You are choosing an analytics tool for a small product team. The shortlist is Plausible, PostHog, and Fathom. Every homepage says the same three things: privacy-first, lightweight, easy to set up. Reading three pricing pages and three feature grids side by side takes an afternoon, and you still end up with a browser full of tabs and no clear answer.
Here is a faster path that does not need you to open any of those tabs. You ask the AI assistant you already use to run a Cobalt comparison study, and it comes back with a sourced, dated matrix and a lean.
The exact thing you type
Once Cobalt is connected to your assistant (one time: paste https://cobaltcapture.com/mcp as a custom connector in Claude, or run claude mcp add --scope user --transport http cobalt https://cobaltcapture.com/mcp in Claude Code), you ask in plain words:
Run a comparison study of Plausible, PostHog, and Fathom for a 5-person product team that wants privacy-friendly web analytics without self-hosting.
No prompt engineering, no code. The /workflows command lists every study and the exact phrasing if you want it. You name the products and the one decision you are making, then you walk away. The study renders the public pages itself and adds web search for the facts a product page leaves out, so there is nothing for you to click through.
The decision framing matters more than the product names. "For a 5-person team that wants privacy-friendly analytics without self-hosting" tells the study which row of each pricing table is yours and which features to weigh. Swap in "for a solo founder on a free tier" and the same three products produce a different lean.
What comes back
The review leads with the verdict, then backs it with a matrix. Rolled up, it reads something like this:
| Dimension | Plausible | PostHog | Fathom |
|---|---|---|---|
| Cost at team-of-5 volume | Mid, usage-tiered by pageviews | Generous free tier, pays for product events beyond it | Flat, simplest to predict |
| Scope | Web analytics only | Analytics plus session replay, flags, experiments | Web analytics only |
| Setup for a non-self-hoster | Script tag, minutes | Script plus product setup, more steps | Script tag, minutes |
| Fine print worth reading | Pageview caps per plan | Event-based pricing can surprise you | Clear limits, fewer surprises |
Every cell carries its source and the date it was checked, because a pricing page you read six months ago is not evidence anymore. The matrix is a screenshot-backed snapshot of what those pages said on the day it ran, not a claim pulled from memory. That dating is the part a feature grid you build by hand almost always skips, and it is the first thing that goes stale.
The lean at the top is a plain sentence: for a small team that wants only web analytics and predictable billing, Fathom and Plausible both fit and PostHog is more than you need unless you also want replay and flags. It does not pretend one is the single best tool on the market. It tells you which one answers the question you asked.
Why the output beats a tab marathon
Reading three pricing pages yourself, you anchor on whichever you opened first and you forget half of page one by the time you reach page three. The study holds all three to the same dimensions and the same team size, so you are comparing like for like instead of comparing your memory of one page against a fresh read of another.
The review ships as a shareable link, so you can drop it in a Slack thread and let the two other people who care weigh in. People with the link can comment on any single row, and you can mark each comment resolved as the decision firms up. That turns a solo research task into something the team actually signs off on, which is the hard part of any tool choice.
Comparison is one of the public-page studies, which means it runs start to finish from Claude chat with no browser needed, because Cobalt renders the pages. You can kick off your own from the product comparison use case, which lists the phrasing and what each study covers. If you want to see the format before you run one, Cobalt publishes its own studies of well-known products at the teardowns index.
When to reach for a different study instead
Comparison is the right tool when the decision is "which of these two or three." If you have already picked analytics tool and want to know whether a cold visitor to your own marketing page understands the pitch, that is a positioning audit instead. If the question is whether your own pricing page tells a buyer what they would pay, run a pricing teardown from Claude chat on your own site. And if you want to feel the first-run experience of a tool you are leaning toward, a usability pass that signs up and tries real tasks catches friction a comparison matrix never will.
You can try all of this free. Cobalt asks for an email when a study is ready so it can send you the link, with no credit card. Reviews stay editable for 7 days, then go read-only for you while the shared link keeps working for everyone you sent it to; Pro is $5 a month if you want to keep reopening and re-running them. Re-run the same comparison in three months and the new review leads with what changed, so the next tool decision starts from a dated baseline instead of three fresh tabs.