Search term negotiations go badly for one reason: each side argues about terms in the abstract. Your opponent wants “contract” and “agreement”; you say that is every email the company ever sent. Nobody has a number.
Courts settled the standard in 2008. In Victor Stanley v. Creative Pipe the court required the producing party to show how its keyword results were obtained, and the “we searched, trust us” era ended. The first search term reports appeared that year, built to document every query and its hit count so a search could be defended.
What the report says
A term report runs every proposed term against the processed collection and returns, for each one: documents hit, documents hit by that term alone, and families pulled in. “Agreement” hitting 41,000 documents, 38,000 of them hit by no other term, is not an argument. It is the end of one.
The negotiation then becomes arithmetic. Narrow “agreement” to “agreement w/5 (license OR supply)” and the report reruns. Volume comes down by narrowing terms, never by curating results, so every number matches a query the other side can run themselves.
Why the number is money
RAND's 2012 study of production costs put processing at 19% of spend and review at 73%. Every document a narrower term removes is a document nobody bills to read. On a matter with a $150,000 review budget, a term list that cuts volume by a third is worth $50,000 before anyone opens the platform.
How Marbury does it
Paste the request as it arrived: the other side's letter, the numbered list, the connectors, the date limits, all of it. The agent reads it, plans the run, and executes it once you approve. The report comes back as a table with hits, unique hits and families for every term, and an audit of exactly what was run, so each number traces to a query. Export it and attach it to the meet-and-confer letter.
Then narrow a term and it reruns. A negotiation can go dozens of rounds, and because the collection was processed once, each round takes minutes, not a vendor ticket.