Free worksheet · 4 minutes · no email, no tracking

The Ceiling Test

Before you build the site, compute what the site is worth at its best. Twelve lines of arithmetic. If the answer is small, the four hours you were about to spend are a donation.

The formula

Why a ceiling and not a forecast: a forecast assumes you execute well, get indexed, keep the domain, and still care in month nine. A ceiling assumes nothing. If the ceiling is $9/month, no amount of execution gets you to rent.

The calculator

Everything here runs in your browser. Nothing is sent anywhere, nothing is stored.

Three worked examples, with my own numbers

From a portfolio of ten live sites. The red row is the one that cost me the most.

The first two were "quick" builds. Both shipped. Neither pays for the domain. The third was a worse idea on paper and is the only one of the three whose curve points up. Gate 0 does not grade your taste; it grades the arithmetic.

Four numbers that will lie to you

1. Ad RPM is a gross number, and networks pay less than they advertise

Take the published RPM and halve it before it enters the formula. If your whole idea depends on the difference between $4 and $2 RPM, the idea depends on a rounding error.

2. Sessions are not clicks

On one of my sites GA4 recorded roughly one session per four Search Console clicks — consent mode, blocked scripts, the usual. Every "revenue per 1,000 sessions" metric derived from it was inflated 4×, which made a dead site look nearly profitable. Measure per 1,000 clicks.

3. The price you paid is not the price you will pay

¥12 for the first year, ¥93 at renewal. 7.75×. Put the list renewal price in the cost box, not the promo you grabbed. Same trap on the AI subscription that "just" helps you write pages.

4. A balance is not revenue

$26.96 sitting under a $100 payout threshold is not money. It is a hostage: it converts to zero the week you stop, and it makes your "lifetime earnings" look 20× better than your monthly rate. Count money when it arrives.

The six lines you must be able to write before you build

query universe   : ____ clicks/mo (source + date)
realistic share  : ____ % (page-1-for-a-new-domain is 5-10%)
monetisable rate : $____ per 1k clicks (measured, or published ÷ 2)
ceiling          : $____/mo   → under $30 = do not build
cost to stay alive : $____/yr (renewal at list price + hosting + shared subs ÷ n)
kill date + gate : if <____ clicks/mo AND <$____/mo by YYYY-MM-DD → stop, no extension

The gate, in the shape you can paste into a repo

{
  "domain": "example.com",
  "renewal_est_usd": 13,            // list price, not the promo
  "kill_date": "2027-02-28",
  "hours_budget_30d": 10,
  "measured_revenue_usd_month": 0, // 0 is a value, not a blank
  "gates": [
    { "label": "search clicks ≥ 300/mo", "metric": "search_clicks_month", "op": ">=", "threshold": 300 },
    { "label": "measured revenue ≥ $10/mo", "metric": "measured_revenue_usd_month", "op": ">=", "threshold": 10 }
  ],
  "note": "why this site exists, and what would prove it should stop"
}

Two rules make this file worth more than the code you are about to write. Write it while you are still excited — in month five you will argue with it. And 0 is a value: a blank field gets read as "not measured yet", which is how a site survives eleven consecutive months of nothing.

If the ceiling is fine, build it — but in four hours

The ceiling test kills bad ideas. It does not tell you how to execute a good one, and a good idea executed for three weeks is still a loss. The rest of the method I use lives in a small kit: a runbook that ends at "live + pushed to Bing" in four hours, an AGENTS.md contract that stops your AI agent from rewriting titles while your ranking is falling, seven zero-dependency check scripts that fail the build on duplicated pages and dishonest schema, and a ledger that recomputes this arithmetic monthly and names the site you are wasting hours on.