Perception Failures
The hardest errors in credit are not arithmetic errors. They are seeing errors. The number is right and you still get it wrong, because you read it under the wrong frame, in a flattened dimension, with an eye that was never trained to see what mattered. This lesson is about the three ways perception fails — because a tool that only fixes arithmetic fixes the easy half.
Almost everyone who blows up on a CRE credit call did the math correctly. That is the unsettling thing to absorb. Their DSCR was right, their cap rate was defensible, their spreadsheet balanced. What failed was upstream of the math — in how they saw the loan. There are three distinct failure modes, and naming them is the beginning of not committing them.
One: the mislabeled cup. A measurement is only meaningful inside a frame. DSCR of 1.2x is “sufficient” — but only within a correctly identified regime. If the loan is labeled “suburban office recovering” when it is really “obsolete office in structural decline,” every number you compute is honest and every conclusion is wrong, because you poured a correct measurement into the wrong container. The failure is not in the number; it is in the label upstream of the number. This is the deepest reason a language model earns its place on the platform: not as a ninth signal, but as a detector of the frame — is this cup labeled correctly? — a job that is upstream of, and invisible to, any ratio.
Two: the flattened dimension. This is the lesson of Flatland: a creature living in two dimensions cannot perceive a sphere passing through its world — it sees only a circle that mysteriously grows and shrinks. When you collapse a high-dimensional situation onto too few axes, you don’t just lose detail; you can lose the exact dimension that owns the crisis. A portfolio can look diversified in two dimensions and be catastrophically concentrated in a third you didn’t plot. A metro can look calm on the axes you charted while the distress hides on the one you flattened away. The danger of a clean, low-dimensional picture is that it feels like understanding. The Flatland trilogy is the long-form version of this idea, and it is worth reading precisely because the failure is so seductive: the flattened view is always the comfortable one.
Three: the eye never reared to see it. Blakemore and Cooper raised kittens in a world of only horizontal stripes; grown up, the cats were functionally blind to vertical lines — not damaged eyes, but a visual cortex that had never been trained to represent verticals, so they walked into table legs they could not perceive. Analysts are reared the same way. If your entire career taught you to read office as a stable, coupon-clipping asset, you may be perceptually blind to the vertical line of a structural regime change — not stupid, not lazy, just never reared to see it. This is the most humbling failure because you cannot introspect your way out of it: the whole point is that the thing you can’t see leaves no gap where you’d notice it missing. Blakemore’s Kittens develops this into a discipline for deliberately re-rearing your own eye.
Put the three together and you have a map of how confident, competent people get CRE credit wrong: they misread the label (mislabeled cup), they trust a picture missing its crucial axis (flattened dimension), or they are perceptually blind to a whole category of risk (unreared eye). None of these is fixed by better math. All of them are fixed by changing what you can see — which is what the platform is ultimately for. The Loupe is the concrete instance: it lets you recut a securitized book at a finer resolution and watch an asset’s “office costume” fall away — the thing hiding one flattened dimension up, made visible by changing the frame. Seeing is the product. The arithmetic was never the hard part.