fivebar

fivebar vs Pirsch

Pirsch Standard and fivebar cost about the same. fivebar also times your pages, shows the bots it left out, counts errors without code, and has an MCP server of its own.

Pirsch and fivebar have a lot in common: no cookies, servers in Germany at Hetzner, a visitor counted once in a day, an AI channel for visits from ChatGPT and the other assistants, and an API for events from your own server. Where they differ is what the dashboard tells you beyond visits, and which plan you need for it.

How fast your pages load

Pirsch doesn’t time your pages: none of the statistics its docs explain is a measure of speed. fivebar’s tag times every page load in your visitors’ own browsers, and the Speed page rates each measure by Google’s thresholds, the Core Web Vitals among them, with the pages that load slowest listed first.

fivebar’s Speed page for djangodocs.dev: six tiles, TTFB, FCP, LCP, INP, CLS and PLT, each rated, then LCP for each day against Google’s thresholds
djangodocs.dev, whose stats are public, from 12 to 18 October 2026. One slow Sunday put its LCP past poor.

The bots it left out

Pirsch filters bots, and on purpose doesn’t say what it dropped, so that nobody can learn how to get past its filter. fivebar counts everything it leaves out on a Bots page: how many pageviews, which pages, which networks, and why each was taken for a bot, by rules its docs set out. When a visit of your own goes missing, that’s where to look.

fivebar’s Bots page for djangodocs.dev: 94 pageviews taken for bots, 30% of them all, by day, and why: a hosting network, an automated browser or a crawler
Nearly a third of djangodocs.dev’s pageviews that week were bots’, most of them from servers.

Pages that fail

To count a 404 in Pirsch, you call pirschNotFound() on each 404 page, since its script can’t tell a missing page from any other, and each one counts toward your bill. Chrome, Edge and Firefox tell fivebar’s script the status of every page they load, so the Errors page fills itself: which addresses fail, how often, and where their visitors came from. Safari needs one attribute on your error pages’ template. Error pages are never billed.

Asking an app

fivebar has its own MCP server: add https://fiveb.ar/api/mcp to Claude, ChatGPT, Cursor or any app that takes one, sign in and choose Allow. The app reads your stats with fivebar’s rules about what they mean, and if you tick a box, adds goals and changes settings when you ask.

Pirsch has none of its own. Its docs list one made by the community, which Pirsch doesn’t support: you run it on your own computer, with a client ID and secret from your Pirsch account, and it reads your stats without changing anything.

Prices

Pirsch has two plans with prices, and Enterprise by quote. Standard costs about what fivebar does up to 10M pageviews and less above it, for up to 50 sites. Plus, for any number of sites, adds funnels, revenue, tags, A/B tests by tag and teams, at two to three times the price. fivebar has one plan, for up to 1,024 sites, with custom properties and A/B tests in it.

Pageviews a monthfivebarPirsch StandardPirsch Plus
10k$5$6$12
100k$13$12$27
1M$50$54$159
10M$165$159$479
50M$650$499$1,499

A month, in US dollars, from fivebar’s prices and Pirsch’s. Both sell a year for the price of ten months. Pirsch bills the 404s it counts; fivebar never bills an error page, nor a bot.

What else differs

What Pirsch has that fivebar doesn’t

Switching

Add fivebar’s tag beside Pirsch’s, and run both for a few weeks, so you have both counts for the same days before you take Pirsch’s out. Where your pages call pirsch() for an event, call tally() with the same name, and its meta as the data; a button with data-pirsch-event takes data-tally, and its data-pirsch-meta- attributes become data-tally- ones. Events your server sends go to fivebar’s API with a token.

Checked on 19 October 2026, on Pirsch’s own pages, each linked where it’s used.