Metrics & Early Warning
Everyone can read a level. The trained eye reads the derivative. Early warning is not about the number being bad — it is about the number moving, and about one small, unglamorous number that moves before the others.
You know DSCR as a level: above 1.0x the loan covers its debt, below it doesn’t. That level matters at the extremes, but as an early-warning tool it is almost useless, because by the time coverage crosses 1.0x the trouble is no longer early — it has arrived. The reader’s instrument is not the level. It is the trajectory. A loan at 1.6x drifting toward 1.3x over four quarters is telling you a story; a loan flat at 1.25x for three years is telling you a different, calmer one, even though the second sits lower. You are watching the slope, not the height. Distress is a process, and a process has a direction before it has an outcome.
Now the sharpest early-warning signal in the whole discipline, and it is deliberately humble: the 30-to-89-day delinquency. Not default. Not special servicing. Just a borrower who was thirty to eighty-nine days late — still current-ish, not yet seriously delinquent, easy to dismiss as a clerical hiccup. It is precisely this near-invisibility that makes it valuable. When you look at the record, an elevated 30–89 bucket leads serious noncurrency by a wide margin — populations that light up here go on to real distress at two to three times the rate of those that don’t. The borrower who is chronically a few weeks late is showing you a cash-flow strain that has not yet reached the coverage ratio. The late payment is the tremor before the number moves.
But this signal comes with a discipline you must internalize, because it is where careless readers go wrong. Read it relative; condition it absolute. Read relative means the signal lives in the change — a book whose 30–89 rate is climbing against its own history, or running hot against its peers, is the alarm. A single absolute reading in isolation tells you little. Yet you also condition on the absolute band, because “30–89 is elevated” means something different in a benign credit environment than in a stressed one; the same relative jump sits inside a different baseline. So you hold two things at once: the movement is the signal, and the level is the context that tells you how loud the movement is. Miss the first and you’ll never see anything coming. Miss the second and you’ll cry wolf at every wiggle.
Why does such a small number work so well? Because it sits early on the same axis everything else sits on. A commercial mortgage really only fails two ways — it can’t cover the debt, or it can’t exit at maturity — and the 30–89 delinquency is the earliest visible tremor along the first of those, the coverage axis. It is not a different signal from DSCR; it is DSCR’s early cousin, the cash-flow strain showing up in payment behavior before it shows up in the reported ratio. This is the low-rank truth the whole course rests on: a few axes, watched early, carry almost all the information. The reading What the Finance Class Was Teaching traces where that conviction comes from — a thirty-year-old lesson about bond duration, rediscovered on the tape.
The platform’s Track Record is where you can see this hold up out of sample: the 30–89 signal replays across every annual cohort since 2001 (the year the bucket began to be collected), and in cohort after cohort the flagged population goes on to distress at meaningfully higher rates — the same lift through the GFC, COVID, and the current cycle. That stability across regimes is what turns a plausible idea into a signal you can lean on.