orbcrawl

Essay · 02

The arithmetic that killed our first plan.

We sized a plan before we had measured the funnel. Once measured, the numbers were smaller than the plan assumed, and the plan changed.

Sam, Orbcrawl · · Written from the 5 September 2026 reading, while discovery was still under way; see Evidence for later readings.

The plan as written

In late August, with seven districts crawled, we wrote a plan built on assumptions about volume. Then we extended the crawl across as much of the country as it had reached and did the arithmetic. The plan did not survive it.

Predicted against measured

Seven districts were extrapolated to the country and missed on every line, for the reason the plan named: large urban districts hold more businesses, more of them have a domain, and more of those domains are already fine. The district map shows how uneven the register is.

Schematic, not to scale: three quantities predicted from seven districts, then measured. The prediction missed on every line.

The order of magnitude was the finding. The pipeline refuses about ninety-seven in every hundred records it ingests, as of that reading, and most of the refusal is correct. What survives is small, chiefly because most sites are fine.

The chain, on the plan’s own rates

The plan assumed a rate at each step between a business being found and a business becoming a result. None of those rates has been measured. Applied to the whole country in one pass, they gave a total well short of the target the plan had set itself.

Illustrative, unmeasured rates: the plan's own assumptions, applied to the whole country once and then to what changes each year.

The list refreshes a little each year as sites change, but the yearly top-up is small beside the first pass.

The legal number is smaller still

A list of businesses is not a list of businesses we can approach until it has been filtered by what the law on marketing email allows. We hold back every business we are not prepared to act on.

Illustrative, not to scale: a grid of records, dimmed stage by stage until one remains.

At that standard, the businesses we intend to draw on are a small fraction of the register. We say so because the smallness shapes the plan, and because the temptation in every pitch is to quote the larger number.

What the arithmetic points at

The funnel discards nearly everything, and the shape of the register decided which kinds of business were left: the survivors were a market nobody chose. The plan has to start from that market, not the one we assumed. The ranking already reads more about a business than the state of its site — two scores, deliberately — so it can be pointed at a different kind of business.

Two more things follow, neither built. Every audit is dated, so the record can one day be read for change. And the sites refused because they were fine are not waste: a fine homepage says nothing about whether a business can take a booking. That is a different question, and we have not tried to answer it.

Why publish arithmetic that argues against ourselves

Because the alternative is to keep quoting the larger numbers and let the smaller ones arrive later anyway.

This essay gets a second figure when those rates are measured. Until then the dated readings show how far the crawl has got, and the Method page lists its known limits.