Wish · WishDash

The problem beneath the brief

The brief was about merchants. The fix ended up in a driver's pocket. Fewer merchants were using Wish's first-mile shipping; the ask was to improve usage. The screens merchants saw were never the lever.

The real users weren't the merchants — they were the drivers. First-mile is the stretch from a merchant's door to the origin warehouse, and it ran on paper sign-offs, typed handover codes, and pickups where the two sides missed each other. Merchants felt the symptoms; drivers lived the process. After the first interviews, the project changed shape: design for the people the process actually runs through.

Validation, not a hunch

The iceberg was a hypothesis — so I went and checked it. Drivers don't visit usability labs, so the research went to them: 252 surveys, 15 interviews, 3 field studies at the loading door.

The answer came back consistent. Neither external nor internal drivers cared what was inside the packages; they cared which bags, how many, where to, confirmed how. The hunch held — the work lived with the drivers — and it narrowed to the two flows a driver actually lives in.

Designed, then iterated

Two constraints shaped the build: a WeChat Mini-Program, not a native app; and data spread across separate systems that had to feel like one process.

The pickup itself, rebuilt for the loading door: scan the merchant's code, enter the bag count, snap a photo, done. A paper sign-off and a typed handover code became four taps — the legitimate replacement internal drivers had been improvising around. Testing set the details: the bigger text external drivers asked for, and only the confirmation steps that actually get skipped when they're slow.

Between pickups, a driver works a queue. Testing pointed the same way every time, so the list leads with the actions — hand-off, warehouse sign-in, transfer — and each card carries only what a driver checks before moving: whose, where, how much. Tap an order and the detail leads with the map: the relevant bags, and a live route to the stop — what a driver on the road actually needs.

The transparency merchants had been promised showed up on the operations side — a live overview of every task and driver, error cases surfaced early, and a direct line to call a driver when something stalled.

None of it landed in one pass. Each flow went back to the drivers in round after round of testing — the bigger text, which confirmations were worth the tap, the order of a card — and every detail above is what those rounds settled on, not what I guessed at the start.

Reusable parts, drawn once

Underneath it all, one shared visual layer: common patterns and a reusable illustration library, drawn once and recombined across the mini-app, the portal, and every empty state. Partly to move faster — but as much to keep one coherent brand across surfaces built at different times, so first-mile always looked like one product rather than three teams' screens.

It launched in late 2021, and the paper finally lost. Internal drivers took to it quickly — the sign-off they'd been skipping now had a faster, legitimate replacement — and a package's first mile became visible to everyone who needed to see it.