How to Combine Indicator Results Across Projects Without Double Counts

When Can Results From Different Projects Be Added?

Two projects can report the same indicator name and still not be addable. Before summing, check four things. If any answer is no, report the results side by side instead of as one figure.

1. Same definition. "Farmers trained" means the same thing in both: the same minimum attendance, the same kind of training.
2. Same unit. Both count people, or both count households, or both count sessions. A person count and a household count cannot be combined.
3. Same period. Both cover the same reporting period, or the same cumulative window.
4. No overlap, or overlap removed. No unit appears in both counts, or the overlap has been found and taken out.

Put the rule in the indicator reference sheet for the combined indicator, so every project applies it the same way. The sheet should state whether the figure is a total, a unique count, or both.

Total Counts vs Unique Counts

A total count (also called participations or reach events) counts every time a unit is served. A unique count counts each person or household once, however many times or through however many projects they were served.

A caregiver who attends three nutrition sessions is three participations and one unique person. A caregiver who attends a nutrition session run by one project and a hygiene session run by another is one unique person but two participations. Neither figure is wrong. They answer different questions: total counts show how much service was delivered, and unique counts show how many people were reached. Reporting a total as if it were a unique count is the most common form of double counting.

What Can Be Added and What Cannot

The rule depends on the type of indicator. The table gives the usual treatment.

Where Overlap Comes From

Overlap is not a sign of poor work. It usually reflects projects that target the same communities. The common sources are:

• Two or more projects in the same village or district, each registering the same households.
• The same person served across years by one project, then reported again in a cumulative total.
• Several services to one person inside a single project, each counted as a separate beneficiary.
• Several partners delivering to the same person under one award, each reporting its own list.
• Re-registration, where a person gives a different spelling of a name or a different phone number at each contact.

List the sources that apply to your portfolio before choosing a method. The method should fit the source of overlap.

Four Ways to Remove Double Counting

Whichever method you use, write it down and apply it the same way each period. A change in method between periods makes the trend impossible to read.

Protect the data. Matching people across projects means sharing personal data. Share the minimum needed: an ID, or a coded version of the name and birth year, rather than full records. Check that consent forms cover the sharing, and agree who holds the combined list. See data management.

Percentages and Averages

Percentages and averages are not additive, and the usual error is to average them. Suppose one project reports 62% coverage of a target group of 5,000, and another reports 40% of a target group of 8,000, with no overlap. The combined figure is not 51%. Go back to the counts: 3,100 plus 3,200 people covered is 6,300, out of 13,000 in the target groups, which is 48%. Add the numerators, add the denominators, then divide. The same applies when a figure is built from disaggregated groups: recombine the counts, not the percentages.

Steps to Combine Results

1. Write the definition, unit and period for the combined indicator.
2. Check each project's indicator against it and note any differences.
3. Decide whether the figure is a total, a unique count, or both.
4. Identify where overlap could occur and which method will remove it.
5. Collect counts from each project, plus the minimum identifying data if matching.
6. Deduplicate and record how many records were removed.
7. Compute the combined figure from counts, not from percentages.
8. Report the figure with its method, its period and any project left out.

Add a check from data quality assurance: confirm that the totals add up, and inspect a sample of matched records to see that they are true matches.

Worked Example: Healthy Futures and Safe Wells

Healthy Futures is a fictional child nutrition program. Safe Wells is a fictional water project working in some of the same villages. Both train caregivers, and the organization wants one combined figure. The numbers are invented to show the arithmetic.

Total participations. Adding the two rows gives 7,300 caregiver trainings. The team labels this "total trainings delivered."

Unique caregivers. Both projects use a household registry ID. Sorting the two lists by ID finds 900 caregivers on both. The unique figure is 4,200 + 3,100 - 900 = 6,400 caregivers.

Coverage. The two target lists share 2,000 caregivers, so the unique target group is 8,000 + 5,000 - 2,000 = 11,000. Combined coverage is 6,400 divided by 11,000, or 58%. Dividing 7,300 by 13,000 would give 56%, which counts the overlap twice in both the top and the bottom of the fraction.

Report line. "6,400 unique caregivers were trained, out of 11,000 targeted (58%). Trainings delivered: 7,300. 900 caregivers attended both projects. Method: matched on registry ID."

Common Mistakes

• Reporting total participations under a label that implies unique people.
• Adding two indicators that share a name but differ in definition.
• Averaging percentages instead of recombining counts.
• Deduplicating one year and not the next, so the trend jumps.
• Leaving out the method, so a reader cannot tell what the figure means.
• Building a cumulative total by adding each year's unique count, which counts returning people again.