Songlinks › Bulk lookup

Bulk songwriter lookup: match a whole roster at once

Send us the writers you administer. We send back, for each one, every registration we can group to them — across spellings, stage names and more than one IPI name number — with the evidence behind each join, and the records we deliberately left out.

Why matching a catalogue is harder than running a lot of lookups

A roster is a list of people. Registration data is a list of names, and the two do not line up. Three things get in the way, and all three are ordinary rather than exceptional:

• One writer is several records. A legal name, a short name, a romanisation, a misspelling, a second IPI name number: each is its own record with its own slice of the catalogue, and nothing in the data says they are one person (worked through here).
• One name is several people. In a random sample of registration records, 8% of songwriter records that carry an IPI name number share their exact registered name with a record under a different number. Matching a roster on name alone merges those catalogues.
• Most records carry no number to match on. About six in ten songwriter records in that sample — 1,639 of 2,657 — have no IPI name number at all. A roster reconciled on IPI never reaches them.

Sample figures measured 2026‑09‑15 on the registration data Songlinks reads: 2,826 records drawn at random. The guides show the working.

What a run over a hundred writers actually found

We took a hundred real registered names at random, sent them through as a roster and counted what came back. Not a demonstration built from chosen examples — a random list, run the way yours would be.

So the searching is the smaller half of the job. Deciding which of the records a name reaches are the same human, and which are a different person with the same name, is the work — and it is the part that does not get faster when you do it a thousand times.

That decision is what our proprietary algorithm makes. It was built for this one question, scores every pair of records for whether they are the same human, and returns the reason with the answer — so an audit is a set of findings you can check, not a verdict you have to believe.

Reading those hundred names took 95,976 registrations, because a name does not tell you which records it might mean until they have been looked at. Nine of the writers had their catalogue split across more than one registration, one across four. Thirteen pairs came back unsettled rather than decided — the evidence did not reach either answer, and saying so is the point.

What you send: your writer list

• The writers. Names as you hold them. A spreadsheet or CSV is fine.
• IPI name numbers where you have them. Useful, never required — the writers without one are usually the reason you wanted the audit.
• Your own identifier for each writer, if you want one carried through the output so the result joins straight back onto your system.

What comes back: every registration, per writer

For every writer on the list:

• The records grouped to them, with the works counted once across the group rather than once per record.
• The evidence for each join — shared works, shared co-writers, shared publishers, matching numbers — record by record, so each one can be checked rather than taken on trust.
• The records we reached and did not join, with the reason. This is the half worth reading: it is where a wrong merge would otherwise have been.
• The pairs the evidence does not settle, marked unsettled rather than decided either way.
• A Songlinks person ID, permanent and never reused, and a page of its own.
• The writers we found nothing for, kept separate from the ones we found and then ruled out. They are different answers and they need different follow-up.

Whole-database matching for societies and registries

The same question asked of a whole catalogue rather than a roster: which of your own registrations are the same human being. The proprietary algorithm is the same one either way; only the input changes. That run is one we already do, on the data behind this site, and it is what mints the person IDs a lookup here gives back.

Record count measured 2026‑09‑18; the person IDs are the production registry released with version 0.9.0. Person IDs are minted for writer registrations carrying an IPI name number, so the two numbers are not a ratio. A bulk run is that machinery pointed at your list.

Run for you as a managed service

Nothing to install, integrate or learn. Send the list and we run it on the full registration data, on hardware sized for the job, and send the results back in a format that joins onto your own system. Repeat it whenever your roster or the registrations move.

Tell us how many writers are on the list and we will tell you what a run looks like. A dozen names is enough to see the output before anyone talks about a whole catalogue. If you would rather call it from your own code, say so — a batch API is built on demand, and demand is what decides the order we build things in.

What an audit answers

Which registrations belong to the same person — identity, with the evidence attached. Not a valuation, and not a statement of what anyone is owed: the word audit here means the writer list reconciled against the registrations, nothing about money. Every join lists the shared works, co-writers, publishers and numbers it rests on, so each line can be checked against the registrations themselves. Songlinks is in beta and no person has checked the groupings, which is exactly why the evidence ships with them.

Ask for a sample run

Email alex@songlinks.to with roughly how many writers you hold and what you already have for them. A dozen names is enough to see what the output looks like before anyone talks about a whole roster.

Email alex@songlinks.to

Or try a single lookup first — free, with nothing to sign up for.