Work

A worked example: we audited our own site

Listing the sections of an audit tells you nothing about how good it will be. So here is a real one, run against branddirr.com in September 2026, with the findings we acted on. Same structure you would receive: what we saw, the evidence, why it matters commercially, what we changed, and how we would know whether it worked. Nothing here is anonymised, because it is our own site.

This is our own site, not a client result. It shows how we reason and what a finding looks like written up. It is not evidence of revenue impact for anyone else, and the checks below are how we would measure it, not outcomes we are claiming.

1. Finding 01

   Our homepage headline explained our philosophy before it said what we sell.

   Evidence

   The H1 read "We diagnose what's holding your brand back - and do whatever it takes to fix it." Nowhere above the fold did the page name a service category or an audience.

   Why it matters

   A visitor who arrived looking for a website build or ad management had to read several sections before recognising themselves. Every extra inference is a chance to leave. This is the single most expensive thing to get wrong, because it filters everyone.

   What we changed

   Lead with audience and outcome: "Websites and marketing that help service businesses win more customers." Keep the philosophy line for the brand, but move it below the point where the buyer has decided this page is for them.

   How we would check it worked

   Ask five people in the target market to explain what we sell after eight seconds on the page. Then watch scroll depth and the rate at which homepage visitors reach a service page.

2. Finding 02

   One of our own prices bought something narrower than its label suggested.

   Evidence

   The pricing page listed "AI Automation from $800/month". That $800 is the Supervised Content Operations tier. An actual agent build starts at $4,000 per project. The page showed the cheapest tier's price under the service's name and dropped the tier name entirely.

   Why it matters

   The number was not wrong, but it set an expectation the scope could not meet. A buyer who anchors on $800 and then hears $4,000 does not conclude they misread - they conclude they were misled, and that costs trust at exactly the moment it matters.

   What we changed

   Print the tier name next to every price, and separate one-time projects from monthly retainers so the two are never compared as if equivalent.

   How we would check it worked

   Count how many sales conversations open by disputing or re-anchoring the price. If the label is doing its job, that objection should mostly disappear.

3. Finding 03

   Our service pages argued from a false dichotomy.

   Evidence

   The SEO page paired "Zero organic traffic" with the cost "100% reliance on paid; CAC keeps rising." Both halves are unproven: a business can acquire customers through referrals, partnerships, or outbound, and no organic traffic does not establish that costs are climbing.

   Why it matters

   A sophisticated buyer notices an overclaim immediately, and it makes every other number on the page suspect. Overstating the problem to sell the fix is the fastest way to lose the buyers most able to pay.

   What we changed

   Describe it as a risk worth investigating rather than a proven cost: acquisition leaning on fewer channels than it could. Then find out what the actual mix is before asserting anything about it.

   How we would check it worked

   Track whether discovery calls now start from the client's real acquisition mix instead of correcting a claim we made.

4. Finding 04

   Our navigation menu stayed open on top of the page you had just asked for.

   Evidence

   The mega-menu panel spans the full header width. Clicking "Pricing" navigated correctly, but because the pointer was still inside the panel's container, no mouseleave fired and nothing closed the menu. The destination page sat behind the overlay until you clicked elsewhere.

   Why it matters

   This one is not a copy problem, it is friction on the exact click that signals buying intent. Someone comparing prices had to work out that the page had loaded underneath.

   What we changed

   Close the panel when the route changes, and on the trigger's own click for same-route cases. Keep the existing Escape and focus-out handling.

   How we would check it worked

   A test that clicks a nav item and asserts the destination heading is visible and unobstructed, so the regression cannot come back quietly.