By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.

where every test makes the next one smarter

Experimentation Growth Engine

A fully structured, AI-powered experimentation programme built on top of your analytics stack. We design and run the system — hypothesis generation, AI-powered prioritisation, experiment setup, result evaluation, and documentation — so your programme compounds over time instead of starting from scratch each sprint.

10x Hypothesis Speed
AI-powered generation and prioritisation vs. manual research alone.
15 Min Per Experiment
Results previously taking 2–3 hours per experiment, compressed into a repeatable automated workflow.
Compounding System
Every sprint builds on the last. The system gets smarter without adding headcount.

Who This Is For

Programme Feels Slow & Manual

You’re running experiments but the programme feels slow, manual, and hard to scale.

Subjective Prioritisation

Hypothesis generation takes weeks and prioritisation feels subjective or inconsistent.

Results Don’t Compound

Results get documented but don’t systematically feed back into the next round of tests.

Want Scale Without Headcount

You want to run more experiments without linearly increasing headcount.

Solid Analytics Foundation

You have a solid analytics foundation (Amplitude, Mixpanel, GA4) and want to activate it for experimentation.

Scope of Work

Three continuous phases running every sprint.

Phase 1

Plan & Setup

  • Hypothesis bank built from analytics data (Amplitude MCP), session replays, UI/UX audits, and customer feedback
  • AI-powered ICE prioritisation calibrated to your business model, traffic, and metric tiers
  • Client scoring incorporated to remove subjectivity bias
  • Experiment design with control and variant clearly defined
Phase 2

Evaluate Results

  • Claude connected to your analytics evaluates results post-experiment
  • Upstream and downstream impact assessed automatically
  • Structured result write-ups populated into results database
  • What previously took 2–3 hours per experiment compressed into a repeatable workflow
Phase 3

Document & Reiterate

  • Results feed back into the hypothesis bank automatically
  • Winners scaled, patterns identified, new hypotheses generated from experiment learnings
  • Each sprint builds on the last

Deliverables

  • Live hypothesis bank (Airtable) with AI scoring and prioritisation
  • Automated result evaluation workflow
  • Documented experiment history: hypothesis, design, result, learning
  • System review and iteration every 4–6 weeks
Book a Free Scoping Call

How the System Works

Five AI-powered workflows running continuously.

W1

The Prioritiser

  • Scores every hypothesis automatically using a custom ICE framework
  • Impact tied to reach and metric tier, confidence built from evidence type, ease from build complexity
  • Populates scores in Airtable. No more gut feel prioritisation.
W2

The Hypothesis Generator

  • Mines completed experiment results to draft the next round of hypotheses
  • Iterates winners, surfaces patterns, scores losers low
  • The compounding engine — the more experiments you run, the smarter it gets
W3

The Result Evaluator

  • Pulls analytics data post-experiment and assesses upstream and downstream impact
  • Writes structured result documentation into the database automatically
  • 2–3 hours per experiment → 15 minutes
W4

Meeting Sync

  • Captures client input from weekly syncs via Fireflies — decisions, subjective scores, context
  • Adds structured input to the hypothesis bank
  • Prevents context loss between sessions
W5

Variant Design (optional)

  • AI-assisted variant generation used selectively alongside manual Figma design

Roadmap

From system setup to a compounding programme running at full velocity.

Weeks 1–2

System Setup & Hypothesis Bank

  • Kickoff: analytics access, experiment history review, tool setup (Airtable, Claude, Amplitude MCP)
  • Hypothesis bank populated from analytics, session replays, UI/UX audit, and client backlog
  • W1 Prioritiser run: all hypotheses scored and ranked
  • Top 3–5 experiments aligned with client for first sprint
Weeks 3–6

First Sprint

  • Variants designed (Figma) and experiments set up in testing tool
  • W4 Meeting Sync running: client input captured weekly
  • Experiments live and running
Weeks 7–8

First Results & Iteration

  • W3 Result Evaluator: results pulled, impact assessed, findings documented
  • W2 Hypothesis Generator: next round of hypotheses generated from results
  • System review and refinement
  • Next sprint planned and prioritised

FAQs

What analytics tools does this work with?

Amplitude is our primary tool — it’s deeply integrated into the hypothesis generation and result evaluation workflows via the Amplitude MCP. We also work with Mixpanel and GA4, though some workflows are more powerful with Amplitude.

Do we need Airtable?

Yes — Airtable is the backbone of the hypothesis bank and results database. We set it up for you at kickoff. If you already use Airtable internally, we can work with your existing workspace.

How is this different from just running A/B tests ourselves?

The difference is compounding. Most internal programmes run tests and document results — but the learnings don’t automatically feed the next sprint. The Growth Engine closes that loop: every result generates new hypotheses, every sprint builds on the last. The system gets smarter without adding headcount.

What do we need to provide?

Admin access to your analytics (Amplitude or equivalent), Airtable access, product access for session replay review, and a dedicated POC who can join weekly hypothesis prioritisation sessions. The weekly sync is what keeps the system compounding.

How quickly will we see results?

The first experiments are typically live by Week 4–5. Compounding value — where the system is generating better hypotheses faster than manual research — usually becomes visible after the second full sprint cycle (Weeks 7–8).

What’s the investment?

€3,000–€5,000/month, 3-month minimum. Scope and exact pricing determined at kickoff based on the number of active experiments, your analytics tool, and level of team involvement.