Why check before reporting?
Once a number reaches a donor report, it is hard to take back. A wrong figure can mislead decisions, and a donor who finds one error may doubt the rest. Checking takes hours; fixing a published error takes weeks.
Data quality is not perfection. It means the number is good enough for its purpose, and that you know where it is weak.
What are the five checks?
| Check | The question | Example problem |
|---|---|---|
| Completeness | Are all required fields and records present? | Missing age, empty location, a site that never reported |
| Consistency | Do the parts agree with each other? | Group totals that do not add up to the overall total |
| Plausibility | Is the value believable? | 500 people at a session in a village of 300 |
| Duplicates | Is anyone or anything counted twice? | The same person entered at two distributions |
| Source | Can you trace it to evidence? | A figure with no attendance sheet behind it |
How do you run the checks?
Use the same routine every reporting period.
- Count what you expected. List the sites and activities that should have reported, and mark who did.
- Scan for blanks. Look for empty required fields.
- Compare totals with parts. Add the groups, such as women plus men, and compare with the total.
- Look at the extremes. Sort values from highest to lowest and read the top and bottom. Ask whether each is real.
- Compare with last period. A jump of ten times needs an explanation. So does a sudden drop to zero.
- Look for repeats. Search for the same name, date of birth or identifier appearing more than once.
- Spot-check the source. Pick a few records and compare them with the paper or original record.
How should you handle problems you find?
Do not guess, and do not quietly delete.
- Query the collector. Ask what happened. Many errors have simple causes, such as an extra zero.
- Correct with evidence. Change a value only when the source confirms the right one.
- Log the change. Record the original value, the new value, who changed it, the date and the reason.
- Mark what you cannot fix. If you cannot resolve a value, flag it, exclude it or footnote it, and say which.
Which problems appear most often?
- Typing errors, such as 1,000 for 100.
- Different people using different definitions. See how to write an indicator reference sheet.
- Late submissions that miss the reporting window.
- Dates in different formats.
- The same person registered at several sites. The article on reach totals and double counting shows why this matters for your totals.
How do checks fit with approval?
Checks find problems. Approval is the decision that a figure is ready to count as a result. They go together: a reviewer who has checked the data can approve it with confidence, and a reviewer who has not should not. Read more in why field data needs an approval step.
Here is a short example of the checks at work. A district officer receives a monthly figure of 4,800 people trained from one site. Last month the figure was 480.
- Plausibility: the jump is tenfold, so the officer asks the site.
- Source: the site finds an extra zero typed into the form. The attendance sheets show 480.
- Fix: the site sends the corrected figure with a note.
- Log: the officer records the original value, the corrected value, the reason and the date.
Without the check, the error would have raised the district total by thousands and reached the donor report. The whole fix took one phone call.
Build a short checklist on one page and use it every period. Over time, add the errors your own team makes most often. A checklist that grows from your own mistakes is the most useful kind.
Think about timing too. The best moment to check is soon after collection, while the enumerator still remembers the visit and can be reached. A weekly check on a small batch is easier than a large check at the end of the quarter. Agree a cut-off date for submissions, tell sites about it early, and decide what happens to data that arrives late: it may go into the next period, with a note. Clear timing rules make the checks calmer and the reports more reliable.
Where EMPOBase fits
In EMPOBase, a field team or partner submits a value through a collection link, and each submission keeps its submitter, note and timestamp. A reviewer then reads the submitted value beside the current recorded actual and approves or rejects it with a note. The software gives the reviewer the information; the checks above are still done by a person. EMPOBase does not deduplicate beneficiaries, and collection links need an internet connection. See the field-to-results workflow for the steps.