One document a buyer or a founder can read in ten minutes. Built from what anyone can see from outside: the website, the cases, review platforms, LinkedIn, search results, AI-assistant answers, DNS records. Whatever cannot be seen becomes a question, never a guess.
Two uses:
- Before a first call with a prospect: the audit is the call plan.
- Self-audit by an agency founder (course homework): the founder scores their own company the way a stranger would, then picks the three things to close first.
Language of the document = language of the user. Company, tool and competitor names stay Latin.
Hard rules
- No evidence, no score. A criterion that cannot be seen from outside gets
?, is excluded from the block mean, and turns into a call question. Never infer a score from "they probably have it". - Torn between two scores: take the lower one.
- Results first. The page opens with the overall score and a 4–5 line thesis, then the dashboard. Methodology never comes first.
- Every number has a source line (page URL, review count and date, DNS record, post count and period). Anything derived is tagged as a conclusion.
- Confidentiality. Never mention the author's other clients or their cases. Competitors of the audited company may be named: they are public.
- One next step at the end (a call link or the homework hand-in), never two CTAs.
- Ukrainian text: no em dashes (use
?as the "not visible" marker, en dash only between numbers), decimal comma (2,7), plain words, block names in the user's language.
Scale (same for every criterion)
| Score | Meaning |
|---|---|
| 1 | absent, or present but brings no buyers |
| 2 | weak |
| 3 | exists, but it is not a system |
| 4 | strong |
| 5 | best in class, visibly better than the neighbours |
? |
not visible from outside; not scored, becomes a call question |
Block score = mean of its scored criteria, one decimal. Overall = mean of all scored criteria. Card colour: below 2.5 red · 2.5–3.4 amber · 3.5 and above green. Show the formula next to any composite number and say how many criteria were scored ("mean across 20 criteria").
The 6 blocks and 23 criteria
Full checklists, tables and sources are in references/rubric.md. Read it before scoring.
| # | Block | Criteria |
|---|---|---|
| 01 | GTM strategy (foundation) | ICP (3 components) · positioning · offer · delivery process (8-point checklist) · lead magnets · case depth (5 cases × 7 features) |
| 02 | Team (foundation) | GTM roles · owners of announced markets · ? sales hiring and development |
| 03 | Lead generation (pipeline) | SEO · AI visibility · LinkedIn and content · events · outreach infrastructure · inbound capture |
| 04 | Sales (pipeline) | path to the first meeting · buyer materials · ? first call and proposal |
| 05 | Account management (pipeline) | retention · reviews and referrals · ? account expansion |
| 06 | Processes (foundation) | tracking and analytics · forms-to-CRM link · ? metrics and rhythms |
Criteria marked ? are almost never visible from outside. Keep them in the table, score them
only when the user supplies inside evidence (answers, exports, call notes).
Procedure
- Collect (60–90 minutes of agent time for a prospect; 2–3 hours for a founder doing it by hand). Site crawl: menu, service pages, cases, pricing, how-we-work, about, blog. Review platforms: count, dates, client sizes and industries. LinkedIn: company page and the two most visible people, 12 months of posts. Search: brand and category queries in the main market. AI assistants: 2–4 topics × 3–4 unbranded prompts, note who gets named instead. DNS: MX, SPF, DKIM selectors, DMARC policy, separate sending domains.
- Score every criterion with 2–4 evidence bullets and a one-line verdict, using the tables
in
references/rubric.md(ICP, delivery process, case anatomy are mandatory tables). - Write the page in the fixed order below. Then run the self-check list.
- Deliver as a single HTML file (the template in
references/template.htmlif the pack ships it, otherwise the structure below with the user's design tokens or neutral ones).
Page order (fixed)
- Header: company, date, "prepared from public data only", overall score, 4–5 line thesis that names the weakest block and the first thing to fix.
- Dashboard: 6 block cards with scores and criteria pills, plus 3 headline findings with numbers.
- Bridge: why strategy decides the rest (6 block scores as bars, strategy highlighted).
- Blocks 01–06. Each block: H2 = its conclusion with a number, criterion rows (score · evidence · verdict), the mandatory tables where they belong, a "What to do" box, a "Not visible from outside · questions for the call" box with 2–3 questions.
- Next step: one link or one hand-in instruction.
- Sources: every URL and record used.
Self-check before delivering
- Every score has evidence; every
?has a question. - No block H2 is a bare block name; each carries its conclusion.
- Headline numbers appear at most twice (thesis and its block).
- Reviews are counted as reviews, not as clients.
- Every "since", "years", "average" has a dated source or is removed.
- No client names of the author, no em dashes in Ukrainian, decimal comma.
- The page has
<meta name="viewport">and does not scroll sideways on a phone.