Conversion rate: choose the right calculation and avoid false progress

8 min read

Calculating a conversion rate requires a numerator, a denominator and a common perimeter. To measure the share of sessions that result in a valid request, divide sessions with at least one valid request by eligible sessions, then multiply by one hundred. Do not replace this numerator with the total number of requests if you claim to measure sessions. Google distinguishes in particular between a key event rate per session and a rate per user.^[54] Therefore define the unit before the calculation, then check the quality of the result. A higher percentage may hide a change in population or a drop in useful demands.

To remember: a rate must be able to be told

A complete definition says “among this population, over this period, what proportion accomplished this validated action? ". If the team cannot explain this sentence, the indicator is too ambiguous. Keep the formula and starting values ​​next to the percentage, along with any important exclusions.

Define conversion rate calculation according to decision

First choose the expected result: order, valid request, confirmed reservation or intermediate step. A progress event can be useful for diagnosis, but should not be renamed as a business outcome to improve the curve.

Then determine the unit of the denominator. A session represents a measurement journey; a user represents an identity unit according to the tool; a contact represents a job profile. These populations are not interchangeable. Contact qualification rate does not directly measure site session conversion.

The API reference GA4 defines sessionKeyEventRate as the percentage of sessions in which a key event was triggered and userKeyEventRate as the percentage of users who triggered a key event.^[54] Check the metric and event included in your report, rather than inferring their meaning from the word “conversion” alone.

Construct a numerator that corresponds to the denominator

For a converted session rate, count sessions satisfying the condition, not each repetition of the action. A single session can contain several requests; it does not become multiple sessions. If you instead measure the number of requests per session, name this ratio differently.

For a qualified contact rate, define the eligible contacts and the qualification criterion. HubSpot, for example, distinguishes a marketing qualification from a sales qualification in its life cycle stages.^[45] These supplier categories must be adapted and validated by your team, not taken as a general truth.

Review exclusions: tests, obviously invalid requests, duplicates, and off-offer populations. Document them before comparing. An exclusion applied only to the numerator of one period or the denominator of another can produce artificial progress. Each line must have a stable definition and a person responsible.

Check the action before counting its success

Clicking on “send” does not demonstrate that the form has been accepted. A confirmation page displayed multiple times may produce a repeated event. Verify the end-to-end journey and the outcome in the appropriate business system.

Distinguish between technical validation and commercial validation. A correctly transmitted request may be out of scope, not usable or simply too recent to be qualified. Do not correct the report by deleting these cases after the fact without a rule; set understandable steps and reasons.

The 4SYSTEM marketing dashboard guide allows you to link indicators to decisions. Tie your rate to volume and quality: the business needs to know if it's receiving more of the useful results, not just if a fraction increases.

Fictitious example: two calculations on distinct populations

The following data are educational hypotheses, not site or customer observation. In a first set, the site has 600 observed eligible sessions and 12 sessions having at least one valid request. The verified calculation is: 12 / 600 × 100 = 2 %. It describes the share of these sessions that satisfied the condition.

In a second set, distinct from the first, a CRM contains 30 contacts to examine and 12 contacts qualified according to the chosen rule. The verified calculation is: 12 / 30 × 100 = 40 %. This rate describes the qualification of this population of contacts. It is not a second reading of the same sessions and should not be presented as the site's conversion rate.

The fact that both numerators are equal to 12 does not create any link between the populations. Connecting pathways would require appropriate definitions and mapping, as well as a validated treatment framework. The example only demonstrates why an exact formula can answer the wrong question if its unit is not made explicit.

Indicator sheet to copy

Add your period, your manager and the last check date. The form must accompany the figure in the dashboard. An accessible explanation prevents the same indicator from changing meaning when it passes from marketing to management.

Compare periods without mixing populations

Identify the same units, rules and exclusions. A new channel can bring a wider audience; a change in form may modify eligibility; a campaign can shift the weight of mobile visits. The overall variation may therefore come from the composition of traffic rather than from a common improvement.

Look at relevant, predicted segments, not just looking for the one that gives the best upside. Keep the volume of each group visible and avoid sweeping conclusions about populations that are too small or unstable. An observed difference is not automatically causal evidence for your change.

Note the collection breaks. If an event has been corrected, the increase may represent a more comprehensive measurement rather than new behavior. If the CRM criteria change, the qualification rate must be annotated. A stable definition is not an administrative detail: it conditions any comparison.

Take commercial maturation into account

A recent application may not have received a definitive qualification. Comparing an old, closed period to a recent period still in progress introduces a gap. Organize reading by cohort of requests or by processing status, with a common maturity rule.

HubSpot presents its stages as a way to understand the flow of leads between marketing and sales.^[45] Use this idea to distinguish between “to be considered”, “qualified” and “not considered”. A lack of decision should not automatically become a poor quality request.

Also keep the response capacity visible. If the files are not processed, the qualification rate may drop without a change in targeting. The calculation should guide an investigation of the entire chain, not be used to mechanically attribute responsibility to the site or to a campaign.

Common errors and validation rule

The main errors are mixing requests/sessions, changing exclusions, using an unvalidated event and comparing commercial populations with different maturity. Also check that exports do not group together several sources without controlled deduplication.

Before publishing a rate, ask another person to reconstruct the formula from the data and the sheet. It must obtain the same result with the same perimeter. If it has to interpret ambiguous notes or choose the exclusions itself, the indicator is not yet sufficiently defined.

Frequently asked questions

What conversion rate should you aim for?

There is no relevant universal target without knowing the result, the unit, the offer and the audience. Start with a stable calculation and a verified measurement. Then define an objective linked to your commercial constraints, with the expected volume and quality. Do not use an uncomparable benchmark as a verdict on your site.

Can a rate rise while activity falls?

Yes, a percentage also depends on its denominator and the composition of the population. Review the starting volumes and useful results before concluding. A relative increase does not automatically mean that the company is receiving more qualified requests or that a page edit has caused progress.

Should we count contacts or sessions?

Choose according to the question. Sessions are used to read a measurement path; contacts are used to examine a relationship and its qualification. GA4 distinguishes its rates per session and user.^[54] The CRM uses its own definitions. Present the indicators separately, then link them only when the correspondences are documented.

Next step: publish definition before rate

Write a record for the indicator used in your next decision and check its events. To coordinate site, measurement and commercial process, 4SYSTEM services can help frame definitions. A useful calculation remains explainable and comparable, even when the result is not flattering.

Sources

• Use contact and company lifecycle stages (accessed 2026-10-08)
• Google Analytics Data API : dimensions et métriques (accessed 2026-10-08)