The 7 Human-Plane Conversion Health Audit Dimensions

In a US courtroom, a bloodstain that matches the defendant proves nothing on its own. Under Federal Rule of Evidence 901, the party offering it has to authenticate it: show who collected it, when, how it was sealed, who held it at every step between the scene and the lab. Break that chain anywhere, and the match becomes inadmissible. Not wrong. Inadmissible. The finding can be correct and still be worthless, because nobody can walk it back to where it came from.

Conversion health audit dimensions are the named properties a page gets scored against, one score per property, each one a question a person could answer by standing in front of the screen. There are fourteen in the full set: seven for how the page works on the humans reading it, seven for how it works on the AI agents parsing it. This piece is about the human seven, which are the ones your team already argues about without having names for them.

The transfer to your page is direct. A dimension score is a conclusion. If nobody can trace that conclusion back to a specific element that was actually observed on the page, you are being asked to act on an opinion wearing a number's clothes. That is why each finding in an audit has to be pinned to annotated evidence on the screenshot rather than asserted in prose. The score is the claim. The evidence is the chain of custody. A skeptical growth lead should be able to check the second before trusting the first, which is the entire reason this article exists.

What does a conversion audit measure, and why does the breakdown matter?

A conversion audit measures the specific, checkable properties of a page that determine whether a visitor understands it, trusts it, and acts on it. The breakdown matters because a single composite number is unfalsifiable. If a tool tells you your signup page scores 61, you cannot argue with it, you cannot verify it, and you cannot act on it, which means in practice you will ignore it by the second week.

Most tools in this space fail in one of two directions. Analytics and session tools give you raw observation with no interpretation: a funnel chart showing 67% of visitors leave before submitting the form, a heatmap showing nobody scrolls past the fold. That is evidence without a finding. It tells you where, and where is not why. On the other side, generic "site health" scores give you a finding with no evidence. A number arrives, its provenance is a black box, and when you bring it to engineering they ask what specifically is wrong and the chain breaks in the meeting.

The 7 human-plane dimensions exist to sit between those failures. Each one is a question a person could answer about a page by looking at it. Each score is attached to findings, each finding to an element. You should be able to disagree with any individual score and point at the thing that produced it while you do.

The 7 human plane audit dimensions

Each dimension is scored 0 to 10. Eight to ten is strong, five to seven needs work, zero to four is critical. A plane's composite is the sum of its seven dimension scores multiplied by 100/70, which puts it on a 0 to 100 scale. That arithmetic is worth doing by hand once. If a page scores 7 across the board, the composite is 70, and you will notice immediately that "70" feels more reassuring than "nothing on this page is strong." That gap between how a composite feels and what it contains is the reason the breakdown is the product and the composite is the summary.

Cognitive Clarity: can the brain process this without strain?

This measures how much work the page makes a visitor do before they can understand it. Dense paragraphs, unlabeled comparison tables, jargon that assumes category knowledge, and too many parallel options all cost processing effort the visitor did not agree to spend.

A typical finding: a pricing page presents six plans in a row with no recommended default and no "most teams pick this" marker. The comparison work has been handed to the visitor. The evidence is the plan grid itself, annotated, with a count.

Decision Clarity: is the next action obvious?

This measures whether a visitor arriving cold can tell what they are supposed to do. It is distinct from Cognitive Clarity. A page can be perfectly easy to read and still leave you unsure which button is for you.

A typical finding: "Start free trial" and "Book a demo" sit side by side at identical visual weight, with nothing indicating which is meant for a first-time self-serve visitor and which is meant for a buying committee. Both are reasonable. Neither is signposted. The evidence is the button pair with their computed styles.

Trust Signal: does this feel legitimate and safe?

This measures whether the page gives a visitor reason to believe the company is real, the product works, and their data will not be misused. Trust signals include customer logos, named case studies, security and compliance marks, review platform badges, and visible company identity.

This is where a worked example from a Nudgent audit is instructive. On one homepage-to-signup pair, the audit found twelve enterprise logos on the homepage and zero on the signup page. That is not "the page needs more social proof." That is a specific placement failure: the proof exists, the company has earned it, and it was left behind at the exact step where the visitor is deciding whether to hand over a work email. A finding you can act on in an afternoon. A general note about social proof is a finding you can nod at for six months.

Motivation Strength: is the value clear before the ask?

This measures the ordering of value and cost. Every page asks for something, attention, an email, a card. Motivation Strength asks whether the visitor has been given a reason before the request lands.

A typical finding: the signup form is above the fold and the explanation of what the product does is below it. The visitor is asked to pay before being told what they are buying. The evidence is the vertical ordering of form and value proposition, with pixel positions.

Comfort Level: are anxieties and risks addressed?

This measures whether the page names and defuses the specific fears attached to the action it is requesting. Getting charged unexpectedly. Getting stuck in a contract. Getting a sales call an hour later. Having data leak.

A typical finding: a paid signup CTA with no "no credit card required," no trial length, and no cancellation language anywhere within the viewport that contains the button. The visitor's unanswered question is not whether the product is good. It is what happens to them if it is not.

Flow Coherence: do the steps connect logically?

This measures whether a multi-step sequence makes sense as a sequence. It applies to signup flows, onboarding, and checkout, anywhere a visitor is moved through states.

A typical finding: step 1 collects an email and promises workspace setup. Step 2 asks for company size and industry, which have nothing to do with setting up a workspace and everything to do with routing a sales lead. The visitor notices. The evidence is the two steps side by side with their stated purposes.

Identity Match: can the visitor see themselves in this?

This measures whether the page's language, examples, and imagery match the person who is actually landing on it. Mismatch here is invisible to the team that wrote the page, because the team already knows who they meant.

A typical finding: copy written for enterprise procurement, "trusted by Fortune 500 security teams," "SOC 2 Type II across all regions," on a page whose traffic is dominated by five-person teams arriving from a product-led content post. Nothing on the page is false. None of it is addressed to whoever is reading it.

How cognitive clarity, decision clarity, trust signal, and the rest become ranked fixes

Dimension scores are the diagnosis, not the work order. Two things happen between a score and a fix.

First, every low score decomposes into findings, and each finding is pinned to an annotated region of the screenshot. This is the chain of custody step. You can look at the score, click the finding, and see the element on the page that produced it. If you think the finding is wrong, you are now arguing about something visible instead of about a number.

Second, findings are ranked by estimated impact rather than by dimension order or by severity alone. This matters more than it sounds. A 3/10 on Trust Signal on a high-traffic signup page almost always outranks a 6/10 on Flow Coherence buried in a settings screen that 2% of users ever reach. The dimension scores tell you what is weak. The ranking tells you what is expensive. Teams that work down the dimension list in order end up fixing the cheapest problems first, because the cheapest problems are usually the easiest to see.

I will state a claim I could be wrong about: I think impact ranking is the harder half of this and the half where our method is weakest. Scoring a dimension against observable page properties is tractable. Estimating what a fix is worth requires assumptions about traffic, intent mix, and the counterfactual, and those assumptions are ours, not measurements. We surface the reasoning behind a rank so you can overrule it with what you know about your own funnel. Anyone telling you their impact estimate is precise is selling you the second half of a chain they have not authenticated.

The audit methodology page walks the full scoring procedure, and the glossary covers the individual terms if you want definitions rather than narrative.

What to do next

Run an audit on the one page where a percentage point matters most, usually signup or pricing. Then read the per-dimension breakdown next to this piece and check three things before you touch anything: does each low score have findings attached, does each finding point at something you can see in the screenshot, and does the ranking match what you already believe about your traffic. Where it does not match, that disagreement is the useful part. It means either the audit found something you did not know or you know something the audit could not see, and both of those are worth ten minutes.

For the underlying concept of what these dimensions are collectively detecting, conversion friction is the shorter definition. The rest of the blog goes deeper on individual dimensions.

Now do the harder thing. Take the last conversion claim anyone made out loud in one of your meetings. "Users don't trust the pricing page." "The form is too long." "Nobody understands the value prop." Follow it backward. Who observed what, on which page, on what date? Somewhere in that chain you will hit a step where the answer is that someone felt it was true, and everything downstream of that step, including whatever got built, is inadmissible.