How does double counting happen?
Imagine three activities in one village: a hygiene session, a food distribution and a vaccination day. Fifty people attend each. A simple sum gives 150. But many of the same people came to all three. The real number of different people may be 70.
Neither number is wrong. “150 participations” is true. “150 people reached” is not.
What are the usual causes?
- Repeat attendance. The same person joins several sessions.
- Several activities, one person. Services overlap in the same community.
- Several sites. A person registers at two distribution points.
- Several partners. Two organizations each count the same household.
- Several periods. A person is counted each month and the months are added.
- Cumulative totals. A running total adds people already counted last quarter.
What is the difference between records and people?
| Measure | What it counts | Example |
|---|---|---|
| Transactions or records | Each event, entry or item | 150 attendances |
| Unique individuals | Each different person, once | 70 people |
| Households | Each different household, once | 40 households |
Pick the measure that fits the indicator. Write it in the indicator reference sheet, so everyone counts the same way. See how to write an indicator reference sheet.
How do you count unique people?
- Give each person an identifier. A unique number, a card, or a registry entry that stays with them.
- Record the identifier at every activity. An attendance list with names only will not do; names are spelled in many ways.
- Define the counting period. For example, “unique people who received at least one service this year”.
- Remove repeats. Compare entries by identifier, and review near matches by hand.
- Keep the original records. Deduplicate in a copy of the data, so you can always go back.
Identifiers hold personal data, so protect them. Our article on keeping a beneficiary registry that protects people covers this.
How should you report when you cannot deduplicate?
Be honest and clear.
- Label the figure: “attendances”, “distributions” or “records”.
- Avoid the word “people” unless it is true.
- Add a note explaining the limit: “People who attended several sessions are counted each time.”
- If you can, estimate overlap from a small sample and state that it is an estimate.
- Where several projects overlap, do not add their totals and call the result unique reach.
Donors usually prefer a clear, honest limit to a total that cannot be explained.
How does this connect to disaggregation?
Every breakdown inherits the problem. If the total is inflated, so is the number of women, children and people with disabilities. See disaggregating by sex, age and disability. Fix the count first, then split it.
It helps to decide what you are trying to say before you pick a measure. Two sentences make the point:
- “We held 150 participations across three activities.” This is a count of records, and it is true.
- “We reached 70 different people.” This is a count of unique people, and it needs deduplication.
Both are useful. Participations show how much service you delivered. Unique people show how many were reached. Donors may ask for either, so read each indicator definition closely and ask the donor if it is unclear.
If you can only give one, give the one you can support, and label it. A clear label stops readers from drawing the wrong conclusion. It also protects your organization: a figure that is clearly labeled cannot later be called misleading.
For future projects, plan for unique counts from the start. Choose an ID method, train staff to use it at every activity, and agree who runs the deduplication and how often.
Also consider how you present trends. If you switch from counting records to counting unique people between periods, the numbers will drop, even though the work has not changed. Explain such changes in the report and, where you can, restate earlier figures on the same basis. Readers who see a sudden fall without an explanation may assume the project is failing. A short note on the method keeps the series honest and comparable from one period to the next.
Where EMPOBase fits
EMPOBase totals are counts of records. A sum of submitted values or distributed items counts transactions, not distinct people, and nothing in EMPOBase deduplicates beneficiaries. Report those totals as records, or deduplicate your data yourselves before you report unique reach. The software can hold your records, indicator actuals and approval history in one place, but the counting rule and the deduplication are decisions for your team. See the field-to-results workflow, which states this limit.