Two team percentages reach a dashboard. Someone adds them and divides by two. The answer looks clean. It describes a different question from the one above the chart.
An overall request completion rate needs completed requests and eligible requests. A team score can give each team equal influence. Those are separate measures, even when both appear as percentages.
Start with the counts underneath the display. Decide what one observation represents. Then combine only records that belong together.
This guide uses fictional service teams and reproducible arithmetic. The worksheet is a proposed reporting method. It assumes the metric definition is already agreed. It does not estimate future performance or infer results for an unobserved population.
Give the Denominator a Name
A denominator says what the rate is out of. For an on-time request rate, that might be eligible requests due during a specified reporting window.
The numerator then counts the included requests that meet the agreed on-time rule. A business metric dictionary gives those boundaries a durable home.
Before combining teams, confirm that both use the same rule. A team counting requests due this month cannot merge directly with a team counting requests opened this month. The column headings can match while the underlying populations differ.
Keep the unit visible. Completed requests divided by eligible requests produces a request rate. Completed tasks divided by employees produces a different measure. Neither column becomes compatible because a spreadsheet formats it with a percent symbol.
For the synthetic examples below, each eligible request appears once. Each request belongs to exactly one team. The reporting period and on-time definition match across both teams. These are explicit assumptions, not facts about an operating business.
Rebuild the Overall Rate From Counts
Team A completes two of ten eligible requests on time. Team B completes forty-five of ninety on time.
| Fictional group | On-time requests | Eligible requests | On-time rate |
|---|---|---|---|
| Team A | 2 | 10 | 20% |
| Team B | 45 | 90 | 50% |
| Combined requests | 47 | 100 | 47% |
Team A's rate is 2 / 10 × 100 = 20%. Team B's rate is 45 / 90 × 100 = 50%.
The simple mean of the two displayed percentages is (20 + 50) / 2 = 35%. That calculation gives each team half the influence.
The overall request rate is (2 + 45) / (10 + 90) × 100 = 47%. That calculation gives each eligible request equal influence.
Check it through the failures as well. Team A has eight requests outside the on-time numerator. Team B has forty-five. Together, fifty-three of one hundred miss the on-time condition. The remaining forty-seven satisfy it.
The twelve-percentage-point gap between thirty-five and forty-seven comes from the weighting rule. It does not reveal a data-entry error in either team's rate.
Write the question next to the result: “Across the one hundred eligible requests in the fictional reporting window, forty-seven finish on time.” That sentence gives the total a population.
Weight by the Quantity the Rate Describes
The equivalent weighted calculation is ((20% × 10) + (50% × 90)) / 100 = 47%. Ten and ninety are the relevant denominators.
They also imply normalized weights of 10 / 100 = 0.10 and 90 / 100 = 0.90. Applied to rates expressed as decimals, 0.20 × 0.10 + 0.50 × 0.90 = 0.47.
OpenStax explains frequency-weighted means by multiplying values by their frequencies and dividing their sum by the total frequency. The example here applies that arithmetic to known request counts.
Do not choose employee counts because they are available. Four employees on one team and nine on another do not establish how many requests each rate represents. Weighting by employees would answer a separately defined question, if the inputs support it.
Do not use the successful-request count as the weight. The rate already contains that outcome in its numerator. For an overall request rate, the weight is the count of eligible requests.
Check that every weight describes the same unit and boundary. A shared total only has meaning when its components belong in that total.
Keep an Equal-Weighted Team Score When That Is the Intended Measure
An organization may deliberately choose to give every team equal influence in a score. Under that rule, the fictional score is thirty-five percent.
Label it “equal-weighted mean of two team on-time rates.” Keep the team count and calculation beside it. Do not rename it the percentage of requests completed on time.
The choice belongs in the metric definition before the result is reviewed. Selecting whichever weight produces the preferred headline changes the question after seeing the answer.
An equal-weighted score can also conceal volume differences. Show the two team rows beside the summary so the reader can see ten requests and ninety requests. Let the reader inspect both the score and the workload distribution.
If the business needs both measures, report both with separate labels. “Team score: 35%. Overall request rate: 47%.” Neither measure replaces the other. Each has its own denominator and intended use.
Combine Duration Averages Through Total Time
Now consider handling duration for a different synthetic set of completed requests. Team C handles four requests with an exact mean of twelve minutes. Team D handles six with an exact mean of twenty-two minutes.
The underlying fictional durations are ten, twelve, twelve, and fourteen minutes for Team C. Team D has eighteen, twenty, twenty-two, twenty-two, twenty-four, and twenty-six minutes.
| Fictional group | Completed requests | Exact mean duration | Reconstructed duration total |
|---|---|---|---|
| Team C | 4 | 12 minutes per request | 48 minutes |
| Team D | 6 | 22 minutes per request | 132 minutes |
| Combined requests | 10 | 18 minutes per request | 180 minutes |
Multiply each exact mean by its observation count. Then add the totals: 12 × 4 + 22 × 6 = 180 minutes. Divide by ten completed requests: 180 / 10 = 18 minutes per request.
The direct row sums confirm the result. Team C totals forty-eight minutes. Team D totals one hundred thirty-two. The combined ten durations sum to one hundred eighty minutes.
A simple mean of twelve and twenty-two produces seventeen minutes per request. It gives the four-request group the same weight as the six-request group.
Keep minutes and hours separate until converted. If a source reports hours per request, document the conversion before combining it with minutes per request. Preserve the original unit in the review record.
Use the count behind the duration average. Requests excluded from duration measurement do not belong in that weight. An eligible-request count from another metric is insufficient.
Withhold a Combined Rate When the Inputs Cannot Support It
Suppose a summary supplies twenty percent and fifty percent but omits both denominators. The overall request rate remains unknown. The two percentages alone cannot establish forty-seven percent.
Ask for the underlying numerator, denominator, population boundary, and reporting window. Until those arrive, present the separate rates with the missing evidence labeled.
Rounded rates create another boundary. Multiplying a displayed thirty-three percent by three requests produces 0.99 completed requests. It does not reconstruct the original count. Prefer retained counts to reverse-engineering a rounded display.
The same limit applies to duration means. A displayed average of twelve minutes could have been rounded from a more precise value. Its product with a count is then an estimated total. Label any accepted approximation and its rounding basis. Otherwise, withhold the precise combined mean.
Overlapping cohorts also block direct addition. If five requests appear in both team exports, adding the denominators counts those requests twice. Resolve identity and cohort membership before calculating the union's rate.
Incompatible periods require a decision about the question. Nonoverlapping weeks can form a longer reporting period when definitions match. Overlapping weeks cannot be added as distinct observations. A monthly snapshot and a year-to-date total do not form a new population through addition.
Use dashboard reconciliation to resolve those boundaries. More arithmetic cannot repair incompatible definitions.
Copy the Aggregation Worksheet
Use one record for each combined metric. The fictional entries below document the forty-seven-percent request rate.
| Field | Fictional entry or review instruction |
|---|---|
| Reporting question | What share of included requests finishes on time? |
| Numerator | 2 plus 45 on-time requests, totaling 47 |
| Denominator | 10 plus 90 eligible requests, totaling 100 |
| Period | One shared fictional reporting month, with exact boundaries recorded before use |
| Population boundary | Requests due in that month, each assigned to one team and counted once |
| Source precision | Exact integer counts; no counts reconstructed from rounded rates |
| Aggregation rule | Sum numerators, divide by summed denominators, multiply by 100 |
| Weight basis | Eligible requests, giving teams weights of 0.10 and 0.90 |
| Rounding | Retain full calculation precision; display 47% for this exact result |
| Independent check | 100 eligible minus 53 outside the numerator equals 47 on time |
| Evidence | Retained synthetic rows or identified source exports with matching cutoffs |
| Unresolved conditions | None within the fictional assumptions; verify them before using real data |
Record the actual reporting boundaries before replacing the fictional values. Include whether the ending timestamp belongs inside or outside the window. Keep the selected time zone with the period definition.
For a duration measure, replace the numerator with total duration and the denominator with measured requests. Record the time unit. Add whether source means are exact or rounded.
Keep the rejected calculation in the review note when it already reached a report. State why thirty-five percent answers a team-score question. Preserve the original label and the corrected label so another reader can understand the change.
Write the Finding With Its Boundary Attached
A neutral report sentence reads: “In this fictional month, 47 of 100 eligible requests finish on time, an overall rate of 47%.”
The separate team score can follow if needed: “Giving each of the two teams equal weight produces a 35% mean team rate.” Those sentences make the aggregation choice visible.
An operator dashboard can carry the short label. Keep the full worksheet accessible behind it. The display should retain enough context to identify the numerator, denominator, and reporting period.
Before distributing a revised total, check which earlier reports used the old label. Decide whether their audience needs a correction. Record the reason and the source version used.
Forty-seven percent describes these synthetic requests under these rules. It establishes no cause, forecast, or statistical significance.
Keep the denominator attached to the answer.
Sources
- OpenStax: Measures of the Center of the Data. Arithmetic means and frequency weighting. Checked 2026.10.28. All business examples, worksheets, and calculations here are original synthetic illustrations.