compare analytics tools

smolanalytics vs the other analytics tools

Honest side-by-side comparisons, including a straight "when to keep them" on every page. The short version: the others are query surfaces, smolanalytics is an investigator. It reads your event log on its own, finds what changed, names who it hit, prices it when the metric carries revenue, replays your history so findings are dated by when they would first have surfaced, and shows the receipt when it heals itself. One snippet, one isolated instance per project, your data exportable in one file, any time. If you build AI apps, it also computes agent observability (tool-call latency, error taxonomy, conversation health) from the same events.

Analytics tools, side by side

The dimensions that actually differ. Verified against each tool's current product, tap a name for the full breakdown.

Tool	What to fix (a verdict)	Ask in your editor	Self-host	Own your data	Entry price
smolanalytics	✓	✓ your own model, free	— isolated instance per project	✓ exports in one file	from $19 · 14-day trial
GA4	—	—	—	— (Google's)	free, sampled
Plausible	—	—	✓ + a database	✓	from $9
PostHog	—	MCP (SQL)	~ Kafka+ClickHouse	✓	free + usage
Mixpanel	—	paid, their AI	—	—	free tier + usage
Amplitude	—	paid, their AI	—	—	free + usage
Umami	—	—	✓ + Postgres	✓	from $20
Matomo	—	—	✓ PHP + MySQL	✓	free · from ~$26
Fathom	—	—	—	✓	from $15
Heap	—	—	—	—	free + sales

“—” = not offered · “~” = partial. smolanalytics is the only one that tells you what to fix and answers in your editor on your own model for free.

“Why not just use Mixpanel or PostHog?”

Fair question. They're deeper. But there are things they structurally can't give you, because it would break their own business, or because they'll never live in your codebase.

It tells you which commit moved the metric

Every other tool shows you the graph dropped. It can't tell you which ship dropped it, it doesn't have your code. We do: record a deploy and ask “did my last deploy move signups?”, and you get the deterministic before/after with the commit named. They're data silos; they'll never be in your repo.

The AI is yours, so it's free

They rent you their AI by the query and gate the good version behind higher tiers. We run no model; you point your own coding agent at it. They literally can't make their AI free without eating their margin.

Answers are computed, never generated

Their AI writes SQL and hallucinates. Ours never generates a query; it calls exact, deterministic reports that return the real number, or nothing, and every answer names the report it came from.

Your data stays yours

Every project gets its own isolated instance, nothing pooled with other tenants, no phone-home, and your data exports in one file, any time. It never trains anyone's model.

same depth where it counts (funnels · retention · paths) · a fraction of the price, flat · and it tells you what to fix instead of making you dig.