Resolve who’s who. Infer who matters. Then act on it.
DarkMath connects the customer records your systems scatter, adds the market context they can’t see, and turns the result into decisions your team can act on. Automotive first, and anywhere identity decides the outcome.
One person, 4 records. Confidence 96%.
Same name, same address, different person.
Prepared for review
Three products. One path from messy records to the next move.
Each product stands on its own. Together they answer three questions in order: who is this, what does it mean for us, and who else is like them.
DarkMatch
Identity resolution that recognizes the same person across nicknames, typos, moves and missing fields.
- Deduplicated customers and households
- Persistent IDs and updated contact data
- Layers on top of your current matching
DarkAgent
A business workspace. Ask in plain language, see the evidence, and move to a supported next step with approvals built in.
- Answers in market context
- Audiences with counts and contact coverage
- Recommended and completed actions, never confused
DarkMirror
Semantic audiences that find people who behave like your best customers, even where the fields your model needs are missing.
- Lookalikes built on behavior, not just demographics
- Exclusions applied before expansion
- For marketing reach, never eligibility decisions
Context changes the answer.
Sales up 8% looks like a good quarter. If your comparable market grew 14%, you lost ground.
Your CRM can’t tell you that, because the answer lives outside it. DarkAgent reads your data next to the market around it, shows where the gap is concentrated, and prepares the audience and offer to close it.
Quarter over quarter
Your share of the market: 9.6% to 9.1%
Illustrative scenario with synthetic figures. Not a customer result.
Measured, not claimed.
In a June 2026 benchmark, DarkMath and a leading identity graph processed the same 1,000,000 records representing 250,000 real people. Here is what came back.
F1 score, which combines precision and recall
of true links found
of DarkMath matches correct
identities against a true 250,000. The incumbent graph returned 459,791.
Your data stays yours.
Choose where DarkMath runs. Every option keeps re-identification behind your authorization.
Hosted by DarkMath
Send files or connect a share. The fastest way to start a test or a production feed.
In your environment
DarkMath software runs inside your cloud. Connected data travels on anonymous IDs, match keys and audiences, not personal data.
Dedicated clean room
For enterprise programs. Raw records stay inside. Only anonymous IDs and aggregates come out.
- Re-identification only at your authorization gate
- Regulated data handled under the FCRA, DPPA and state privacy laws by a registered data broker
- You decide which actions DarkAgent may prepare or run, and who approves them
Built for industries where identity decides the outcome.
Start with a test, not a contract.
Anyone can quote a match rate. We would rather show you yours. Most match tests take about two weeks, and you keep the report either way.
Request a match testSend a sample
Hashed or raw, your choice.
We build known answers
Controlled errors go into copies of records whose true identity is known.
Same input, both methods
DarkMatch and your current method process identical files.
Standard report
Match rate, precision, recall, duplicates, households and net-new identifiers.
Decide how to connect
Batch, API, in your environment, or a dedicated clean room.
Questions buyers ask
What is DarkMath?
DarkMath is an identity and decision platform with three products: DarkMatch for identity resolution, DarkAgent for analysis and supported actions, and DarkMirror for semantic audience expansion. It runs on regulated data infrastructure operated by its sister company, DecisionLinks.
What is identity resolution?
Identity resolution decides whether different records describe the same person, household or business. Rules-based matching links exact or near-exact fields. Probabilistic matching estimates the odds of a match. DarkMath adds semantic matching, which compares what records mean, so it recovers links that text rules miss and avoids merging people who only look alike.
How is DarkMath different from a CDP?
A CDP stores and syncs customer records. DarkMath improves what those records mean: it resolves identities the CDP’s rules miss, connects them in a Cognitive Data Graph, and lets your team ask questions across your data and relevant market context. Keep your CDP. DarkMath makes it more accurate.
What is DarkAgent?
DarkAgent is the business workspace for DarkMath. You describe an objective in plain language, DarkAgent shows the evidence, recommends a next step and prepares the audience, message or order for review. Execution follows the permissions and approvals you set.
Do we need DecisionLinks data to use DarkMath?
No. Start with your own data. DecisionLinks and other licensed sources can add reference data or market context when a question needs it and your use is permitted.
Where does our data go?
You choose. DarkMath can host the work, run inside your cloud environment on anonymous IDs, or operate in a dedicated clean room for enterprise programs. Re-identification happens only behind your authorization.
How do we evaluate DarkMath?
Run a match test on your own data. We build a test set with known answers, process the same input through DarkMatch and your current method, and return a standard report on match rate, precision, recall, duplicates and households.
Which industries does DarkMath serve?
Automotive first, plus lending and finance, data providers, advertisers and agencies, and demand-side platforms. The common thread is customer data where identity decides the outcome.
See what your data means in context.
Bring one question your team needs answered. We will show you the evidence, the audience and the next step.