Calculated metric

Define inputs and compatible units

Name each input and explain its source, meaning, and exclusions. Confirm that numerator and denominator describe compatible populations. Dividing purchases from all channels by sessions from one channel produces a number, but not a meaningful conversion rate for that channel.

Align time periods and identity rules. A customer, user, session, order, and event are different units. Revenue from customers acquired last year may not belong over this month's new-customer count unless the intended measure explicitly requires that relationship.

Check currency and scale. Dollars and cents, daily and monthly values, or raw fractions and displayed percentages can be mixed accidentally. Document any conversion and rounding. Handle missing inputs separately from genuine zeros: unavailable cost data does not mean acquisition was free. These checks prevent plausible-looking results from hiding inconsistent source data or an unsuitable denominator.

Choose the formula and exceptional behavior

Write the formula in ordinary language as well as mathematical notation. For a rate, explain what qualifies for the numerator and who had the opportunity to qualify in the denominator. For an average, identify the unit being averaged and whether weighting is needed.

Decide what happens when the denominator is zero. Returning an undefined or unavailable state with context is often more accurate than displaying zero or infinity as if it were an ordinary business result. Keep negative values when they are meaningful, such as a net amount after adjustments, rather than automatically removing inconvenient observations.

Define rounding for display without discarding precision prematurely. A small rate rounded to zero can still represent real activity. Keep the underlying calculation auditable and distinguish a display choice from the stored value. Exceptional cases are part of the metric definition, not merely a technical detail for whoever builds the chart.

Aggregate components before calculating ratios

Ratios generally need appropriate weighting when combined. Averaging several percentages equally can misrepresent the overall result when their denominators differ. In many common cases, sum the eligible numerators and denominators first, then calculate the combined ratio.

For example, a fictional campaign produces 10 conversions from 100 visits, while another produces 10 from 1,000. Their rates are 10% and 1%. The simple average is 5.5%, but the combined rate is 20 divided by 1,100, or about 1.82%.

The 5.5% figure answers an equal-campaign-average question, not the overall visitor conversion question. Neither calculation should be labeled ambiguously. The same issue affects cost ratios, average order values, and many dashboard summaries. Specify the intended weighting and test totals, subtotals, and filtered views so the metric behaves consistently when people change the report.

Worked example: calculating approved lead cost

Imagine a fictional team spending $3,600 in media and receiving 120 form submissions. Thirty submissions are duplicates or fail the predefined qualification criteria, leaving 90 approved leads. Media cost per submission is $30, while media cost per approved lead is $40.

Both measures can be useful, but their labels must preserve the distinction. Calling the $30 figure “qualified lead cost” would misrepresent the denominator. Calling the $40 figure fully loaded acquisition cost would also be wrong if production, agency work, and customer conversion are outside its scope.

The team records the qualification rule and waits for the relevant review period before comparing cohorts. If recent submissions are still pending, it reports that limitation rather than treating all of them as approved or rejected. The example shows how a simple division depends on operational definitions and timing, not just arithmetic.

Validate and maintain the metric

Recalculate a small sample independently and compare it with the report. Test zero denominators, missing values, duplicate records, adjustments, and aggregation across groups. Inspect whether joins multiply rows or filters apply to only one component.

Keep a metric record with its purpose, owner, formula, source fields, population, time basis, and known limitations. Version meaningful changes so an apparent performance shift is not confused with a revised definition. If history can be recalculated consistently, explain that choice.

Review whether the metric still supports the intended decision. A formula can remain technically correct after the business process changes and become less useful. Avoid adding complexity simply to create a proprietary-looking score. A clear, trusted measure with visible limitations is usually more actionable than an opaque index whose movement nobody can explain or connect to a practical response.

Related terms

• Custom metric: a measure configured for a particular need. It may use a calculation but still requires a clear definition.
• Conversion rate optimization: improving a defined conversion outcome. Its analysis depends on consistent rate calculations.
• Customer acquisition cost: a cost measure requiring explicit scope and customer definitions. It should not be confused with narrower media ratios.
• Cohort analysis: comparing groups with shared starting conditions. It supports compatible timing and populations.
• Event tracking: recording actions used as metric inputs. Reliable formulas depend on accurate events.
• Campaign taxonomy: consistent campaign classification. It helps ensure calculated results are grouped and compared correctly.