Ask any advancement shop or finance director where their capacity scores come from and you will get a surprisingly uniform answer. They come from one of three vendors that license the same underlying wealth database, wrapped in different front ends. The price tag is six figures annually. The consensus in the industry is that the data is flawed but tolerable.
The data is not tolerable. It is wrong in a specific, structural way that we have seen reproduce on every donor file we have rescored since 2024. Roughly 20 to 25 percent of major-gift prospects in these databases are entity-resolution failures. The vendor thinks two people are one person, or thinks one person is two. In both cases, the capacity estimate is meaningfully wrong.
The mechanism is simple. The vendors resolve entities on name, city, and date of birth when available. They do not resolve entities on the richer signals that matter. Household address history. Business affiliation. Political giving patterns. Property holdings. Family relationships through estate records.
When two senior executives at the same firm share a name within a metro area, the vendor will often collapse them into one record. When one person goes through a divorce and a name change, the vendor will split the record in two. In the first case, capacity is dramatically overstated because the vendor is summing two people's wealth onto one person's profile. In the second case, capacity is halved because half the signal is attributed to a phantom.
We see this pattern most clearly on the rescoring passes we run against university advancement files. On a typical R1 file of 100,000 prospects, we find between 1,200 and 2,400 collapsed entities (over-stated capacity) and between 2,800 and 4,500 split entities (under-stated capacity). The fundraising implication is direct. Development officers are spending time on false positives, and they are missing high-capacity prospects hiding inside the under-stated pile.
The good news is that this is a fixable problem. We fix it by running every name on the existing file through a deeper entity resolver that uses public-record signals the wealth vendors ignore. Property records, SEC ownership filings, campaign contributions, and corporate officer registrations produce enough signal, in aggregate, to catch most of the collapses and recover most of the splits. The bad news is that the vendors have no incentive to fix it themselves. Their product is priced on coverage, not precision, and the collapses look like more coverage.
If you are running capacity scoring on anything larger than a thousand prospects and you have not rescored the file against a second-source resolver, there is a meaningful chunk of your capacity pool that is wrong today.