How the Launch Readiness Score works

Forty checks across eight weighted dimensions, each answered No, Partly or Yes and scored zero, one or two. Six of the forty are answered by fetching your live site rather than by asking you. The weights, the reasoning and the exclusion rules are all on this page so you can argue with them.

Why publish the methodology at all?

Because an instrument whose workings are hidden is a quiz. The weights below are judgements, and reasonable people will disagree with several of them. Publishing means the disagreement happens with the numbers in front of you, which is more useful to you and considerably more useful to us.

The scoring, in full

Each check is answered No, Partly or Yes, scored 0, 1 or 2.
A dimension score is the percentage of available points in that dimension. Five checks means ten available points, so three Yes answers and two No answers gives 60.
The overall score is the weighted average of the eight dimension scores, using the weights in the table below.
Where a check is verified against your live site, the verified value replaces whatever you answered. If we could not reach your site, the check reverts to your answer and the result is labelled self-assessed.

The eight dimensions and their weights

Dimension	Weight	Why that weight
Positioning and message	15%	This carries the heaviest weight because no amount of launch day execution rescues a launch that nobody understood. Positioning failures are invisible on the day and fatal over the quarter.
Demand evidence	15%	Weighted equally with positioning. Every other dimension improves a launch. These two decide whether there was anything to launch.
Site and conversion path	13%	Launch traffic is the most expensive and least repeatable traffic you will ever receive. A conversion path that leaks wastes it permanently.
AI visibility	13%	Weighted level with conversion and above traditional search, on purpose. This is the dimension where a 2026 launch is most likely to be silently invisible, and the one almost nobody self-assesses. Only a minority of AI-cited pages also rank in the organic top ten, so this is a separate game with separate rules.
Offer and pricing	12%	Pricing set at launch is expensive to change later, because every change carries a grandfathering decision and a trust cost. Getting the value metric right at launch compounds for years.
Product and activation	12%	Launch traffic converts once. If the product cannot carry a new user to value without a human, the launch produces signups rather than customers.
Search foundations	10%	Search is the compounding channel, not the launch day channel. Four of these five are verified against your live site rather than asked, because they are the checks people are most confident about and most often wrong about.
Launch ops and measurement	10%	The lowest weight, because good ops cannot save a weak product, but the easiest dimension to score full marks on. Failing it is a choice rather than a constraint.

Which six checks are verified, and how

If you supply a URL, our server fetches that page, its robots.txt and its sitemap, once, and answers these six itself. Each one reports what we actually found rather than a verdict on its own.

Check	Dimension	What we look at
Retrieval crawler access	AI visibility	Your `robots.txt`, evaluated per agent for OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, Bingbot and Googlebot
Structured data	AI visibility	JSON-LD blocks on the page, looking for an Organization, Product or comparable type, and whether it carries a stable `@id`
Indexable and readable without JavaScript	Search foundations	Meta robots tag, `X-Robots-Tag` response header, whether your own robots.txt disallows Googlebot, and how much readable text the server sends before any script runs (under about 80 words with scripts present is scored Partly, because most AI retrieval crawlers do not execute JavaScript)
Canonical	Search foundations	Presence of a canonical link, whether it is absolute, and whether its host matches the one we fetched
Heading and title	Search foundations	Count of H1 elements and the character length of the title tag
Sitemap	Search foundations	Any sitemap declared in robots.txt, then the usual locations, checked for valid XML and a URL count

What the verification does not do

It reads one page, not your whole site. It cannot execute JavaScript, so a site that renders its content entirely client-side may return less than it should, and we say so rather than marking you down. It does not judge content quality, and it never marks a check as failed because a fetch failed. An admitted unknown costs less credibility than a confident false negative.

What we do with your URL

We fetch it once, cache what we found against the hostname for 24 hours, and store the hostname with your result so the result page can say which site was checked. The full URL is not retained. No credentials are ever sent and nothing is submitted to any form on your site.

Three checks that used to be here and are not

Earlier drafts scored internal link depth, unique meta descriptions across the site, and whether a category page exists targeting how buyers search. All three are real and all three need a multi-page crawl to answer properly, which is a different and slower kind of tool. They appear as guidance inside the full report instead of as scored checks. Saying so here seemed better than letting you wonder why a search dimension has no internal linking check in it.

Benchmark exclusion rules

Every completed assessment is stored anonymously and feeds the benchmark. A run is excluded from the published figures if any of these apply, and the count of qualifying runs is what gets published.

Completed in under three minutes. Nobody reads forty checks that fast.
Thirty six or more of the forty answers identical. That is a pattern, not an assessment.
A second run from the same source on the same day. The first is kept.
Flagged as a test or staff run.

Runs reporting "already launched" are reported separately rather than mixed into the pre-launch figures, because the two populations are not comparable. Every published figure states its n.

The limitations, stated plainly

Thirty four of the forty checks are self-reported, and self-assessment is generous. People assessing their own work score higher than an outsider would, consistently, which is exactly why six checks were moved to verification and why the instrument opens by asking you to answer as your harshest advisor would.

The benchmark population is self-selected as well. Founders who seek out a launch readiness assessment are probably more diligent than average. Both biases push the same direction, so the fair reading of any benchmark figure here is that it represents a ceiling rather than an average.

Finally, the framework knows nothing about your market. A score of 62 in a category with no competition is a different situation from a 62 against three funded incumbents, and no instrument that runs in seven minutes is going to tell you which one you are in.

Versioning

The current scoring version is 2. Changing the number, order or weighting of checks creates a new version, and results scored under an older version are shown with a note rather than silently rescored. Comparing across versions would be misleading, so we do not.