How the Pressure Reading is computed.
Every Find The Leak diagnostic is a 0-100 weighted composite across seven GTM dimensions, computed from twenty-plus live public data sources, rendered into a roughly thirty-page PDF. Nothing is hidden behind a proprietary black box. Here is the full mechanism.
We publish this for two reasons. Buyers should not be asked to trust a number whose provenance they cannot inspect. And the firms we sell to, the operators inside CMO and founder seats, want to know whether the score they see in their report holds up under their own scrutiny. We think it does. The math is below.
One number, seven dimensions, transparent weights.
The Pressure Reading is a 0-100 weighted average of seven dimension scores. Each dimension is itself scored 0-100 from a specific public-data measurement. The composite is then converted into one of three bands.
| Weight | Dimension | What it measures |
|---|---|---|
| 20% | Competitive Positioning | Demand trajectory vs. competitors |
| 15% | AI Share of Model | How often AI assistants recommend you vs. competitors |
| 15% | Page Speed | Lighthouse performance score (mobile) |
| 15% | SEO & Indexability | Indexed pages, robots.txt, schema.org coverage, AI crawler access |
| 15% | Review Presence | Aggregate rating and volume across major public review aggregators |
| 10% | Security Headers | HSTS, CSP, X-Frame-Options, certificate health |
| 10% | Tech Stack | Marketing, analytics, attribution, and CDP tools detected |
The bands are interpretive shorthand for the score. Below 50 is past redline. Between 50 and 74 is pressure dropping. 75 and above is holding tight. The bands are not category-tuned; the underlying score is.
What we measure and why.
Every dimension exists because there is independent evidence that it correlates with real pipeline outcome. The weights reflect both the strength of that correlation and the actionability of the signal. A signal you can fix this quarter weighs more than a signal that takes a year.
Predicts whether your category is rising, flat, or falling around you.
Buyers ask AI before they ask Google. This is the new top of funnel.
First-impression latency. Direct conversion impact.
If you're not indexed, you're not in the consideration set.
Social proof gates B2B purchase. Buyers read reviews before they take a call.
Enterprise procurement filters on basic hygiene. Failing this filters you out.
Stack maturity correlates with attribution depth and revenue operations rigor.
How we measure who AI recommends.
For each diagnostic we generate a fixed number of buyer-intent queries grounded in the target's category, then route every query through all major large language models in parallel. For each response we extract the top-five vendor recommendations and run a rank-weighted, word-boundary brand match against the target and every named competitor.
Rank-weighting matters. Being recommended first is qualitatively different from being mentioned fifth. Word boundaries matter too. A naive substring match produces false positives we will not honestly ship. We match against the actual recommended brand, not raw text inside it.
The final share is computed as the brand's rank-weighted mention count divided by the total brand mentions across all queries and models, expressed as a percentage. When fewer models are reachable, the report explicitly labels the result and flags it with a published caveat. We do not pretend a single-model sample is a multi-model consensus.
Public signal only.
Every number in every report is derived from public data. We do not access internal systems, CRMs, analytics platforms, or anything behind a login or paywall. The engine reads across the families below.
| Family | What it covers |
|---|---|
| Performance & technical health | Industry-standard performance APIs, security header probes, SSL/TLS handshakes, DNS inspection. |
| Search visibility & indexability | Search-index probes, sitemap and robots inspection, structured-data extraction, AI-crawler access checks. |
| Messaging & brand history | Long-form archives of public web content, including multi-year snapshots of homepage and key landing pages. |
| Voice of customer | Major public review aggregators, community discussion platforms, video review search, B2B and B2C review surfaces. |
| Press & media | News feed aggregators and tier-1 media coverage indexes used by AI summarisers to ground their answers. |
| Hiring & GTM motion | Public applicant-tracking systems and careers pages, parsed for active roles and motion signals. |
| Competitive ad activity | Public ad-platform transparency archives. |
| Tech-stack fingerprinting | Live DOM inspection, runtime probes for marketing and analytics tools, outbound-request analysis. |
| AI Share of Model | Live endpoints from all major large language models. |
Every individual diagnostic report ships with a methodology appendix that lists the specific sources that informed that customer's numbers. The publishing of the inventory is the report itself, not this page.
What we will and will not claim.
A category-level peer median requires at least ten completed diagnostics in the same cohort before we will surface it. Below that threshold the cohort is too small to be statistically honest, so we suppress the comparison entirely rather than display a noisy number. Above it, the report shows your score against the category median as a second tick on the Pressure Reading bar.
When fewer than five of the seven dimensions could be measured on a given diagnostic run, typically because a provider blocked us or rate-limited us, the Pressure Reading is rendered as a partial measurement. The number is dimmed, the band reads “partial,” and a callout tells the reader exactly which dimensions succeeded. We do not let a two-component score read as authoritative.
The honest caveats.
The Pressure Reading is a snapshot of public signal at the moment we ran the audit. Markets move fast. A number computed in January may not hold in March. We render the audit date prominently on every report cover for this reason.
We do not measure intent we cannot see. Closed-deal velocity, win rate by segment, average contract value, sales-cycle length, churn cohorts, the actual content of your CRM, none of this is in scope. The diagnostic is a public-signal lens on a private business. It is highly useful, but it is not omniscient. The consultation debrief exists specifically to connect what we measured to what only you know.
We do not back-fit. Every weight in the score formula was set before any client report was generated and has not been retroactively tuned to flatter individual customers. If the math says 47, the report says 47.
See the math against your own domain.
The fastest way to evaluate the methodology is to put your own company through it. Submit your domain and three competitors. We respond within one business day to schedule the debrief.
Request a diagnostic →