I design tools for people who work with data, and I write about how they get made.

The first build of Tidepool had a screen I was proud of for about a week. It showed tide readings for the station a researcher had chosen, drawn as a smooth curve with high and low water marked. In the office, on fast wifi, it looked finished.

Then we took it out on the institute’s survey boat, forty minutes beyond Vinga lighthouse, and the curve was gone. In its place was a grey box that said “No data.”

No data is not one thing

When we read the logs, “No data” covered at least four different situations. The tablet had never synced this station. It had synced, but the station had stopped reporting. The readings existed but were older than the six hours the app treated as fresh. Or the station was healthy and there were simply no readings yet for a day that had just begun.

To the app, all four were the same thing: an empty list.

{
  "station":  "Göteborg Torshamnen",
  "date":     "2026-03-14",
  "readings": [],
  "lastSync": "2026-03-13T22:40:00Z"
}

To a biologist holding the tablet in a wet glove, they were completely different. One meant “sync again when you’re back in range.” Another meant “the sensor is broken, log a fault.” Another meant “wait an hour.”

An empty state is the one screen where the user knows less than the system. Its job is to close that gap.

Writing the empty states first

We rebuilt the screen so each case had its own message and its own next step. “Last synced yesterday at 22:40. Newer readings will appear when you’re back in coverage.” “Torshamnen hasn’t reported since Tuesday. Switch to Fiskebäckskil, the nearest station, or log a fault.” Plain sentences, each with a timestamp and an action.

The surprise was what this did to the rest of the app. Once we had written down what people needed to know when nothing was there, we could see what they needed when everything was. The sync time moved from a settings page to the top of every chart. Station health got a small indicator in the station list. The full screen improved because the empty one made us name what mattered.

On every project now, I write the empty states before I draw the full ones. I list every reason a view could be empty and give each reason a sentence and an action. It takes an afternoon, usually in a spreadsheet, and it has never been wasted.