Why the Death Master File is not enough on its own

What the Death Master File actually is

The Death Master File is a record of deaths reported to the Social Security Administration. It exists because the SSA needs to stop benefit payments, and it grew into a de facto industry reference because it was large, cheap and easy to obtain.

It carries name, Social Security number, date of birth and date of death, which is enough to be genuinely useful. Many institutions built their deceased controls on it a decade or more ago and have not revisited that decision since.

That is the problem. The file changed, and a lot of downstream controls did not.

The two structural gaps

The first gap is source. Deaths reach the SSA from several places, including funeral homes, family members, financial institutions and state vital records offices. State reported deaths were removed from the version of the file made publicly available, which took a very large volume of records out of what most commercial users can see. If your provider is repackaging the public file, that volume is missing from your screening too.

The second gap is time. Access to the most recent records is restricted to certified users with a legitimate fraud prevention purpose and appropriate safeguards, under a certification program with real requirements attached. Everyone else waits. The waiting period is measured in years, and the entire fraud exposure window sits inside it.

Put those together and you get the outcome teams actually experience. The file confirms that someone who died in 2019 is deceased, which nobody was going to lend to anyway, and says nothing useful about someone who died last month, which is exactly the identity being used against you right now.

How death data reaches a commercial file

Understanding the chain explains why providers differ so much even when they all claim nationwide coverage.

A death is certified by a medical certifier and reported by a funeral director. It is registered with the state vital records office, increasingly through an electronic death registration system, though adoption and processing speed are not uniform across states. The state maintains the authoritative record, and onward sharing is governed by state law and by agreements that vary considerably.

Commercial providers assemble their file from whichever of those channels they can lawfully access, then match, normalise and deduplicate. Every hop adds latency, and every missing channel adds a gap. Two providers can both be honest about nationwide coverage while being weeks apart on the same death.

• Certification and funeral director reporting happens within days
• State registration follows, with speed varying by state and by system
• Onward availability depends on the channel and the agreement behind it
• Aggregation, matching and publication add the final stretch

Questions that separate providers

Coverage claims are almost useless as a comparison tool because everyone claims the same thing. These questions are not.

Ask them in a proof of concept rather than in a questionnaire, because the answers only mean something when you can check them against your own book.

• How many days from date of death to the record being queryable by us, at the median and at the ninetieth percentile?
• How many new confirmed records do you add each week, and how does that compare to national mortality?
• Which states are slowest for you, and by how much?
• What is your process when a record is later corrected, and how do we learn about it?
• Do you return date of death and last known address, or only a deceased flag?
• What license or certification governs your access, and can we see it as part of vendor review?

What good looks like in practice

A workable deceased data input is fresh enough that the median record arrives within days rather than months, complete enough that it does not have a state sized hole in it, and specific enough to return a date of death and an address so you can adjudicate rather than guess.

It should also be honest about what it is. A provider that tells you which states are slow and what their correction process looks like is a provider that has actually measured it.

None of this means abandoning the DMF. It means treating it as one input into a deceased decision rather than as the answer, and covering the two gaps it structurally cannot cover.

Common questions