Match two databases without handing either one over.
DataMatched finds the people two databases have in common. Neither side sends its file to the other side, or to a third-party bureau. Nobody sees anyone else's records, and each side learns only the result you agreed.
Now inviting early-access partners.
Your customers vs. a partner database
31,240
The problem
Data matching usually means trusting someone with your file.
To find out who you have in common with another company, someone normally ends up holding both lists.
- You upload your customer file to a bureau, and hope it's handled and deleted properly.
- "Hashed" emails or names are sent instead. But simple hashes can be guessed back by trying every likely value.
- Copies end up in inboxes, servers and backups you don't control.
- With DataMatched, your file stays on your own system. Only locked codes are compared, and only the agreed answer comes out.
How it works
The double lock.
Each side locks its own records with a secret key. Each then adds a second lock to the other side's codes. Codes only line up when the same person is in both files.
You lock your records
On your own system, each record is standardised and locked with a secret key only you hold. The key never leaves your system.
They lock theirs
The other side does the same with its own secret key. Each side adds its lock on top of the other's codes.
Only matches line up
A double-locked code is identical on both sides only when the person is in both files. Everything else stays locked and unreadable.
Agreed result only
Only the agreed answer comes out, such as the overlap count. The keys for that job are then thrown away.
The method is called Private Set Intersection. It uses elliptic-curve cryptography (P-256), the same family of encryption that protects online banking and card payments. No one can work out a key, or a record, from the locked codes.
Security
No one sees your data. Not the other side, not us, not anyone in between.
Security is the whole point of DataMatched, so here are straight answers to the questions we'd ask.
What if someone intercepts the data?
They only see locked codes, sent over an encrypted connection. Without both secret keys, the codes are meaningless.
What if our servers are hacked?
Your file was never on them. There is nothing of yours to steal except locked codes, and your key stays on your system.
Can one of our staff look at my data?
There is nothing to read. Job logs hold counts and job numbers only, never names, addresses or other personal details.
Could the other side guess who's in my file?
Each match is tied to an agreed audience and capped in size. Your system enforces the cap, so no one can run long lists of guesses against your data.
Can the codes be reversed?
No. Simple hashes can be reversed by trying every likely value. Our codes are locked with secret keys nobody else has, so guessing doesn't work.
What does each side end up knowing?
Only what you agreed in advance. For a count, you learn the overlap and we learn only the size of your list. To leave your customers out of an order, we learn only which of our own records to remove.
Is anything kept afterwards?
No. Match results are deleted once the job is reconciled, and the secret keys for each job are discarded.
How strict is a match?
A match needs the exact full name and full address. Incomplete or unclear records are reported separately and are never counted as a match, or as new.
Compared with the usual way
What changes with DataMatched.
| Upload to a bureau | Swap hashed lists | DataMatched | |
|---|---|---|---|
| Your file leaves your system | Yes | Yes, scrambled | No |
| Someone else holds both lists | Yes, the bureau | Yes, whoever compares | No one |
| Records can be guessed back | Not needed, they're readable | Often, by trying likely values | No, codes are keyed |
| What each side learns | Everything | Which records match, on both sides | Only the agreed result |
What it's for
One secure match. Several jobs.
Find new prospects
Pick an audience in ProspectCount and leave out the customers you already have, so you only buy new names.
Improve your own data
See which details can be added or updated for your customers. Results come back against your own customer references.
Measure overlap with a partner
Find out how many customers you share with another company before agreeing a joint campaign or data deal, without either of you showing your list.
Clear quotes, before you pay
See how many customers are affected and how many fields can be added, based on actual matched numbers, not an estimated match rate.
Good to know
Questions
Is DataMatched available now?
We're inviting early-access partners for a pilot. The matching method has been tested on made-up data, including checks that no known customer slips into an order. The software each side runs is being built now and will be independently security-reviewed before any live use.
Do I need to install anything?
A small piece of software runs on your side to prepare and lock your records, so your file never has to leave your system. We'll help you set it up.
Which details are matched?
Full name and full address, matched exactly. Records with missing or unclear details are listed separately so you can see and fix them.
Can it match against databases other than ours?
That's the plan: any two organisations should be able to compare lists the same way. Talk to us if you have a partner in mind.
What about data protection rules?
DataMatched is designed so each side learns as little as possible: only the agreed result. Your data protection officer should still review each use, and we'll give you full documentation of what is exchanged.
Match with confidence.
Tell us what you'd like to match and we'll walk you through a pilot.
Talk to usor email help@datamatched.com