I had 12 months to prove a scanner could replace the tablet.

The app ran only on expensive tablets, so customers were ready to leave. I led design on the scanner workflow that replaced it, and rebuilt the design system underneath it.


Role

Design lead

Team

1 design, 1 PM, 5 eng

Surfaces

Zebra handheld, Tablet, Web

Before

1h 30m

Manual ordering by counting every item

After

20m

Count targeted items

Future

5m

Fully automated order

Afresh is the software that powers grocers.

No one knows how much of an item will sell on a given day. If the order writer guesses low, the shelf goes empty and the store loses a customer. If they guess high, you get shrink. The real number depends on volatile factors like weather, ads, promotions, local demand, and seasonality. Our AI/ML forecasting system recommends how much to order That recommendation prevents 200M+ lbs of food waste yearly.

Business problem

People problem

Afresh was exclusive to tablet devices which is expensive. You also could not scan on a tablet. The rest of the industry was already on handheld scanners and perspective customers preferred scanners.

Corporate customers wanted centralized AI ordering and not level control since stores were gaming the system in order to produce the recommendations they wanted. So the strategy was to shift labor away from stores to corporate.

An order guide can have tens of 100’s of items. Ordering makes an order writer decide item by item. A scanner is built to capture data fast and it is poor at displaying much of it. So the mental model had to change at the same time remove the bloat that was added over time.

Before

1h 30m

Manual ordering by counting every item

After

20m

Count targeted items

Future

5m

Fully automated order

The strategy

Capture

Scan the items running low or already gone. Count only the targeted items, not the whole department.

Generate

Afresh forecasts demand and turns the captured inventory into a recommended quantity per item.

Review

Corporate reviews the flagged items. The job becomes approving an order instead of writing one.

Notable designs

Since this scanner was a new modality, there hasn't been a mobile design system built with the scanner in mind. Design considerations had to be made for its small form factor, environment and conditions. A new design system had to be created to satisfy these requirements.

A new design system

When I joined, the design system was incomplete and lacked mobile considerations. Designers were creating custom components that quickly proliferated into inconsistent variations. Engineers had also hard-coded values over time.

Scanner workflow

End to end. Category structure, item card, count entry, unit switching, completion states, failure paths.

A new design system

When I joined, the design system was incomplete and lacked mobile considerations. Designers were creating custom components that quickly proliferated into inconsistent variations. The team urgently needed consistency to strengthen brand equity and accelerate both design and engineering velocity.

Strategy

Old tablet workflow

Store managers took inventory on all items + placed orders in same session

Bloated UI with 12+ data fields per item

1.5 hours average time

High cognitive load leading to errors, fatigue

New scanner goal

Inventory on AI-selected set of items only

Ordering handled centrally by corporate (still in scanner for some customers)

Single-task focus leading to speed + accuracy


Next

Customer Care Platform