[ Case study ]
The product looked its age and customers had adapted to its quirks — power users had muscle memory for exactly where things were. A previous 'modernization' attempt had been reverted within a month after a support storm.
CLIENT a 10-year-old vertical SaaS product — FOCUS Map the muscle memory first
Representative examplesEvery case study in this library is an illustrative composite of the kind of engagement we deliver — written to show our method and standards, not to name clients.
Independent pharmacies run their daily operations on a ten-year-old vertical SaaS product — dispensing, inventory, compliance records. The interface was designed when the product shipped and has aged with its customers: power users operate it at speed through keyboard paths and exact positions they've held for years. The company ships continuously with a small engineering team, and its support queue is the most sensitive instrument it owns — customers call support the moment something moves.
The product looked its age and customers had adapted to its quirks — power users had muscle memory for exactly where things were. A previous 'modernization' attempt had been reverted within a month after a support storm.
We proposed refreshing the visual layer without touching the behavioral layer: heatmaps of the five most-used workflows would define untouchable zones — same positions, same labels, same keyboard paths — while typography, spacing, color, and states modernized at the token level. Rollout would be per surface with support-ticket monitoring as the regression detector, and a planned revert path for any section that provokes the queue. Because the customer base reads change as risk, the refresh had to feel like maintenance — the product looks newer without behaving differently, and nothing moves that a daily user's hands know.
Just as important is what we ruled out, and why:
Heatmaps of the five most-used workflows defined the untouchable zones — same positions, same labels, same keyboard paths.
Typography, spacing, color, and states modernized at the token level; screen structures stayed — the product looks newer without behaving differently.
Changes rolled out section by section with support-ticket monitoring as the regression detector — the revert scenario was planned, not feared.
Delivered by the experience pod — 2 designers over 10 weeks, with working increments reviewed with the client every week.
Obstacle
One heatmap contradicted the others: the dispensing workflow's usage pattern turned out to include a stale toolbar from a deprecated feature that a third of users still ran out of habit.
Handled: We shadowed five pharmacy calls to separate habit from need, removed the stale path from the untouchable set, and added a deprecation notice ticket to engineering's backlog.
Obstacle
Internal dogfooding found three contrast failures in the new tokens a week after the first customer surface shipped — our own team spotted what customers would have found less politely.
Handled: We paused the rollout one surface deep, fixed the token values, published the correction internally for two weeks, and only then resumed; no customer ticket ever mentioned contrast.
The headline: reversion requests in the two quarters after full rollout (the prior attempt: 1 within a month) — 0, read from Support ticket audit. A second check: points of nps improvement among daily active users, same survey at +38.
Nothing moved that daily users' hands knew, and the queue stayed quiet through the full rollout — the planned revert path was never used, which the team treats as the real verdict. Demos stopped opening with an apology for the interface, and new prospects see a product that looks maintained rather than abandoned. Power users noticed the refresh and, mostly, approved it in the exact terms it was designed in: nothing to relearn. The company regained its nerve about touching its own interface, which had quietly frozen feature work too.
The result was read from Support ticket audit against the pre-engagement baseline over the stated window, with a guardrail check on points of nps improvement among daily active users, same survey. Where platform-reported numbers and business outcomes differ, this record says which layer it is quoting.
What we would do differently
We would have published the token changes internally for a month before customers — our own team found three contrast issues that external users would have found less politely.
[ Related service ]
[ Related builds ]
[ Next step ]
Next case study
