We audited our own website. 14 pages. Zero passed.
Five weeks ago, we ran our own website through an audit we use on client sites. The Refinery23 “Friction and Filter” method — score each page as a busy, overwhelmed business owner scanning for 60 seconds. Quick read. No benefit of the doubt. Either the page earns attention in the first scroll or it fails.
Fourteen pages. Zero passed outright.
We are a marketing agency. We have been doing this for years. Our clients pay us to fix exactly this problem on their sites. We found it all over ours.
What the audit actually looks for
The “Friction and Filter” method has two axes. Friction: anything that slows a reader down or creates confusion — competing CTAs, jargon, unclear structure. Filter: whether the page quickly tells the right person they’re in the right place, and tells the wrong person to leave.
Most agency websites fail both. They’re written to impress peers, not to convert buyers. They hide what they actually do behind positioning language. They assume the reader will work to understand the offer.
Our homepage had three competing calls to action. “START.” “Get Started.” “START YOUR ENGINE.” Three separate buttons, no visual hierarchy, no indication which one mattered most. If you’re a buyer who arrived with intent, you’d have to decide which of three things to click before you’d even finished reading the headline.
That’s not a design problem. That’s a thinking problem. Someone — us — decided all three of those CTAs belonged on the page at the same time.
The second issue was jargon density. Our headlines were sharp. Below the fold, the body copy was written for other marketers. Terms our buyers don’t use. Frameworks without explanation. The kind of writing that signals expertise to someone already inside the industry and signals confusion to the CFO who’s actually signing off on the brief.
We knew this, abstractly. The audit made it specific.
What we fixed first — and why
After the April 15 audit, we had a long list of failures. We didn’t try to fix everything. We prioritized the three changes with the highest signal-to-effort ratio.
First: one CTA, everywhere.
We picked “Let’s Talk Strategy” and applied it consistently across all 11 active pages. No exceptions. This was entirely a content decision — no development work, no design changes, no new pages. We opened the CMS and rewrote buttons.
It took a few hours. It was the single biggest structural problem on the site, and fixing it cost nothing except the willingness to admit it needed fixing.
Second: every hero had to lead with a specific stat or a sharp problem statement.
We had pages where the main headline was a positioning claim. “The growth agency for serious brands.” That sounds confident. It tells you nothing. A buyer who arrives with a specific problem — “I’m spending too much on leads and not converting enough” — has no reason to keep reading.
We rewrote every hero. B2B: “1,000 Leads. 4 Closed Deals. That’s a targeting problem.” Email marketing for ecommerce: “Recover the 70% of customers who abandon their cart — on autopilot.” Insurance lead gen: “34% Lower Cost Per Sale. In 12 Months.” OTA direct bookings: “15% of Your Revenue. Gone to the middleman.”
Every headline is now either a number that means something or a problem the buyer recognizes. Neither requires the reader to do translation work.
Third: add two real testimonials with named outcomes.
Generic testimonials don’t convert. “Great agency, highly recommend” tells a skeptical buyer nothing. We had almost no social proof on the site with specific numbers attached.
We added a real testimonial to the B2B lead gen page — 200% growth in webinar attendance, named client category, named metric. We added a CFO testimonial to the paid media ecommerce page. Two pages went from zero credible social proof to something a buyer can evaluate.
What five weeks of changes actually produced
We ran the same audit on April 20. Same method, same scoring, same 60-second-scan discipline. Here’s where 11 pages landed:
| Page | April 15 | April 20 |
|---|---|---|
| Homepage | FAIL | PARTIAL PASS |
| SEO & GEO Lead Gen | FAIL | PARTIAL PASS |
| SEO & GEO Ecommerce | FAIL | PARTIAL PASS |
| RTFM (Book) | FAIL | PARTIAL PASS |
| Paid Media Lead Gen | PARTIAL PASS | PARTIAL PASS |
| Paid Media Ecommerce | FAIL | PASS |
| Insurance Lead Gen | FAIL | PARTIAL PASS |
| Email Marketing Ecommerce | FAIL | PARTIAL PASS |
| Email Marketing Lead Gen | FAIL | PARTIAL PASS |
| B2B Lead Gen | 6.5/10 | PASS |
| OTA Direct Bookings | 6.5/10 | PASS |
Three pages now pass. Eight more are partial passes. Nothing is an outright fail.
That’s real movement. It’s not finished. The honest read is that we’ve addressed the structural problems — CTA confusion, weak heroes, missing social proof — and we’re still mid-work on everything else.
The things we caught that stung a little
The Paid Media Lead Gen page had a stat we added — “312 Sales Last Quarter” — and a typo in the next line: “Saless.” That page had been live. We found it during the re-audit.
The Email Marketing Ecommerce page now leads with “70% abandoned cart” in the hero. But the social proof counter beneath it shows R0K — a broken dynamic counter that hasn’t loaded correctly since we updated the page. The headline works. The page still looks unfinished if you scroll far enough.
The SEO Ecommerce page had the wrong FAQ section on it. Someone had copy-pasted questions from the email marketing page during an update. Email platform questions appearing on an SEO page. Those have been fixed now, but they were live for long enough.
We’re an agency that runs conversion audits for clients.za/conversion-rate-that-disappeared/" class="linkosaurus-link" data-linkosaurus-source="2025" data-linkosaurus-target="2033" data-linkosaurus-score="0.495">conversion audits for clients. Every one of those was a live error on our own site.
What’s still open
Social proof is the remaining gap. Two pages now have real testimonials. Nine don’t. We know the testimonials we want — specific client outcomes, named metrics — and we’re producing the case studies that will feed them.
The homepage still has a social proof problem. CTAs are sorted. Hero is sharper. But a buyer who reaches the bottom of the page looking for evidence we deliver what we say we deliver won’t find enough of it yet. That’s a known gap. We’re working on it.
The RTFM book page still has navigation that competes with the primary CTA — we stripped the full service menu but haven’t simplified the header completely. That’s on the list.
We’re not going to wait until everything is finished to say what we learned. The point of doing this publicly isn’t to show a perfect outcome. It’s to show the process.
The actual lesson
When we run this audit on client sites, we find the same two problems almost every time: CTA confusion and hero copy written for the wrong reader. We know what they look like. We know how to fix them.
We had both problems on our own site. Badly.
There’s a specific kind of embarrassment that comes with finding your own failures in your own methodology. We built the Friction and Filter audit to catch exactly this. We used it on ourselves and caught exactly this.
The reason it happens — on our site and on client sites — is that the people writing the copy know too much. They know what the product does. They know the terminology. They know how the CTA is supposed to work. The buyer doesn’t know any of that, and the people writing the copy stop being able to see what the buyer sees.
An outside audit catches it. Your customers see it immediately. You see it only when you force yourself to read the page as someone who has never heard of you.
Do that first. Then fix the CTAs. Then fix the heroes. The social proof comes after, because you need to know the page is doing the right job before you add evidence that it delivers.
The three changes that cost nothing: one CTA, one stat-driven headline, two real testimonials. We shipped them all in under two weeks. The site isn’t finished. It’s materially better, and it cost us almost nothing to get there.
