Vol. 020: What a Legal Pad Costs

Current Conditions

Last issue named a gap: the cost of reading the public record fell, and the practice built around expensive reading did not move with it. The natural objection is that the old way was never that expensive in the first place. “Development is local,” meaning a person good at their job can sit down with a legal pad, make a few calls, watch a couple of hearings, and arrive at a defensible read in a week. That objection deserves a real answer, because if the manual baseline is cheap, there is no gap worth discussing.

So this issue puts a number on the read itself, in hours and in dollars, for one deal in one jurisdiction. First the manual baseline, the legal-pad method, on its own terms. Then the same read priced at today’s cost of compute, because a baseline only means something next to the thing it is a baseline against. The first number explains a behavior that otherwise looks like negligence. The second explains why that behavior is about to start looking expensive.

Signals Observed

Start with the size of the thing being read. Take one jurisdiction and a two-year lookback, which is the shortest window that reliably captures a board’s posture toward a product type. In the public record for the Triangle, that window holds roughly 106 meetings in Wendell, 168 in Cary, 189 in Holly Springs, 194 in Raleigh, and 309 in Apex. Measured in the searchable segments of actual discussion, the same window runs from about 4,400 in Wendell to roughly 13,900 in Holly Springs and 18,500 in Apex. That is before the ordinance, which in Cary alone is 2,918 sections, before the comprehensive or small-area plan that governs the parcel, and before the fee schedule that prices it. The relevant record for a single deal is hundreds of meetings, thousands of segments of testimony and motion, and a few thousand pages of code and plan.

Now put a person against it. A genuinely complete read, watching or reading the meetings that touch the product type, working through the relevant ordinance and plan sections, reconciling the fee schedule, and writing it down, is not a one-week task. It is closer to 150 to 300 hours of senior attention, the kind of attention that can tell a routine continuance from a meaningful one. Most of that time is in the meetings, where the binding signal lives and where there is no index. At a loaded senior rate of $150 to $300 an hour, the complete read costs somewhere between $25,000 and $75,000 per jurisdiction, per deal. Those are estimates with visible inputs, so move them as you like; the conclusion holds across the range. The complete read is a five-figure line item, and no single-deal diligence budget carries it.

Which is why it is not spent. What happens instead is the thin sample: the ordinance section that names the use, the two or three hearings someone flagged, the calls institutional memory suggests. Against the record just described, that is coverage in the low single digits, and it is triage, not laziness, under a cost the legal pad cannot escape. For a long time it was the only available strategy, so it hardened into the definition of competent diligence. But the unread remainder does not vanish. The cost moves, and it is paid later, in the surprises this newsletter has spent eighteen issues cataloguing: the denial the prior hearings predicted, the fee spread published before the contract, the rule that moved while nobody was reading. The legal pad does not remove the cost of reading the record. It defers it and randomizes when it lands.

Now price the same read with the record already structured. These are not projected numbers, they are today’s posted rates for compute. Ingesting three years of one jurisdiction’s meeting transcripts, the raw material behind that 200-hour read, runs ten to twenty dollars. A synthesized lookback off the structured corpus is a few dollars more. Call it roughly twenty-five dollars in compute for a one-time, two-year read of a single jurisdiction, plus about an hour of senior time to review the output and decide what it means. Set the two columns side by side. The manual read is roughly $50,000 and 200 hours. The machine-assisted read is roughly $25 and one hour. The senior judgment, which was always the actual value, is the one line item in both columns, and it is the smaller one.

And most of that gap does not depend on trusting the machine to write the summary, which is the objection that usually arrives here. The synthesized report is the part that asks for trust, and it is the cheap part, a few dollars. The expensive part of the manual read was never the thinking. It was the sourcing: figuring out which meetings exist, pulling the minutes and video, digging through them for the handful of passages that bear on the deal. A structured record collapses that to a search. The practitioner who distrusts a generated summary and wants to read every relevant segment himself still skips the entire sourcing step, and the sourcing is where the hours were. Trust the synthesis or don’t. The database removes the digging either way.

One caveat keeps that honest. The twenty-five-dollar figure is the marginal cost once the record is already structured, and structuring a corpus this size is a real cost in its own right, paid once and spread across every read that draws on it. So the fair comparison is not free against fifty thousand. It is one large fixed cost amortized across thousands of reads against a five-figure manual cost re-paid in full on every deal, because the legal pad amortizes nothing. Each deal starts from a blank page.

Forecast

The thin sample was rational for as long as the read was a five-figure manual cost. That condition has lifted on one side and not the other: senior judgment still takes senior time, but the reading those hours went to is no longer the expensive part. With that constraint gone, the thin sample stops being triage and becomes a choice that is hard to defend on cost. The likeliest path is not a sudden conversion but the divergence Vol. 19 forecast: a small set of operators who have repriced the read and ask larger questions earlier, and a larger set still running legal-pad math and attributing the surprises to a hard market. The gap is small on any single deal and compounding across a portfolio.

Watch Items

• The loaded cost of senior land and entitlement talent, which prices every manual reading hour and is rising with turnover
• Whether machine-assisted reads start appearing as a billed line item in third-party diligence scopes, marking the baseline shifting from hours to a fixed cost
• Lookback length as a discipline; the two-year window used here is itself a sample, and the boards with the longest memory reward a longer read
• The meetings specifically, which carry the binding signal and the most expensive manual hours, so any change in the cost of reading shows up there first
• Deal cadence, since operators doing more deals per year amortize a better reading method across more transactions and cross the line sooner

Confidence Level

Medium-High. The record sizes are measured directly from the public corpus. The manual hour-and-dollar ranges and the compute figures are estimates from visible inputs, senior billing rates on one side and posted compute rates on the other; the argument holds across the plausible range rather than depending on a single figure.

An aside. The legal pad earned its place honestly. For most of this business’s history it was not the cheap tool, it was the only tool that fit inside a deal budget. The question now is not whether it was ever wrong. It is whether it is still the cheapest read available.

Why This Matters

The decision to read the record deeply or sample it thinly is made implicitly, at the start of every deal, by people who have never seen the baseline written down. Once it is written down, the choice changes character. Reading a few percent of the relevant record stops looking like prudent diligence and starts looking like a bet that the unread 95 percent contains no surprises. That bet was forced when the complete read cost $50,000 in hours. At a marginal cost closer to $25, it is no longer forced, it is a choice, and the firms that notice the difference will price the read, and the risk, more accurately than the ones still reaching for the pad.

Arc: The Reading Problem (Vol. 20 of 19-24). Vol. 19 named the adoption gap as a phenomenon. This issue sizes the manual baseline that made the gap rational: a complete read of the record behind one deal has always cost more in senior hours than a deal budget could carry, which is why the thin sample became the standard. The next issues examine what becomes readable, and askable, once that baseline moves.