Every valuation team has inherited The Workbook: fifteen linked worksheets, hundreds of line items, three generations of analysts deep. Nobody fully understands it. Everybody signs off on it.
The orthodoxy: keep it simple
The best valuation teachers have been saying it for decades, and Aswath Damodaran says it most memorably: he values companies with about a dozen line items, and his rule is blunt — don't add detail unless you can bring something to that detail.
The reasoning is sound, and it rests on two observations about giant models.
First, you stop knowing who runs whom. When an analyst says "the model valued the company at…", that's not a methodology statement — it's a dodge. Responsibility has quietly moved from the person to the spreadsheet.
Second, complexity hides error. And here's the uncomfortable part: errors don't happen in the hard cells. The terminal-value formula gets checked five times, by everyone, before the review. Errors live where nobody looks — the hard-coded FX rate on sheet 11, the row that stopped summing after last year's insert, the adjustment someone made in 2023 for reasons no one recorded. Not because those cells are difficult. Because they're invisible.
Why the rule exists — and what it actually rations
Notice what the orthodoxy is really about. It isn't a claim that detail is worthless. It's a rationing rule for a scarce resource: comprehension.
A small model you understand beats a big model you don't — every time, no argument. But data got cheap a long time ago. A terminal can pull a company's full history, and its whole sector, into a spreadsheet in minutes. Understanding didn't get cheap with it. So the discipline of the last few decades has been to throw data away until the model fits inside one analyst's head.
That was the right trade — under one assumption: that comprehension per line item is fixed and expensive.
What changes when every line can explain itself
Suppose every figure in the model carried three things with it, permanently: the exact source it came from (document, page, table cell), the logic that transformed it, and an explanation available on demand — in plain language, at review time, under client questioning.
Then the ranking changes. Small-and-understood still beats big-and-opaque. But big-and-understood beats both — because the extra detail now brings something with it, instead of costing something. Damodaran's rule isn't violated; it's satisfied at a scale that used to be impossible. You're not adding detail you can't bring anything to. The detail arrives already carrying its own explanation.
And the "errors live where nobody looks" problem inverts. A verification pass that cross-checks every extracted value doesn't get tired on sheet 11, doesn't skip the boring rows before a deadline, and doesn't assume last year's link still holds. Completeness is exactly the property machines are good at and stressed teams are not.
What this looks like in practice
Concretely, "understanding every line" is not a slogan; it's four working habits, each of which can be systematized:
Every figure keeps its source. When a client challenges a number, the answer is one click, not an excavation.
Checks run where nobody looks. Cross-source consistency, stale inputs, arithmetic breaks, unexplained movements — checked everywhere, every cycle, not just where attention happens to fall.
Adjustments carry their reasoning. The 2023 normalization has a recorded basis, so 2026's analyst doesn't have to choose between blind trust and re-deriving it.
Inherited workbooks get audited, not worshipped. The Hydra you inherit can be reviewed line by line — flagged, explained, and either defended or retired.
The honest caveat
None of this is a license to build Hydras. Judgment still decides which drivers matter, and a valuation with nine drivers you can argue about is still better than ninety you can't. The point is narrower and, we think, more useful: the ceiling on model size should be what you can defend — not what you can hold in your head.
The orthodoxy told you to shrink the model until you owned it. The better version of the same principle: own every line, and let the engagement decide the size.
"The model valued the company at…" was always a dodge. The goal is to be able to say, at any scale: I know where every figure comes from.