HomeWorkAbout
Industry work

A forecast you could plan around.

One of the first public forecasts of when COVID-19 would peak — and whether hospitals would have the room to meet it.

View Project
Role
Data viz · design eng
Year
2020
Org
IHME · UW
How it was made

A daily forecasting model, made readable in seconds.

In 2020 I was at IHME: projected deaths and hospital demand for every U.S. state, then every country, each shown as a range rather than one confident number. The first public version shipped in weeks. We rebuilt it every day after that.

Reach
  • Ran daily in national and global news.
  • Used by millions, and briefed by the White House.
A region's daily-deaths forecast: the observed climb so far, a projected curve continuing past today, and a shaded uncertainty band that widens into the future toward a marked projected peak.
One region: what we’d seen, what we projected, and the uncertainty widening ahead.
Why it stuck

Most dashboards counted what had already happened.

This was one of the first to put a date on the peak: when each place would crest, and whether its hospitals would have room.

Several regional curves on one timeline, each peaking on a different date, with the peaks connected to show the wave arriving later from place to place.
Peaks weeks apart by region
A projected demand curve cresting above a dashed capacity line, with the gap between them — the shortfall — shaded in.
Projected demand vs. capacity

Same idea as my transit work: the forecast people can act on isn’t the most confident one, it’s the one that’s honest about how wrong it might be.

← All workGet in touch →
Michael Fernandes
0
500 PX