Why is our donor retention rate falling?
Frequently because the file is getting messier rather than because supporters are leaving. Retention counted on records rather than on resolved people falls every year that duplicates accumulate, since one supporter under three records reads as three people who each gave once. Resolve identity first, then separate donors who chose to stop from those whose card simply failed.
Section 01
Count people, not records
A supporter who gives every year under three records looks like three supporters who each gave once and left. Every duplicate turns one retained donor into two lapsed ones, and the error compounds each year the file grows.
So resolve identity before counting anything, and report how many records collapsed alongside the rate. That second number is the size of the correction, and it is usually the more persuasive figure in the room.
In one worked file, resolving people before counting moved retention from 47% to 58% and removed 312 names from a win-back appeal.
Section 02
Deciding to stop is not the same as failing to pay
A cancelled monthly gift and a declined card look identical in a totals report and need opposite responses: one is a re-acquisition campaign, the other is a phone call.
Look at what happened in the three days before each gap started. Payment failures cluster, often in processor batches, and a cluster is recoverable in a way that genuine lapse is not.
Section 03
Concentration is a governance fact
Income that grew while becoming dependent on eleven households is a worse year than it looks, and the board is entitled to know which it was.
Report the top one, five and ten together rather than a single threshold, and put an end date against any pledge, because a pledge in its final year and a renewing relationship look identical in a share-of-income chart.
Section 04
What to export from the CRM
The gift transactions with dates, amounts and a donor reference, and whatever identity fields the CRM will put on that same export: email, name, address. They have to be on it, because every figure is computed from one table and a separate constituent file is not reached into. That second file is still read and returned as a download, and the report says whether the two look related.
Any CRM that exports a spreadsheet works, and column names differ by system and by year, so there is no need to rename anything first. What matters is that soft credits are distinguishable from hard credits, or a matched gift gets counted twice.
See it on a real project
5,510 people were stored as 6,214 records, depressing reported retention by 10.9 points.
Keep reading