Cleanup Rule Bench

Synthetic data

Switch the cleanup rules on and off. See which rules make fewer mistakes, which find more, and what each one costs.

Change settings

Loading the evaluation data…

How to read this

  • Green is work done right. Red is the costly mistake: things joined that should stay apart. Amber is missed work: the record stays as it was.
  • Big numbers are per 1,000 donor records, so files of any size compare. Real totals are under them.
  • Safe to apply: at most 1 in 100 is wrong. Review first: up to 1 in 20. Risky: more than 1 in 20.

Results

What each switch does

Starting from your current settings, flip one duplicate or household switch on its own. Red bars show the change in wrong merges, amber bars the change in missed ones.

Technical: choose what the chart counts

Board connections

The tool links donors to board members of other nonprofits, found in IRS 990 filings. Move the sliders to see the links change.

The test files have no planted board look-alikes, so the sliders mostly change how many links are found. At the default settings, every link made is right.

How this was tested, and limits