To identify mistaken citations, L.A. Material analyzed LADOT’s parking citation database from March 2021 to the present. We filtered for street sweeping citations, then compared the location where each ticket was issued against the city’s map of street sweeping routes.

But there were a few problems with the city’s data.

First issue: The digital map of sweeping routes has a margin of error. In several instances, using Google Street View imagery, we found that street sweeping signs extended several hundred feet from where the map indicated they did.

To correct for this, if tickets didn’t align exactly with a posted route, our Python script searched for the closest one within 500 feet. To avoid false positives, if at least one route in that radius indicated that the citation was justified, our analysis registered the ticket as legit.

Second issue: Each citation in the city’s database contains two types of location data — a street address and coordinates — but neither is universally accurate.

One set of coordinates was inexplicably in Nebraska. Some addresses simply didn’t exist in the City of L.A.

It proved impossible to get a straight answer from LADOT about how the location data is collected. (Do officers manually enter addresses that a computer system then converts to coordinates? Do their handheld devices automatically enter GPS coordinates?)

In most cases, it looked like the listed street address was an approximation while the coordinates provided more granular data. (In 99% of the cases where the city listed the exact same address on multiple tickets, it recorded multiple sets of different, but nearby, coordinates.)

Because of this, our analysis was based largely on the original coordinates. In rare cases where those coordinates looked suspicious — specifically, when three or more tickets were issued on the same day at the exact same coordinate, but with different addresses — we used ArcGIS to geocode the addresses and relied on those for the route comparison.

Our analysis has probably identified a small number of tickets as bad when they actually weren’t, due to inaccuracies in LADOT’s dataset. It also has almost certainly missed a number of tickets that should be listed as mistaken.

But while there’s a small amount of ambiguity on the margins, the analysis has found a critical mass of tickets that were undoubtedly issued in error, and established a clear pattern. The only question that remains is how many of those tickets were actually canceled or reimbursed.