AI agent for growth marketers
Campaign Experiment Operations Agent
Experiments that are valid, monitored and judged by their own rules.
What it does
A/B tests often go wrong because a variant was set up badly or someone called a winner too early. Before launch, this agent checks the variants and tracking against the experiment charter. During the run it monitors traffic allocation and whether events are arriving completely. At each checkpoint it first checks that the run is still valid. Only then does it analyze results, using only the rules written into the charter in advance. If tracking breaks, it opens a repair task and holds back any winner claim. The growth lead decides on launch, stopping and any change to success criteria. Edge case: a tracking fault makes variant B look like it got 40% of traffic instead of 50%.
How it works
Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.
Read the steps as a list
- Experiment setup or checkpoint
- Validate variants and instrumentation
- Check allocation and event completeness
- Is the run valid?If not: create a repair task; withhold conclusions. Back to step 2.
- Analyze with predeclared rules
- Growth lead decidesThe agent waits here for your OK.
- Experiment log and decision-ready results
How it decides
It runs validity checks before outcome analysis, every time.
- Validity checks always run before any outcome analysis.
- Use only the analysis rules predeclared in the charter.
- No winner claim while any instrumentation issue is open.
Make it yours
Every agent is a starting point. You choose these settings for your own situation.
- Allowed allocation drift before a run is flagged (default 2 points)
- Checkpoint schedule (default day 7 and day 14)
- Minimum event completeness required (default 98%)
- Who receives repair tasks and results
What keeps you in control
It always asks you first
- Launch
- Budget changes
- Changing success criteria
Hard limits
- No mid-test criteria changes.
It stops when
- Done: decision checkpoint reached.
Set it up
We guide you through the set-up, step by step
Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.
- One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
- The agent then walks you through connecting your own data, one source at a time
- A downloadable copy with the flow chart, the rules and the full guide
