A key performance indicator (KPI) is a specific, measurable metric that reflects how well you're achieving a goal. In CRO and growth, KPIs are the numbers your team tracks to determine whether optimization efforts are actually moving the business forward.
KPIs are not the same as metrics. You might track hundreds of metrics, but your KPIs are the 3–5 that directly tie to business outcomes.
Common CRO KPIs
| KPI | What It Measures | Best For |
|---|---|---|
| Conversion rate | % of visitors who complete a goal | Landing pages, signup flows |
| Revenue per visitor (RPV) | Average revenue generated per visit | E-commerce, SaaS pricing pages |
| Customer acquisition cost (CAC) | Cost to acquire one new customer | Paid campaigns, overall efficiency |
| Average order value (AOV) | Average spend per transaction | E-commerce, upsell experiments |
| Bounce rate | % of single-page sessions | Content pages, landing pages |
| Time to convert | Days from first visit to conversion | B2B, long sales cycles |
KPIs vs. Metrics vs. Goals
- Goal — The outcome you want ("Grow trial signups by 30% this quarter")
- KPI — The number that tells you if you're on track (trial signup conversion rate)
- Metric — Any measurable data point, whether or not it's tied to a goal (page load time, scroll depth, button clicks)
Every KPI is a metric, but not every metric is a KPI.
Choosing the Right KPIs for Experiments
When running A/B tests or multivariate experiments, you need to define your KPI before the test starts — not after. This prevents cherry-picking metrics that happen to look good.
Good experiment KPIs are:
- Directly tied to business value — Conversion rate or RPV, not vanity metrics like pageviews
- Sensitive enough to detect changes — A metric that barely fluctuates won't show experiment impact
- Measurable within the test timeframe — Revenue per visitor works for short tests; lifetime value doesn't
Primary vs. Guardrail KPIs
- Primary KPI — The metric you're trying to improve (e.g., signup rate)
- Guardrail KPIs — Metrics you don't want to harm (e.g., ensuring a signup rate increase doesn't reduce paid plan conversions downstream)
A test can "win" on its primary KPI while causing problems elsewhere. Guardrail KPIs catch this before you ship a change that looks good on the surface but hurts the business overall.