Impact

Capture failure dropped from about 26% to about 6% after launch, as measured in Mixpanel.

In validation testing, all 6 participants completed the capture flow without critical questions, in 60-70% of the time the old flow took.

Context

Trax uses computer vision to read store shelves for CPG brands and retailers. I designed new features for the mobile app that field teams use on store visits: task flows, route planning, store information and navigation. 

The app’s core job is to capture shelf images for the recognition engine. The capture flow relied on too many rules for the average user to remember and follow, which led to errors and inconsistent data.

Research

The problem first surfaced through client complaints. The product team’s view was that the training field reps received was enough, and that capture errors would go away as reps followed it.

I wasn’t convinced. I collected feedback from field reps at a global beverage client, who said it was hard to remember all the details of the capture process they’d been taught in training. To see whether the flow itself was the problem, I tested it with 3 pre-sales colleagues who had never used the app in the field. They couldn’t get through it correctly, which meant the flow wasn’t explaining itself.

To make the case with data, I pushed for a dedicated Mixpanel event to track where users dropped, canceled, or were forced out of the flow. Roughly 30% of users failed to complete it. With that number, the team agreed the problem was the flow, not the training, and the redesign was prioritized.

I then looked at how other products guide people through precise capture: native Android and iOS camera apps, shopping cart flows, PDF scanning tools, and AR game mechanics.

 

Solution 

Instead of relying on training that people forget, I moved the guidance into the camera itself, at the moment it’s needed.

The core issue was the lack of real-time feedback. Users couldn’t tell whether they were capturing images correctly until it was too late.

I designed an onboarding flow and in-camera guidance that gives immediate feedback on direction, alignment, blur, and distance. The guide walks users step by step through both the capture and the stitching sequence.

Inside the camera view, a tile-based status panel shows the quality of each captured image. A good image shows as a confirmed tile. A bad one turns the tile red, so the user knows to retake it before moving on.

One strict requirement is a shooting angle below 15 degrees. I added an angle indicator driven by the device’s gyroscope. When the angle goes past the threshold, the capture button is disabled until the phone is positioned correctly.

 

Validation

I tested the new flow with 6 people: a product manager, a pre-sales rep, an operations rep, two customer reps, and a crowdsourced participant who had never seen the app. All six completed the capture flow without critical questions, in 60-70% of the time the old flow took.

Image capturing Process

Trax Mobile App screens