank tool: a dense product catalog table under a row of search filters — Where it started: the bank's tool, with around twenty features this team never used.

Make the table readable at a glance

Products became groups instead of a value repeated in every row. Min and max ranges collapsed into one cell. Status became a badge instead of something you calculate from dates. Search accepts several codes at once, because managers often update a batch.

The rebuilt products table: sub-products grouped under their product, amount and rate ranges in one cell, status as a badge — 14 columns and a two-level header became six readable columns grouped by product.

Close-up of the grouped table: a product group with its count, ranges in single cells and Active badges

Edit the matrix as a whole, not cell by cell

This is the answer to the key rate day. Managers select a cell, a row, a segment or any area, then apply one operation to all of it. Set a value, add or subtract points, multiply, divide or change by a percent. One action replaces dozens of separate edits.

The legacy rate matrix: rows of separate input fields by client group, range and loan-to-value band — Before: the rate matrix as 90 separate fields, edited one at a time.

The rebuilt rate matrix: one grid by client segment, range and loan-to-value band, with tools to edit selected cells — After: one grid, and every edit starts with a selection.

Close-up of the matrix with a whole column selected, a single cell picked and Select all, showing 14 selected cells — Select a cell, a column or the whole matrix, then apply one operation to all of it.

Let large changes wait safely

A rate update can touch many products. Draft saving lets a manager pause a large change without losing it and come back to finish it.

The legacy sub-product page: every parameter on one crowded screen in three columns — Before: every parameter of a sub-product on one crowded screen.

The rebuilt sub-product page: section titles with a short purpose on the left, their fields on the right — After: each section says what it's for, with its fields beside it.

Close-up of the sub-product page header, tabs and the basic parameters section

The rate matrix with edited cells highlighted, a 14 cells edited, not saved note and a marker on Save changes — Edited cells stay highlighted, counted and flagged on Save until the change is saved.

Outcome

Timed tests before and after the redesign, on finding a parameter and changing a rate, showed both tasks getting more than 50% faster. Matrix errors dropped by 76%, based on error logs and the final check by the head of the department. The tool now belongs to Domclick, not to leftover bank infrastructure.

Try the prototype

What's next

The final review stayed manual. The next step is building it into the tool. Show what changed since the last approved version, flag rates far outside their usual range, and let the reviewer approve or send back in one place.

Back