How We Actually Count Rally Attendance — And Why the AP's Number Is Always Wrong

The AP files a rally count within an hour of the event. It is almost always wrong. Here is how we count.

The three-step method

1. Satellite pre-image — We buy the 3-metre Planet Labs image of the venue from twelve hours before the event. This is the baseline: how many square metres of asphalt, grass, field.
2. Drone at peak — Between 15 and 45 minutes after the speech begins, we fly a DJI Mavic 3 at 120 metres AGL over the venue. Grid photography, 80% overlap, orthomosaic composited within the hour.
3. Density model — We apply a 2.2 people per square metre model for the front third (concert-density), 1.4 for the middle third (crowded), and 0.8 for the back third (standing but sparse).

Why the AP is always wrong

The AP takes the venue's official capacity number, applies the fire marshal's rated occupancy, and reports that. This is fine for indoor events. It is wildly inaccurate for outdoor events, where the venue capacity is set by the number of Porta-Potties, not the acreage.

The counting example

A rally in Youngstown, Ohio on 2026-08-18: AP reported 4,200. Our drone method: 41,300 in the field, another 8,200 in the overflow parking lot watching on jumbotrons. AP filed a correction six days later, on page A14.