I Was Wrong

This page logs various times I was wrong or otherwise made mistakes I wanted to own up to. It’s not comprehensive.

Credit to my former employer, GiveWell, which inspired me with its mistakes page.

Past Mistakes

Post About T-Mobile’s 100GB Hotspot Plan

In March of 2021, I published a blog post criticizing T-Mobile for ceasing to offer a 100GB hotspot plan just a few months after the company drew a ton of attention to the plan’s launch. My post had serious issues.

I didn’t adequately distinguish between a prepaid version of the plan that was no longer available and a postpaid version of the plan that remained available. I also, likely incorrectly, suggested subscribers who already signed up for the plan would not be grandfathered.

Misconfigured JavaScript

For about 24 hours in November 2021, some misconfigured JavaScript removed text intended to flag that content about a cell phone carrier was a sponsored placement.

Missing Entry On Transparency Page

For about a year from early 2022 to early 2023, US Mobile did not have an entry on my transparency page even though I had a financial relationship. While US Mobile was inadvertently missing from the transparency page, the financial relationship remained disclosed on other parts of Coverage Critic.

In 2022, I removed a list of cell phone carriers that I had once had indirect financial relationships via a 3rd party middleman. One of these companies was US Mobile. While I ended my relationship with the third party, I had previously established a second, direct arrangement with US Mobile. In March 2023, I realized the error add added US Mobile back to the transparency page.

Ongoing Issues

In this section, I acknowledge ongoing issues with Coverage Critic’s content or product. These aren’t really mistakes, but this is as good a place as any to acknowledge the issues.

Coverage Model Bias Towards Verizon

The model I use to estimate coverage quality in cities throughout the US is significantly biased toward Verizon (and against T-Mobile). While this has likely been the case since the model’s inception, the issue probably worsened over time.

I’m working an upgraded model that relies on newer and more finely grained data. I expect the upgraded model will largely resolve the problem.

Overemphasis On Numerical Coverage Scores

On pages where I share details about coverage quality in specific cities, I put a ton of emphasis on Coverage Scores that rate network quality on a 0 to 10 scale. It’s not straightforward how these numbers should be interpreted, and I hope to move to an approach that places less emphasis on these scores.