Join marketplace data to your own records.
Matching links your internal records — clients, matters, accounts, locations — to external datasets, so marketplace data shows up where your teams already work.
Matching versus entity resolution
Entity resolution finds duplicates of the same entity. Matching answers a narrower question: which external record corresponds to this one of mine? The output is a join table you can use in your own systems.
Because the question is narrower, matching is usually faster to scope and easier to verify. You know your own records, so you can judge a sample of matches quickly, and the unmatched report tells you exactly where external coverage stops.
How a matching project runs
- 1
Describe your records
Share the fields you can match on — names, identifiers, citations, docket numbers, addresses, dates — and how complete they are.
- 2
Choose match keys
Exact keys first (an identifier or a normalized citation), then normalized keys, then scored fuzzy matching for what remains.
- 3
Match a sample
Run against a sample and review results with you. Thresholds come from that review.
- 4
Run in full
Produce the match table, the unmatched report and the review queue for ambiguous candidates.
- 5
Keep it current
Re-match when the external dataset or your records change, and report what moved.
What a match table contains
| Column | Meaning |
|---|---|
| your_id | Your record's identifier. |
| external_id | The marketplace record it matched, with dataset slug. |
| method | Exact key, normalized key, or scored fuzzy match. |
| confidence | Score from the matching method, with the threshold used. |
| matched_at | When the match was computed, and against which dataset version. |
| status | matched, ambiguous (several candidates), or unmatched. |
Matching a matter list to case law
A typical legal use: your matters reference cases by citation or name, and you want each one linked to a court opinion record. Citation strings are normalized and compared first; names, courts and dates are used where no citation is present. Shown as illustrative rows, not real data.
your_id,external_id,dataset,method,confidence,status
M-0001,lawco:case:<id>,lawco-court-opinions,normalized_citation,1.00,matched
M-0002,lawco:case:<id>,lawco-court-opinions,name_court_date,0.91,matched
M-0003,,lawco-court-opinions,name_court_date,0.62,ambiguous
M-0004,,lawco-court-opinions,,,unmatchedIllustrative example · final API surface documented in /docs
Where a match table goes next
Warehouse joins
Load the table next to your own data so BI dashboards can join matters, accounts or locations to external attributes. Keep the provenance columns so a chart can be traced back to source records.
Retrieval for AI
Give a RAG pipeline or agent a direct path from your record to the matched external record id, then fetch that record through the API or MCP at query time, with its license and retrieval time attached.
Alerts
Re-run matching on a schedule and post new or changed matches to a queue, so teams see new external records that relate to their work.
Data review
Use the unmatched report to find gaps in your own records — missing identifiers, inconsistent names — as well as gaps in external coverage.
What matching cannot tell you
A match says two records refer to the same thing, with a stated confidence. It does not say the external record is complete or current, and it does not extend the provider's coverage. If a dataset's record count or date range is listed as not yet verified, an unmatched record may simply be outside what the provider holds.
Matching also does not change licensing. A matched external record is still governed by its provider's terms, including attribution and any restrictions on bulk copying or model training.
Included
- Unmatched-record report
- Ambiguous matches routed to review
- Re-matching on dataset updates
- Runs inside your environment if required
- Match keys and thresholds documented
- Provenance kept on every matched external record
Questions buyers ask
Does our data leave our environment?
It does not have to. Matching can run in your environment through a private connection or private deployment.
Can we match against LAW.co case law?
Yes, through the live API. Your references are matched against LAW.co opinions retrieved one request at a time. LAW.co search is sampled (up to 50 matches) and has no citation resolver yet, so a match can miss cases a full index would find.
Can matching use citation resolution?
Not against LAW.co today. LAW.co has no citation-resolution endpoint, so the API's citation resolve operation returns 501 unsupported_operation. Citation matching would rely on normalized citation strings in records we can access.
What does a match cost in API usage?
Matching that queries the API is metered like any other record search, against your plan's monthly quota. Large one-off matches are better scoped as a service engagement.
Do you keep a copy of our records?
Only for the duration and purpose set in your agreement, or not at all if matching runs in your environment.
Want external data next to yours?
Tell us about your records and the dataset you want to join, and we will propose a match approach.