New editorial guide, published 17 October 2026. The historical sources linked below are preserved separately and are not represented as current product or service information.
A report says a page converted at eight percent. Without more information, that figure cannot answer whether the page worked well, improved or should be copied elsewhere. Eight percent of what? Which action counted? Were repeat actions included? Was the measurement available for everyone? This guide works through a fictional enquiry website to show how a small publisher can read a percentage without treating it as a complete verdict. All numbers are invented for arithmetic examples. They are not this site’s analytics, a campaign benchmark, a sales result or a recommendation to spend money on advertising.
Name the action before calculating the rate
Choose the event that answers the question you are asking. For the fictional site, opening the enquiry form, submitting it successfully and receiving an eligible enquiry are three different events. A button click may lead to an error, and a successful submission may still concern a service the organisation does not offer. Do not call all three enquiries. Record the event’s trigger and the system that observes it. If the event definition changes, mark the change in the report rather than presenting the later figure as directly comparable with the earlier one.
Distinguish total events from people or sessions with an event. One person submitting twice can produce two submission events. Whether that counts as one or two outcomes depends on the question and the approved measurement method. Use the same definition in the numerator and in the surrounding prose. A headline saying twelve customers is wrong if the system recorded twelve form events and cannot identify unique customers. Precise labels are not pedantry; they prevent a technical event from quietly becoming a stronger business claim during editing or presentation.
Choose a denominator that matches the question
Suppose the fictional report records two hundred measured sessions that reached the enquiry page and sixteen sessions with a successful submission. Dividing sixteen by two hundred gives 0.08, or eight percent. This is a session-based rate for that measured population and period. It is not automatically a person-based rate, a site-wide rate or the proportion of qualified leads. If the question concerns people who opened the form, use an appropriately defined form-opening population instead. Changing the denominator changes the question even when the numerator remains sixteen.
Write the denominator in the chart label or a nearby note. Do not make a reviewer search a separate technical appendix merely to learn what the percentage describes. Also state important exclusions, such as test traffic removed under a documented rule. An exclusion should have a reason established independently of whether it improves the headline number. If the numerator includes actions outside the denominator’s eligible group, the rate may be incoherent. Review event timing and identity rules with the person responsible for measurement before treating an unusual result as a business insight.
Keep observation windows aligned
A monthly page-visit count and a quarterly submission count do not form a meaningful monthly conversion rate. Use aligned windows and an explicit time zone. Consider actions that occur after the initial visit. A person may begin a form in one period and finish in another. The report needs a rule for those cases that matches its purpose and can be applied consistently. Do not change that rule after seeing the results without documenting the change. A tidy percentage cannot repair numerator and denominator records that describe different populations or periods.
Note when the data was collected and when it was exported. Late-arriving events or later corrections can change a report. Save a version so another reviewer can reproduce the arithmetic from the same inputs. Missing observation is not the same as a recorded non-action. If measurement was unavailable for part of the period, state the gap and its known scope. Avoid multiplying a partially observed total by an assumed correction factor unless there is a defensible method and the resulting estimate is clearly labelled as an estimate rather than a measured count.
Separate percentage points from relative change
Now imagine the next comparable period has two hundred measured sessions and twenty successful-submission sessions. The rate is ten percent. It rose by two percentage points, from eight to ten. The relative increase is two divided by eight, or twenty-five percent. Both descriptions can be arithmetically correct, but they say different things. A headline saying the rate rose by two percent would be ambiguous. Show the underlying counts and the before-and-after rates so a reader can understand the scale without reverse-engineering the wording.
Small counts can produce dramatic relative changes. Moving from one event to two doubles the count, but it does not establish a durable pattern or explain its cause. Do not let a large relative percentage crowd out the actual numbers. Use precision that suits the observations, and avoid ranking tiny differences as if they were conclusive. This guide does not calculate statistical uncertainty or prescribe a sample-size threshold. The practical first step is simpler: identify what was observed and avoid implying a level of confidence that the report has not analysed.
Do not confuse a change with its cause
Suppose the enquiry page was rewritten between the two periods. The higher observed rate does not by itself prove the rewrite caused it. The visitor mix, service availability, traffic source, measurement implementation or form behaviour could also have changed. Record these plausible differences as questions to investigate, not as an invented explanation. A before-and-after comparison can be a useful observation while remaining insufficient for a causal claim. The wording should preserve that distinction: the measured rate increased after the update is narrower than the update increased the rate.
Check the path for obvious measurement or usability faults before celebrating. A duplicate event can inflate the numerator. A broken page-view event can shrink the denominator. A form may report success before the server has accepted the request. These are implementation possibilities, not claims about any particular analytics product. Ask for evidence at the relevant boundary and inspect the real task where access permits. Do not repair a suspicious report by adjusting counts until the result looks plausible. Fix the observation method and document which periods can still be compared.
Publish a metric note someone else can reproduce
A compact metric note can include the question, event definition, unit, denominator, period, exclusions, source and calculation. Add known gaps and the date the note was prepared. W3C’s Data Quality Vocabulary is a Working Group Note offering ways to describe quality information such as measurements and their context; it is not a certificate that a dataset is universally fit for every purpose. The useful idea here is contextual description. A percentage becomes more interpretable when the reader can see how it was produced and what question it was intended to answer.
Finish by stating the decision the evidence can support. The fictional report might justify checking whether the revised page and form function as intended, or continuing a clearly defined observation period. It does not automatically justify a revenue forecast or a spending increase. If the measurement cannot answer the business question, identify the missing information instead of forcing a recommendation from the available percentage. A modest, reproducible result is more useful than an impressive number with an invisible denominator. Keep the counts, definitions and limits attached whenever the figure moves into a slide, summary or article headline.
References & further reading
From the recovered archive
- Best Digital Marketing Strategies To Grow Your Business · captured 22 October 2021
- Are You Sabotaging Your Marketing Success? · captured 26 February 2021