Skip to content

Automation · Operations and reporting

Review an A/B experiment's result in Amplitude on WhatsApp

Ask on WhatsApp how an experiment is doing: an AI agent reads each variant, its significance and the funnel in Amplitude.

How do I know whether a variant of my experiment has won without opening Amplitude?

You write the experiment's name to your Met, for example "annual price with discount". The Met finds it in Amplitude, brings its state and variants, queries the analysis with each variant's performance and its statistical significance, and checks the purchase funnel in the same period to see whether the winning variant also moves the final conversion. It tells you whether there is a winner or more traffic is needed, and leaves a comment in Amplitude with the summary if that action is on. It does not change or stop the experiment.

What starts it

When someone asks for it

It starts when someone on the team asks the Met about an experiment. Each review runs on request; this procedure does not leave a scheduled alert.

A manual task: someone on the team runs it with a button, or asks the Met for it in chat.

Who runs it

Product analytics Met with Amplitude

This is not a named, seeded Met. It is an internal Met with Amplitude connected: its work ends when it replies with what the experiment analysis says, or when it says there is not enough data to decide yet.

Run by the data analyst Met

Channels: WhatsApp

What you need connected

5 steps · 5 tools

The procedure, step by step

Each step shows the tool that runs. They belong to the real connector: if one of them stopped existing, this page would not build.

  1. 1

    Find the experiment

    The Met searches for the experiment by name in Amplitude. If several similar ones show up, it asks you which one instead of picking the first.

  2. 2

    Bring its state and variants

    The Met brings the experiment with its state, each variant's configuration and the decision, if there is one. That tells it since when it runs and which metric it is measured against.

  3. 3

    Read each variant's analysis

    The Met queries the experiment analysis: how each variant does on the primary metric and whether the difference is statistically significant according to Amplitude.

  4. 4

    Check the funnel in the same period

    So it does not decide only on the experiment metric, the Met runs a purchase funnel query for the same period and tells you whether the final conversion moved.

  5. 5

    Reply and leave the summary in Amplitude

    The Met replies on WhatsApp whether there is a winner or more traffic is needed. If the comments action is on, it leaves the same summary as a comment in Amplitude for the rest of the team.

Before you start

What you need ready

The Met solves none of these for you. If one is missing, the automation stops there.

The limits

What this automation does NOT solve

Better said here than discovered halfway through the rollout.

FAQ

Questions about this automation

Yes, one after another. Naming them well in Amplitude helps the search find the right one on the first try.

The analysis usually comes out without significance. The Met tells you and suggests asking again when there is more traffic.

It shows the Amplitude account that authorized the connection in Met, not the person who wrote to the Met on WhatsApp.

Keep going

Other automations built on the same tools

Want this automation running on your own data?

We will show it to you with your accounts connected, not with a canned demo.