A conversion rate can rise while the business performs worse. If low-intent traffic stops arriving, the denominator shrinks and conversion rate improves—even if relevant sales stay flat or fall. Changing the conversion definition can make a chart show progress that exists only in the report.
Conversion rate remains useful when accompanied by its definition, audience and the outcome it describes. Below is a five-step check before drawing conclusions from an increase or decrease.
1. Write exactly what you measure
“Conversion” can mean a button click, checkout start, completed order or enquiry ultimately considered qualified. Choose an event with business meaning, define when it is recorded and preserve history when it changes. Do not compare periods using different definitions as though they measured the same thing.
2. Check the denominator
Explain whether you divide by visits, sessions, users, clicks or eligible opportunities. Visits and users are not interchangeable. Multiple devices or return visits affect interpretation. Also record exclusions: internal traffic, test purchases, bots or duplicate events.
3. Look at who converts
Break results into meaningful groups: traffic source, campaign, device, new or existing relationship, product or region. The overall rate may hide improvements in low-value enquiries masking deterioration in leads that close. However, do not create dozens of small segments afterwards and select only favourable ones; specify in advance which groups determine the decision.
4. Connect the online event to the final outcome
A form submission does not always equal a sale. Agree with sales what qualifies a lead, which stages are tracked and how losses or cancellations are recorded. For ecommerce, examine net sales, returns and margin where available. The easy event is not automatically the useful one.
5. Separate association from cause
A rate increasing after a page change does not independently prove the change caused it. Seasonality, traffic origin, stock or the offer may have changed. Where possible, maintain a comparison group and predetermine the primary measure, time window and required sample. For low volume, show uncertainty rather than decimal-point certainty.
A small scorecard is enough
For each change, note the main conversion rate, conversion count, quality or value, cost per outcome, key groups and any recording changes. Add what changed, when, why and which alternative explanation was examined. This scorecard keeps discussion close to a decision instead of orbiting an isolated percentage.
The checking framework and scorecard were originally developed by DigitalNow.