# Read an experiment's result and say whether to ship

> Read an experiment's result and say whether to ship: 2 companies fully do it, 9 with a limit, 11 carry the tag. Every listing in Product analytics & experimentation with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/product-analytics.experiment-readout/
- Markdown: https://www.anchorterminal.com/niches/product-analytics.experiment-readout/index.md (~1,600 tokens)
- Slim: https://www.anchorterminal.com/niches/product-analytics.experiment-readout/index.min.md (~330 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/product-analytics.experiment-readout/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

Two or three companies' docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 2 | 9 | 11 | 9 | 5 |

## What counts as done

Per-variant metric values, uplift, interval or probability and sample sizes, read through the API or MCP, with the vendor's exposure and test-account filters applied.

Measured, for our own tests: the values are identical to the experiment page, read in at most 3 calls.

How agents ask for it: did the checkout test win; summarise our completed experiments; is this experiment significant; show results by device; which experiments were inconclusive

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **Optimizely Experimentation (Optimizely North America Inc.)**: Get Experiment results takes a baseline, an interval (start included, end excluded), a confidence threshold and one segmentation (device, browser, source or attribute); exp_program_reporting_execute_query returns lift, significance, intervals and visitor counts.
- **Statsig (Amplitude, Inc.)**: experiment_read returns overall and per-metric dimension results with CUPED and confidence options from Statsig's own Pulse, and cumulative_exposures and dimensional_exposures give sample sizes.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Amplitude (Amplitude, Inc.)**: use_amp_experiments analyses variant performance and significance over MCP, but no REST results endpoint was found and the docs don't say the experiment page's exposure filters apply; this is the vendor's tool description.
- **Countly (Countly Ltd)**: ab_experiments_details returns per-variant results and statistical significance, but sample sizes, intervals and exposure filters aren't described, and the ab-testing plugin is Enterprise only.
- **Flagsmith (Bullet Train Ltd)**: get_experiment_results and get_experiment_exposures (beta) return statistical results and exposure counts, but return null until computed, and the docs compute results on a UI Refresh click at most every 5 minutes.
- **GrowthBook (GrowthBook, Inc.)**: GET /experiments/{id}/results and POST snapshot with a dimension (now including combos) return per-variation results, but results still list deleted metrics (open issue 4799) and appear only after a snapshot runs.
- **Harness Feature Management & Experimentation (Harness Inc.)**: fme_experiment_result lists evaluated per-metric results for each comparison treatment, filterable by metric, but only the latest calculation run, and sample sizes and intervals in the output aren't documented.
- **LaunchDarkly (Catamorphic, Co. (dba LaunchDarkly))**: get-experiment-results summarises every metric on an iteration and get-experiment-metric-results gives detailed statistics, but the tool text doesn't promise sample sizes or the page's attribute filters, and the public REST spec has no results endpoint.
- **Matomo (InnoCraft Limited)**: The AbTesting API reads experiment reports, but it is a premium plugin and its result fields (uplift, significance, sample size) couldn't be read from the JavaScript-rendered reference.
- **Mixpanel (Mixpanel, Inc.)**: Get-Experiment returns configuration and results, with interpretation guidance, but the tools are an open beta and the docs don't list the result fields or confirm they match the experiment page.
- **Mostly Good Metrics (Mostly Good Metrics)**: get_experiment and the experiments feature describe enrolment from first exposure and conversions joined on user ID, but the result fields (uplift, interval, sample size) aren't listed in the docs read.
- **PostHog (PostHog Inc. (formerly Hiberly Inc.))**: experiment-results-get returns metric and exposure results with per-metric metadata, so the agent needn't write its own query (acc-17), but dimension breakdowns aren't documented and MCP-created metrics broke the experiment page (acc-07, open).

- not this job: Fullstory (Fullstory, Inc.): No experiments product in the MCP tools or server API.
- not this job: Heap (Content Square, Inc.): No experiments product in the developer reference.
- not this job: OpenPanel (OpenPanel AB): No experiments product in the MCP tools or API.
- not this job: Pendo (Pendo.io, Inc.): listGuideExperiments returns guide A/B experiments' status and configuration only; no tool returns per-variant results, uplift or intervals.
- not this job: RudderStack MCP (RudderStack, Inc.): No experiments.
- not this job: Userpilot (Userpilot, Inc.): The MCP permissions and use cases cover no experiment results.
- not this job: Woopra (Woopra, Inc.): No experiments product.

## What stops agents finishing

- results null until the vendor computes them
- dimension breakdowns only computable from the UI
- agent recomputes results with its own query and gets different numbers
- deleted metrics still listed in results

## Around it

Before: product-analytics.flag-experiment-change. After: product-analytics.flag-experiment-change, product-analytics.stale-flag-cleanup.

## Where this comes from

Weight 9 of 100 in Product analytics & experimentation: 5 of 55 account links; the buyer check names it, but first-hand agent readouts are few and two of the five are vendor-written

As data: https://www.anchorterminal.com/api/v1/use-cases/product-analytics.experiment-readout.json, or the MCP tool get_use_case with id product-analytics.experiment-readout. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
