You can read every "10 ways to boost conversions" listicle ever written and still have no idea why your page loses people. Generic tips treat every site like it has the same disease. Your site has a specific leak, at a specific stage, for a specific kind of visitor. The job here is to find that exact spot.
This is a repeatable diagnostic process. You define the goal, separate your traffic so you're comparing like with like, walk a decision tree across five stages, score what you find in a table, and then route each problem to the fix that handles it. Run it once and you'll have a ranked list of what to repair first instead of a vague feeling that "the site isn't converting."
It pairs with the website customer experience guide, which covers the strategy. This post is the audit you run when you suspect something is broken and want proof, not a hunch.
Step 1, Define the conversion goal and the path to it
You cannot diagnose a conversion problem until you name the conversion. "More sales" is not a goal you can audit. A goal is a single, observable action on a single page or sequence.
Write down two things:
- The primary action. Purchase, booking, form submission, demo request, account signup, quote request.
- The path to it. The ordered set of pages or steps a visitor moves through to complete that action.
A path is usually three to five stops. Map yours before you look at any data, because the path tells you where to look for the leak.
Two running examples
We'll carry two businesses through every step so you can see the same process produce different diagnoses.
- Imagine a small e-commerce store selling home goods. Primary action: completed checkout. Path: ad or search result -> product page -> cart -> checkout -> confirmation. The owner's complaint: people add to cart and never finish.
- Imagine a two-person consulting firm. Primary action: completed contact form for a discovery call. Path: search result or referral link -> service page -> contact page -> submitted form. The owner's complaint: traffic looks fine, the inbox is quiet.
Same five-step method ahead. By the end, the store's leak and the firm's leak land in completely different places.
Step 2, Segment by traffic source so you compare like with like
A blended conversion rate hides more than it shows. Visitors from a branded search, a cold ad, an email to past customers, and a referral link arrive with wildly different intent. Average them together and a healthy segment can mask a broken one, or a broken segment can drag down a healthy one.
Before you judge any page, break your traffic into at least these buckets:
- Branded / direct (people who already know you)
- Non-brand search (people with a problem, not your name)
- Paid ads (intent depends entirely on the targeting)
- Email / existing audience (warm, already trust you)
- Referral / social (curious, often low intent)
Look at how each segment behaves through your path separately. The question is never "is my conversion rate good." The question is "which source drops off, and at which stage." A page that converts warm email traffic fine but bleeds cold ad traffic isn't a broken page; it's a mismatch between the ad's promise and the page's content, which is an intent-match problem you'll meet in Step 3.
For the store, segmenting shows cart drop-off is roughly even across sources, which points at the cart or checkout itself, not the traffic. For the consulting firm, segmenting reveals non-brand search visitors leave the service page quickly while referral visitors convert fine, which points the finger at the service page for cold visitors specifically.
Step 3, Isolate the leaking stage with the decision tree
Now walk each path stage through five questions, in order. The first one that returns a clear "no" is very likely your leak. Stop there; that's the stage to fix. The five stages are clarity, trust, friction, intent match, and follow-through.
The decision tree
1. Clarity. Can a visitor tell, within seconds, what this is, who it's for, and what to do next? Signals: high bounce on a landing or service page, very short time on page before exit, low scroll depth, people leaving without clicking anything. If no -> clarity leak. If yes, continue.
2. Trust. Has the page given enough reason to believe you before it asks for money or contact details? Signals: visitors reach a pricing, cart, or contact page and exit there; they revisit multiple times without acting; they bounce right after the first ask for personal or payment information. If no -> trust leak. If yes, continue.
3. Friction. Is the action itself easy, or does it demand too much effort, too many fields, or an unexpected step? Signals: high form-start but low form-complete rate, cart-to-checkout drop, abandonment on a specific field or step, mobile drop-off far worse than desktop. If no -> friction leak. If yes, continue.
4. Intent match. Does the page deliver what the source promised, to the kind of visitor that source sends? Signals: one traffic segment bounces while others convert, mismatch between ad/search wording and the page headline, visitors arriving for X landing on a page about Y. If no -> intent-match leak. If yes, continue.
5. Follow-through. After the visitor acts (or nearly acts), does anything stall completion or recovery? Signals: started-but-abandoned actions with no recovery (no saved cart, no follow-up), confirmation steps that fail or confuse, no path back for someone who left mid-process. If no -> follow-through leak.
Use whatever analytics you already have to read the signals above; you do not need to fabricate a "good" threshold for any of them. Compare stages and segments against each other and against your own history. Treat an outside benchmark as context only after checking its source, sample, date, and whether its funnel definition matches yours.
Running the tree on both examples
The store. Clarity on the product page is fine (people add to cart, so they understood the offer). Trust looks fine until checkout. Friction: form-start to form-complete drops hard, and the drop concentrates at the point where shipping cost first appears. Diagnosis: a friction leak at checkout, specifically an unexpected-cost surprise. The tree stopped at question 3.
The consulting firm. Clarity on the service page for cold, non-brand visitors is weak; they leave fast without reaching the contact page at all. Referral visitors, who arrive already warm, sail through. The tree stops at question 1 for the cold segment. Diagnosis: a clarity leak on the service page for people who don't already know the firm.
Same five questions. Two different stopping points, because the data differs.
Step 4, Score it: the audit table that turns symptoms into a ranked fix list
A single diagnosis is useful. A ranked list is what actually changes your week. Score each stage so you fix the biggest, easiest win first instead of the one that annoys you most.
For every stage where the tree flagged a problem, rate three things from 1 to 5:
- Severity, how much this stage costs you (drop-off size, revenue weight)
- Confidence, how sure you are the signal is real, not noise
- Effort to fix, how hard the repair is (5 = easy, 1 = major rebuild)
Then a simple priority score: Severity x Confidence x Effort-ease. Higher is sooner.
| Stage | Symptom (signal) | Severity (1-5) | Confidence (1-5) | Effort-ease (1-5) | Priority |
|---|---|---|---|---|---|
| Clarity | Cold-traffic bounce on service page | 5 | 4 | 4 | 80 |
| Trust | Exit on pricing/contact | 3 | 3 | 3 | 27 |
| Friction | Checkout drop at shipping reveal | 5 | 5 | 4 | 100 |
| Intent match | One segment bounces, others convert | 4 | 4 | 3 | 48 |
| Follow-through | Abandoned cart, no recovery | 3 | 4 | 4 | 48 |
The numbers above are illustrative, not a template answer. Fill in your own. The point is the discipline: the store's checkout-friction row scores highest because the drop is large, the signal is unambiguous, and the fix is contained. The firm's clarity row tops its own list. Each business attacks one thing first, and it's the right thing for that site.
This scored table is the heart of the Website Friction Audit, which gives you the worksheet and the routing in one place so you're not rebuilding the grid from memory each time.
Step 5, Prescribe the fix and the spoke that covers it
A diagnosis only pays off when it points to a specific repair. Here's how each flagged stage routes to the post that handles it. This is what turns the diagnostic into the hub of the whole Website Customer Experience pillar: you land on a stage, you follow the route, you fix it.
| Diagnosed stage | Most likely page to fix | Where to go next |
|---|---|---|
| Clarity | Homepage or service page | Rework the page's first screen and message; see the website customer experience guide for structure |
| Trust | Pricing, contact, or proof sections | Review the website conversion mistakes to check for the trust gaps that cost conversions |
| Friction | Checkout, form, or contact page | Audit the form and steps; check the same website conversion mistakes to check list |
| Intent match | Landing page vs. source promise | Align page wording to the traffic; the pillar guide covers message match |
| Follow-through | Confirmation, recovery, follow-up | If the leak sits after the click, diagnose it from the operations side where process and follow-up live |
A note on the boundary: if your tree stops at follow-through, the leak often isn't the page at all. It's what happens after, abandoned-action recovery, confirmation handling, response time on submitted forms. That's an operations problem more than a layout problem, so diagnose it from the operations side rather than redesigning a page that's working fine.
Back to our two:
- The store routes to the friction fix: remove the checkout surprise (show shipping earlier, or fold it into pricing). The page mostly works; one step needs repair.
- The firm routes to the clarity fix: the service page needs a rebuild so a cold visitor understands the offer in seconds without prior knowledge of the firm.
That difference matters for the next step, because one of these needs a page rebuilt and the other does not.
From diagnosis to rebuild: the Website Friction Audit and building the diagnosed page
You now have a method that produces a named stage, a scored priority list, and a route to the fix. Two ways to take it forward.
Run the full audit on paper. Download the Website Friction Audit to get the decision tree, the scoring grid, and the routing map as a worksheet you can run on any page in about an hour. It's the same process above in a fillable format, built so the output is a ranked fix list rather than a pile of notes.
Rebuild the page the diagnosis pointed to, when that's the fix. This only applies if your diagnosis lands on a stage that requires reworking the page itself, like the consulting firm's clarity leak on its service page. If your diagnosis was the store's checkout-step friction, you're editing one step, not rebuilding; skip this. But when the verdict is "this page can't do its job and needs to be rebuilt," that's the implementation stage.
Resource (sibling site, relationship disclosure): Outclimber and Cedros are sibling sites. Cedros is a website builder. If your audit concluded that a specific page, your homepage or service page, needs to be rebuilt for clarity, Cedros is one way to rebuild that page around the clear message your diagnosis identified. We mention it because of that relationship; use it only if your diagnosis actually calls for a page rebuild, not as a default.
In Plain English
What this is: a five-step diagnostic, define the goal, segment traffic, walk a five-stage decision tree (clarity, trust, friction, intent match, follow-through), score the findings in a table, and route each finding to its fix.
Who it helps: any operator who knows their site underperforms but can't name where it leaks: e-commerce owners with cart drop-off, service businesses with quiet contact forms, anyone tired of generic conversion tips.
When to use it: when you suspect a page is losing people and you want a specific, ranked answer before you spend time or money changing anything. Run it before a redesign, not after.
What to do next: map your conversion path, split your traffic by source, walk the decision tree until you hit the first clear "no," score every flagged stage, and fix the highest-priority one first. If that fix is "rebuild a page," do that; if it's "edit one step," do only that. If the leak sits after the click, move to the operations-side diagnosis instead.
The whole point is to stop guessing. One run of this gives you a shorter, smarter to-do list than any tips article could, because it's built from your own site's signals. That's how you out-climb a competitor who's still randomly tweaking buttons: you fix the one thing that's actually costing you, then move to the next.
Download the Website Friction Audit to run the full diagnostic on your own pages, and if the verdict is a page that needs rebuilding, rebuild it around the message your diagnosis surfaced.