Every negotiation of a covenant package includes the phrase "that's market". It is used to justify a grower basket at 30% of EBITDA, a twelve-month lookback on the equity cure, a portability clause with a ratio condition. It is usually said with confidence, and it is usually reconstructed from memory: the last three deals the speaker worked on, a league-table summary from a law firm, and whatever the other side said last time.
The strange thing is that the actual answer is sitting in the bank's own files. A desk that has closed two hundred facilities over five years has two hundred data points on every term it negotiates. It does not use them, because they are in PDFs.
What "market" means when you can count
Once a corpus of agreements has been read into a register, with every basket, threshold, definition and carve-out stored as a field with a citation, "what's market" becomes a query rather than a recollection.
- Distribution, not anecdote. The general debt basket across sponsor-backed deals in the last eighteen months: median 25% of EBITDA, interquartile range 20 to 30, three outliers above 35 and which deals they were.
- Segmented properly. By sector, by sponsor, by size, by year, by law firm on the other side. The clause a sponsor accepted in their last four deals is a stronger argument than an industry average.
- Trends. The equity cure lookback has gone from 24 months to 12 in the sample since 2022. The share of deals with a J.Crew blocker has doubled. The share with a Serta blocker has not.
- Off-market flags on a new deal. A draft arrives and every term that sits outside the interquartile range of comparable deals in the corpus is flagged before anyone reads it, with the comparables listed.
Why this needs a register, not a search box
Full-text search over a document store gets you to the pages. It does not tell you that the basket on this page is 25% of EBITDA and the one on that page is 20%, because the numbers are embedded in different drafting. The register is what makes them comparable: the same field, on the same basis, with the definition chain resolved so that "EBITDA" in one deal and "Consolidated Adjusted EBITDA" in another are recognised as the same leg of the same kind of basket.
The citation matters here as much as it does in monitoring. A precedent argument that can be opened to the page is persuasive. One that cannot is an assertion.
Negotiating leverage is knowing what you have already agreed to, across every deal, faster than the other side can remember what they agreed to.
The same corpus, three uses
We built precedent queries because customers asked for them, but the corpus that answers them is the same one that runs monitoring. The three uses reinforce each other.
| Use | Question | Who asks |
|---|---|---|
| Monitoring | Is this borrower compliant and where is the headroom? | Portfolio, credit control |
| Capacity | What can this borrower do under this document today? | The desk, risk |
| Precedent | What have we agreed before, and what is this draft asking for that we have not? | Origination, legal, credit committee |
A document read into the register for monitoring is automatically a precedent. A precedent query that surfaces an unusual term is automatically a monitoring question about the deal it came from. The corpus gets more valuable with every agreement, which is the argument for reading all of them rather than the ones currently causing trouble.
Start with what you have
The practical advice is short. The last five years of executed agreements, read into a register once, is an asset the desk already paid for and has never used. It is a bounded piece of work with a measurable output, and the first time a negotiator opens the distribution for a basket in the middle of a call, the case for it is made.