All field notes

Field note #3: I built a free tool for PostHog bills in a day

On 18 November my X feed filled up with people angry about their PostHog bills. By the next afternoon I had a free tool out for them. It's called Frumo.

PostHog is an analytics tool that a lot of startups use to see what people do in their apps. It's generous: it says 97% of companies pay nothing. But the bill grows with how much you track, and it's easy to track far more than you need without noticing.

Why this, and why now

One founder's post asking someone to "reverse engineer and open source posthog" passed 300,000 views, and replies put his bill at about $10,000. PostHog's co-founder replied in public: the cause was autocapture, a setting that records every click by default.

That day I asked Hugo to look into it. Hugo is one of the AI agents I run in Grok Bot, an AI assistant app, and his only job is to explore this one idea. He found 19 public complaints from the past two years. Most named the same causes: autocapture, identified users, session replays and feature flag checks. Most of the bills were under $1,000 a month.

He also found a catch. PostHog is building the same audit into its own billing page. It isn't live yet, but once it is, a paid tool doing the same job makes little sense.

So for now I'm running a cheap test, and a business can wait.

What I built

Frumo is a command-line tool. You give it a read-only PostHog key and it lists the events, replays and flags driving your bill, what you'd save by changing them, and the fix to paste in. It runs on your own machine, and your data only goes to PostHog. It's free and open source under the MIT licence, which means anyone can read the code, use it and change it. You can try it on sample data with one line: npx frumo-cli demo.

It's built for PostHog. It isn't made by PostHog and isn't affiliated with them.

There's also a page at frumo.marianoventures.com with a waitlist for a future monitor, something that would warn you before a bill spikes. If nobody wants that, it won't get built.

The test

I wrote the pass and fail rules down before launching, so I can't move them later. Within 7 days of the first post:

  • Keep going if Frumo gets at least 30 downloads, at least 10 visits from links inside the tool or its report, and at least 10 waitlist sign-ups, with 3 or more of those coming from the tool itself.
  • Stop if there are fewer than 10 visits from the tool, fewer than 5 sign-ups, or PostHog ships its own audit to everyone.
  • Anything in between, I decide whether to give it one more week.

Frumo doesn't track anyone. I count downloads from npm, where the package is published, and visits from the links printed in the report.

What's next

Frumo goes out on X on 19 November. In a week I'll write up the numbers here, whichever way they go.

Fluq is still Project 1. It's waiting on Apple to approve my developer account, and Frumo is what I built in the meantime.

If you pay for PostHog, run npx frumo-cli demo or read more at frumo.marianoventures.com, and tell me what it gets wrong.

Follow along on X: @marianocarrica.

Share this note

More on Frumo

All field notes on Frumo