From paper binders to 15,000 farms

Farmers had the app and still used paper. We made it simple enough to use between cows.

Role
Product Designer. I owned the redesign end to end, mentored by 2 senior researchers.
Team
2 senior UX researchers (mentors), product, engineering
Timeline
6 months · 2024
Platform
Agri-tech, B2C mobile
  • ~15Kfarms using the app daily, up from almost none
  • 50+countries, tested on farms in New Zealand and Italy
  • ~10×task events logged, 0.03M (2022) → 0.29M (2024)

Why it mattered

Afimilk's sensors track every cow on the farm, but the data only helps if farmers log what they do: inseminations, treatments, new tags. The field app was made for exactly that, and almost nobody used it. Farmers kept paper binders in the barn and typed things in later, if at all.

The solution

I cut the app down to the day's work. Home is today's tasks, each task is one short flow, and the wand farmers already carry is the fastest way in.

Counters, cards and lists all fighting for attention. Now each task gets one card with today's progress.

Before
Daily tasks, the original appCow card, the original app
After
Daily tasks, the redesignCow card, the redesign
  • Today's work first

    Home is the list of today's tasks with progress for each group. The dashboard of counters is gone.

  • One flow per task

    Find the cow, do the task, move on to the next. It's the same flow from a task or from the cow card.

  • Built for the wand

    Type an ID or tap a tag with the wand. Bulk jobs like tagging stay open for the next cow.

  • Nothing that only looks tappable

    If it looks like a button, it is one. Cards that only looked tappable and tooltips that wouldn't close are gone.

What we learned

Going in, I was sure we'd need tutorials. Then I watched farmers use the app and it clicked. They were fine with tech. What wore them down was every extra tap. My favorite part of this project ended up being deleting things.

The problem

The app could do a lot. Farmers opened it, found a cow, and left without logging anything. The one task that makes the whole system work was the one nobody finished.

Before: the one flow that mattered, and where people dropped off

  1. Open the appLands on a dashboard of counters
  2. Find the cowSearch, or scroll a list

    Most left here

  3. Cow cardGraphs and tiles first
  4. Log the taskRarely reached

After

  1. Open the appLands on today's tasks
  2. Pick a taskProgress per group
  3. Pick the cowOr tap her tag with the wand
  4. Log it, next cowOne tap, list stays open

Most farms never logged anything

Logged in bursts

Big batches entered seconds apart. Farmers wrote things down in the barn and typed them in later.

Tags assigned in bulk

574 farms kept assigning tags long after setup, many at a time.

Two ways in, used equally

So both doors had to lead to the same short flow.

What the app's usage data showed before the redesign.

The evidence

The numbers showed us what was happening but not why, so we went to the farms. We visited, ran about 15 contextual interviews, and joined remote sessions where farmers shared their screen and used the app while we watched.

  • The original app during a remote session: looks tappable, isn't

    Looks tappable, isn'tEvery card has a shadow and an icon. None of them open anything.

  • The original app during a remote session: hints at a scroll that isn't there

    Hints at a scroll that isn't thereThe next cow peeks in from below, so people tried to scroll a list that didn't move.

  • The original app during a remote session: a tooltip that won't let go

    A tooltip that won't let goTap the graph once and the tooltip stays on top until you take an action.

Captured in remote sessions with farmers using the live app. Notes translated from Hebrew.

Too much thinking between cows

These farmers run complex herds and equipment every day, so skill wasn’t the issue. The app kept making them think. Which card opens? Where does the list end? How do I close this tooltip? In UX terms the cognitive load was way too high for a job you do standing next to a cow. Paper never asked them anything.

One flow, built around the barn

Tag assignment is a good example. The data showed farms tag lots of cows in a row, long after setup, so I built the flow for a line of cows and a wand in hand.

  1. Tag assignment, step 1: type or scan

    Type or scanOne field. Type the cow's ID or tap her tag with the wand.

  2. Tag assignment, step 2: confirm the cow

    Confirm the cowHer ID and current tag, then the new tag. One button.

  3. Tag assignment, step 3: next cow

    Next cowA short confirmation, and the drawer stays open for the next one.

We also looked at swipe and multi-select patterns from other apps. Long swipes didn’t fit how our users work, so we dropped them early.

The result

The redesign shipped and held up in field tests in New Zealand and Italy, with only minor bugs. And logging finally moved from paper into the app:

Task events logged in the app, per year. Each one is a task a farmer did in the app instead of on paper. From a check we ran with Afimilk's analysts.

Here’s the flow farmers use now, on the final screens.

The new flow · tap a step to see it

Today's tasks
Pick a task
Open a group
Confirm
Or many at once
1 / 5
Final designs. Demo data.

Pinch to zoom