A field platform, made trustworthy.

01 — The problem

The platform lets brands send field managers to branches to run structured surveys, capture geotagged photos and raise incidents. Multiple brands run on the same codebase under their own colours and domain. Two things were quietly going wrong in production.

• Survey submissions on some branches were returning a server error. The error was being swallowed, so the app reported nothing useful, and the field teams simply assumed the network was bad.
• Reference images that field managers need in order to conduct a survey were not loading. Managers were completing audits without the material they were supposed to compare against.

02 — What we did

• The submission failure was a data-integrity bug: an incident was being linked to its responsible department head twice, and the database's uniqueness constraint rejected the second link inside a catch block that hid the error. We made the link idempotent and surfaced the error path so it can never fail silently again.
• The missing images were an authorisation bug. A media policy we had tightened for security allowed a manager to view a reference image only if its owning brand matched theirs, but head-office content had no brand at all, so every manager on every brand was refused. The fix grants access only when the image belongs to a survey actually assigned to the manager's branch, which is narrower than the original public files had ever been. We did not widen access to make the symptom go away.
• Upload and serving defects found along the way, including server upload limits smaller than a phone photo and files written with the wrong owner, were fixed at the server level on both environments.
• The redesign. With the platform stable, we rebuilt the mobile app's interface against a new design system: brand-derived colour tokens so white-labelling still works, bundled fonts so nothing is fetched at runtime, and a pixel-level pass against the design files for every screen. The web admin's modules, roles and workflows were extended in parallel across two client shipments.

03 — How we proved it

• Every web change was verified live on staging and production with browser automation, logged in as a real brand admin, before it was reported as done.
• Every mobile change was built into a release APK, installed on a device or emulator and exercised against real data, including a department-head login that loaded live complaints and observations.
• A security audit of the media paths accompanied the authorisation fix, with the old public copies of every file removed once the authenticated path was confirmed working.

04 — The outcome

Both silent defects are closed, with root causes written down. Access to reference material is tighter than before, not looser. The redesigned app and the extended web admin rolled out across brands with no regressions reported, and the client now has a verified checklist and test-scenario document for each shipment.

05 — What this means for you

Silent failures are the expensive kind, because your users adapt to them instead of reporting them. If a platform you rely on has bugs that everyone has learned to work around, we can find the root causes and fix them properly, then give the product the interface it deserves. See our Web & Mobile services.