Skip to content

Methodology

Experiment Engine Methodology

How Growomo structures hypotheses, variants, measurement plans, evidence review, and approved next actions.

On this page

Experiment Engine

Experiment lifecycle

A test remains connected to its original question, measures, guardrails, and reviewed outcome.

  1. Falsifiable hypothesis

    Record the question, expected direction, owner, and decision.

  2. Variants and measures

    Define variants, primary metric, guardrails, and duration assumptions.

  3. Evidence monitoring

    Sync available provider metrics and preserve measurement context.

  4. Human review

    Check quality, seasonality, simultaneous changes, and business impact.

  5. Reviewed outcome

    Record the selected next step and the evidence behind it.

Reviewed outcomes

  • ContinueEvidence supports continuing or iterating.
  • StopEvidence does not support the hypothesis.
  • InvestigateData or context requires more review.
  • InconclusiveThe evidence does not support a conclusion.

Reviewed learning returns to future evidence and decision context.

Experiment inputs

A useful experiment starts with a decision, a falsifiable hypothesis, a primary metric, guardrails, eligible variants, a minimum evidence plan, and a clear owner. Growomo stores this context so results are not interpreted without the original question.

  • Hypothesis and expected direction
  • Source type and variant identifiers
  • Primary metric and guardrail metrics
  • Planned duration and traffic assumptions
  • Owner, status, and decision notes

Monitoring and evidence

Growomo can sync available provider metrics and calculate evidence summaries. A test that lacks traffic, duration, comparable variants, or clean tracking should remain inconclusive. Reaching a planned end date does not manufacture a winner.

Human review

The marketer checks sample quality, seasonality, simultaneous campaign changes, attribution, and business impact. The product can recommend continue, stop, or investigate, but the final interpretation and any provider action remains reviewable.

Worked demo-data example

A fictional landing-page test compares two offers. The demo shows variants, trial counts, a primary conversion metric, and an inconclusive state because evidence is insufficient. The example teaches the workflow and is not a customer outcome.

Questions and boundaries

Frequently asked questions

Does Growomo always declare a winner?

No. A test can remain inconclusive when the evidence, duration, tracking, or sample quality is insufficient.

Can Growomo pause or scale variants automatically?

No blanket automatic-action claim is made. Supported provider actions require an explicit approved user action.