Audits

An audit is PlumPMM’s core measurement primitive. Given a public URL, it scrapes the page, runs an LLM analysis across 13 product-positioning dimensions, and returns scored answers plus concrete fixes.

Audits show up everywhere: ad-hoc URL audits on the home page, scheduled monitors, competitor benchmarks, draft-preview audits (treating draft copy as if it were live), and as the input to Audit Compare (narrative deltas between two runs).

What an audit measures

13 dimensions, each rated 0–10. The overall rating is the rounded average.

Dimension	Question it answers	What scores well
target_customer	Who is this product for?	Names a specific role, industry, or company stage in plain language.
problem_solved	What concrete problem does it solve?	A specific pain — not “improve workflows” or “boost productivity.”
aha_moment	Can a cold visitor grasp the value within a couple of scrolls?	H1 communicates a specific outcome; subheading adds new information; concrete benefit appears in the first visible sections.
product_description	What does the product actually do?	A plain-language explanation, not a feature list.
business_benefits	What outcomes does it deliver?	Backed by named customer quotes, testimonials, or case studies with actual numbers. Generic claims like “saves you time” don’t count.
objections	Does the page address common buyer objections?	Pricing transparency, security concerns, switching cost, integration questions surfaced proactively.
headline_appeal	Does the H1 itself communicate value?	The literal H1 names a specific outcome or pain relief — not a tagline or a wordplay.
unique_value_proposition	What’s the one specific claim only this product can make?	”Only platform that does X for Y.” “We are the best” doesn’t count.
market_category	Can a visitor answer “what kind of product is this?” within 3 seconds?	Yes, and the page also produces a clean 2–5 word inferred category (e.g., “developer onboarding tool”, “B2B CRM”).
cta_alignment	Does the primary CTA match the buyer’s stage?	Cold visitors get a low-friction option. “Book a Demo” with no trial scores poorly for cold traffic.
evidence_quality	Does customer evidence directly validate the positioning?	Direct customer quotes tied to the specific problem or outcome. Logos and badges alone don’t count.
icp_alignment	Does every key element speak to the deduced ICP?	Headline, subheading, product description, and CTA all explicitly address the inferred ICP’s role and outcomes.
competitive_differentiation	Why choose this over alternatives?	Comparison tables, “unlike X” framing, named alternatives with specific differentiators.

Scoring rubric

The same 0–10 rubric applies to every dimension:

9–10 — Answers the question clearly and directly in plain language. No jargon. A 10-year-old could understand it.
7–8 — Answers the question but is somewhat vague or uses mild jargon.
5–6 — Partially answers, or hides the answer behind fluffy marketing speak.
3–4 — Barely addresses the question. You have to squint to find an answer.
0–2 — Does not address the question at all.

A dimension that can’t be answered from the page scores 0–2, not 5. The rubric is honest by design — soft scores produce soft fixes.

What you get back

For each dimension, the audit returns:

Answer — what the page actually said, with quoted text where possible
Rating — 0–10
Explanation — 1–3 short sentences for why it scored that way
Suggested fixes — 2–4 concrete, actionable improvements that would raise the rating

Plus an overall summary (2–4 sentences) and the rounded average rating.

The icp_alignment and market_category dimensions return one extra field each — the inferred ICP and the inferred market category the page communicates, separate from how clearly it does so. Useful when your stated ICP and the page’s implied ICP diverge.

Where audits run from

Surface	What triggers it	Cost
Home page audit form	Paste a URL → click Audit	1 audit-unit
Monitor schedule	Backend tick when `next_run_at` ≤ now	1 audit-unit per run
Competitor audit	Click “Run audit” on a competitor record in Research	1 audit-unit
Draft preview audit	Click “Generate preview audit” inside a draft	1 audit-unit
Audit Compare (narrative)	Click “Compare to previous” or “Compare to live”	1 audit-unit
Fasttrak	Click “Start Fasttrak”	10 audit-units (covers the whole orchestration including 2 audits)

→ Audit-unit budgeting: Plans and pricing

How long an audit takes

Roughly 20–40 seconds end-to-end, dominated by the scrape + LLM call. Pages with heavy JavaScript rendering or anti-bot measures can fail to scrape — you’ll see an error, not a mock score.

When to re-audit

Audit the same URL again whenever:

You’ve shipped meaningful copy changes — verify the change moved the score
A monitor flagged a regression — the audit run that triggered the alert is the new baseline
You’re building a draft variant — audit the live page first so the variant has a baseline to compare against

Audits that hit the same URL pile up in the History tab on the page workspace. Older runs are kept indefinitely for paid users (audit_history: true); Free users only see the latest.

What audits don’t measure

By design:

SEO, page speed, accessibility, design — out of scope. Plenty of tools cover those.
Conversion rate prediction — PlumPMM doesn’t promise that a 9.0 page converts better than a 7.0 page. It promises the 9.0 page communicates positioning more clearly. The conversion-rate work is what you do with the fixes the audit hands you.
Brand voice consistency across multiple pages — single-page measurement only. For a multi-page audit, run each page independently and compare.