Weighted percentages need the right base.

Campaign A records nine conversions from ten sessions. Campaign B records nine conversions from ninety sessions. Their conversion rates are 90% and 10%, but the combined rate is 18 divided by 100, or 18%. Averaging 90% and 10% produces 50%, which describes the mean of campaign-level rates instead of the conversion share of all sessions.

If conversions are in B2:B3 and sessions are in C2:C3, the overall rate is =SUM(B2:B3)/SUM(C2:C3). Format the result as a percentage. Check that each conversion count uses the same definition and attribution window and that the session groups do not overlap.

If rates are stored as numeric percentages in D2:D3 and their bases are in C2:C3, use =SUMPRODUCT(D2:D3,C2:C3)/SUM(C2:C3). Store 90% as 0.9, not as the number 90. Ensure both arrays cover the same rows and represent matched rate-base pairs.

Prefer counts over displayed rates when possible. A report showing 33% may represent one of three events, a rounded value of 33.333%, or an independently defined metric. Reconstructing counts from rounded percentages can introduce error. Retain full precision in the calculation and round only the final display.

For conversion, the denominator might be eligible sessions. For average selling price, divide revenue by units. For a defect rate, divide defective items by inspected items. Revenue is not a universal weighting field merely because it is large or financially important.

Sometimes equal group weighting is deliberate. A manager may want the mean branch score with every branch treated equally, regardless of size. Label that statistic as an average across branches and show the spread. Do not present it as the experience of the average customer unless the underlying weighting supports that interpretation.

A group with no eligible sessions has no defined conversion rate. Label it unavailable, retain its zero base, and investigate why the population is empty. A missing base is different: the group may have activity that was not measured or exported. Omitting it silently makes the overall rate appear more complete than it is.

Before calculating, identify valid pairs where both numerator and denominator are known. Report excluded groups and their known activity separately. Do not use a broad error-catching formula to replace every failed division with zero; it can conceal broken references and malformed imports as well as legitimate empty populations.

An overall rate assumes the groups can be combined meaningfully. A person appearing in two channel groups makes a user denominator nonadditive. Daily distinct users also cannot simply be summed into monthly distinct users. If the source provides a deduplicated overall metric, compare against that result instead of reconstructing it from overlapping groups.

Check time zones, date windows, eligibility rules, and currency where relevant. A weighted average is mathematically tidy even when its inputs are incompatible. Neither SUMPRODUCT nor a chart can detect every semantic mismatch for you.

A higher overall conversion rate does not prove that every campaign improved. Traffic can shift toward campaigns with higher existing rates while each individual campaign deteriorates. Compare rates within stable groups and show how their denominator shares changed alongside the overall number.

The PivotTable checking method covers why grand totals and row percentages may differ. For a decision document, show conversions, sessions, the resulting rate, and any exclusions together. To inspect your own prepared export, start a project and review those bases before relying on the summary or buying downloadable outputs.