Metrics Picker (Primary + Guardrails)

What this tool helps you do

Choose one number to decide the test, one number to confirm business impact, and a few safety checks so you don’t accidentally break something.

Why it matters

Most “winning” tests don’t fail because the idea was bad — they fail because the team measured the wrong thing.
This picker helps you match metrics to what you’re changing, and to where users are in the funnel.

Use it when

  • You’re about to launch an A/B test and want to agree on “success” up front

  • You’re not sure whether to measure bounce, click-through, form completion, or sales/leads

  • You want to protect against side effects (errors, speed issues, low-quality leads, refunds)

Skip it (for now) when

  • Your tracking is unreliable or inconsistent in GA4 (we’ll still recommend metrics, but fix tracking first)

  • The change is huge and affects everything (split it into smaller tests so you can learn what actually worked)

How to think about metrics (without overthinking it)

1) Decision metric (Primary)
The one metric you’ll use to decide if the test worked.

2) Business impact (Secondary)
The “so what?” metric — did it actually improve sales, leads, or activation?

3) Safety checks (Guardrails)
Metrics that shouldn’t get worse while you’re chasing the win (errors, speed, churn, lead quality).

If you’re using GA4: we’ll show GA4-friendly definitions and (optionally) the events to track.

Quick shortcuts (pick the closest match)

  • If you’re improving clarity → measure CTA click-through / engagement first

  • If you’re reducing friction → measure completion rate / step-to-step conversion

  • If you’re building trust → measure progression + errors (and confirm impact with sales/leads)

  • If you’re optimising revenue → measure conversion rate or revenue per visitor, and watch quality/returns

How to read the result

  • Decision metric up, safety checks down → don’t ship yet; iterate or limit to a segment

  • Decision metric flat, business impact up → you may be measuring too early in the journey; try a later-step primary metric next time

Common mistakes we see

  • Measuring only sales/leads when the hypothesis is about confusion, bounce, or friction

  • Changing the success metric after you’ve seen the results

  • No safety checks (then a “win” quietly increases errors, churn, or low-quality leads)

Next step

Once you’ve picked metrics, generate a tracking plan and then estimate sample size + duration so you know how long the test needs to run.

Pick your primary metric, secondary metric, and guardrails (GA4-first)




Tip: If the hypothesis is behavioural (clarity/friction/trust), we usually pick a behavioural primary metric and keep revenue/leads secondary.





If revenue/activation lags, we’ll often keep a nearer-term primary metric and use revenue/activation as secondary.









Default is behavioural-first for behavioural hypotheses. Business-first is best when tracking is solid and the change is clearly revenue-driving.