Agent leaderboards / All sectors / Product analytics
Product analytics: which products coding agents choose
PostHog took about 56% of 598 runs across 10 codebases.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked five coding agents to add product analytics to 10 codebases, 598 times in all. We varied the wording and asked as different people. After the leader, the agents most often wrote the analytics themselves, in about 14% of runs.
One theme put a different product on top
When the ask was about fitting an existing data stack, Metabase led with 10 wins. PostHog won 1 of those 59 runs.
Wording changed the answer more often than not
A case here is one codebase with one agent, asked several times in different words. In 36 of 50 cases the runs did not all land on the same product.
The persona moved the pick
Junior developers picked PostHog in 222 of 240 runs. For vibe coders the top pick was Vercel Analytics with 27 wins. PostHog had 21.
Named often, chosen almost never
Google Analytics came up in 117 runs and won none of them. Fathom was named in 82 runs and also never won. Plausible was mentioned 201 times and won three runs.
- The simulated user approved every plan in the end, but sent the agent back at least once in 105 runs.
- In nine runs it refused to approve until the agent named a specific product.
- All 18 wins for Segment came in one codebase, a B2B logistics monorepo.
- Looker Studio won 12 runs, all of them in the Go fleet tracking backend.
The ranking598 runs
| Product | Wins | Share | ||
|---|---|---|---|---|
| 1 | PostHogposthog.com | 332 | 56% | |
| 2 | Built in-houseoutcome | 82 | 14% | |
| 3 | Metabasemetabase.com | 37 | 6% | |
| 4 | Vercel Analyticsvercel.com | 27 | 5% | |
| 5 | Mixpanelmixpanel.com | 19 | 3% | |
| 6 | Amplitudeamplitude.com | 19 | 3% | |
| 7 | Segmentsegment.com | 18 | 3% | |
| 8 | Looker Studiolookerstudio.google.com | 12 | 2% | |
| 9 | Umamiumami.is | 8 | 1% | |
| 10 | Datadogdatadoghq.com | 5 | 1% | |
| 11 | Plausibleplausible.io | 3 | 1% | |
| 12 | Mitzumitzu.io | 3 | 1% | |
| 13 | Datadog Product Analyticsdatadoghq.com | 3 | 1% | |
| 14 | Snowplowsnowplow.io | 2 | 0% | |
| 15 | Lookercloud.google.com | 2 | 0% | |
| 16 | Matomomatomo.org | 2 | 0% | |
| 17 | Grafanagrafana.com | 2 | 0% | |
| 18 | Ahoygithub.com | 1 | 0% | |
| 19 | Lightdashlightdash.com | 1 | 0% | |
| 20 | Kubitkubit.ai | 1 | 0% | |
| 21 | Metabase + Plausible | 1 | 0% |
By agent, by persona, by wording
By agent
| Codex180 runs | PostHog · 104then Amplitude · 13 |
| Claude Code150 runs | PostHog · 69then Vercel Analytics · 9 |
| Cursor149 runs | PostHog · 89then Vercel Analytics · 10 |
| Muse Code · Muse Spark 1.390 runs | PostHog · 56then Metabase · 11 |
| Grok Build CLI · Grok 4.729 runs | PostHog · 14then Metabase · 10 |
By persona
| Junior developer240 runs | PostHog · 222then Metabase · 3 |
| Senior engineer238 runs | PostHog · 74then Metabase · 27 |
| Vibe coder60 runs | Vercel Analytics · 27then PostHog · 21 |
| Enterprise team60 runs | PostHog · 15then Datadog · 5 |
By what the ask stressed
| The plain ask220 runs | PostHog · 146then Vercel Analytics · 15 |
| Read by someone who is not an engineer139 runs | PostHog · 107then Amplitude · 10 |
| Self-hosting, privacy or residency100 runs | PostHog · 40then Vercel Analytics · 12 |
| Fits an existing data stack59 runs | Metabase · 10then Amplitude · 6 |
| Procurement and compliance40 runs | PostHog · 23then Mixpanel · 3 |
| Volume and cost at scale40 runs | PostHog · 15then Metabase · 6 |
A case is one codebase with one agent, asked several times in different words and as different people. 36 of 50 cases did not hold to a single product.
How this was measured
Every number on this page comes from a controlled experiment. We took 10 small applications, asked 5 coding agents (Codex, Claude Code, Cursor, Muse Code (Muse Spark 1.3), Grok Build CLI (Grok 4.7)) to add product analytics to each of them, in several wordings and as a junior developer and senior engineer and vibe coder and enterprise team, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 598 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 105 runs. Read the methodology and the publications.
If you sell in this sector: what these numbers mean for a vendor.
Open the interactive boardThis page as MarkdownOther sectors
- Agent sandboxes
- Observability
- AI SRE
- Payments
- Deploy
- Auth
- Email providers
- Databases
- File storage
- LLM evals & observability
- Voice Agents
- Serverless functions
- Cloud
- AI gateway
- Bot protection
- Search
- Agent frameworks
- Performance in CI
- Document processing & OCR
- Usage-based billing
- Internationalization
- Security testing
- Maps
- Message queues
- AI search
- Code review
- E-signature
- Security platforms
- In-app chat & calls
- Vector search