How to organize a website

The symptom

Audit, Tax, Advisory. Or Litigation, Corporate, Real Estate. Or Wealth Management, Retirement, Institutional.

Somebody arrives carrying a specific problem. A founder buyout going sideways. A lender asking for something the last controller never produced. The menu bar hands them a directory of departments.

So their first job is guessing which department owns the problem. Tax or Advisory? Corporate or Litigation? They pick one, read four paragraphs of capability language, find no version of their own situation in any of it, and back out to try the other tab.

That visit lands in the analytics as a bounce, indistinguishable from traffic that was never going to buy. It’s the firm’s org chart with a menu bar on it, and the person who needed the firm most got handed the decoding job.

Why it keeps happening

Whoever funds the site gets to shape it, and in most professional-services firms budget authority runs along practice lines.

Each group pays for part of the build. Each wants its own section, its own landing page, its own control over the words describing what it does. None of that is unreasonable on its own. The head of Tax has a real interest in how Tax gets represented, and defending it is the job. The site ends up assembled from a dozen locally sensible decisions and collectively useless, because a buyer arrives with a situation, and situations ignore practice-area boundaries.

We’ve shipped the org-chart nav. More than once, with a straight face and a launch email. Every group signed off, which was the problem. Sign-off from eleven practice heads measures internal peace and says nothing about whether a stranger can find their own problem on the site. We were collecting the wrong approval and calling it consensus.

Many established firms stumble right here. They carry real experience, then wrap it in language so broad that the buyer can’t locate the relevant piece inside the specific situation. The experience is genuine. The architecture buried it.

What the buyer is doing at the Relevance gate

Scanning for whether the subject belongs to their world, and doing it in seconds, well before quality enters the picture.

A menu of service names gives them nothing to match against. They’re holding a situation phrased in the language of their own business, and the site answers in the language of the firm’s reporting structure. That gap is narrow enough that nobody inside your firm notices it and wide enough that most visitors won’t cross it. They leave, land somewhere that felt more certain, and never appear in your pipeline as a loss.

The machine layer sits on the same fault. A sorting system reads the navigation as the firm’s own account of what it handles, and internal service names match nothing about how a buyer phrases a problem. A firm with genuine depth in seven situations gets read as a generalist, because generalist is what the architecture says out loud. The point holds for a human skimming the menu, for the same reason.

The market can’t value what it can’t see.

The correction

Rebuild the information architecture around named client challenges and named audiences, with services demoted to a layer underneath.

Start with a list. Eight to twelve situations the firm is genuinely best at handling, written in the words a buyer would use on a first phone call rather than the words that appear on an engagement letter. “Transaction Advisory” is the engagement letter. “You’re selling in eighteen months and the books won’t survive diligence” is the phone call. Those situations become the primary navigation.

Service pages survive. Move them one level down, where they serve the visitor who already knows what they want to buy and needs confirmation that you sell it. Nobody loses a page. What changes is which layer a stranger meets first.

Then file everything else underneath. Every article, every case note, every piece of published thinking gets filed under a situation rather than a practice group. Disconnected content proves you’re publishing. A coherent body of work creates recognition. Pages shouldn’t be treated as isolated compositions, and a library sorted by department is a pile of isolated compositions with nothing holding them together.

The site should read as one controlled argument. Somebody who lands deep inside an article about the situation they’re living through should be one click from the rest of that argument, without passing through a department. The homepage does the same job in compressed form, which piece 04 covers.

Test the list before anyone builds anything. Read the situations aloud to somebody outside professional services. If they can name a company that fits one, you’ve written buyer language.

Before and after

Before. Services: Audit · Tax · Advisory · Consulting · Private Client

After. We work with: Companies preparing for sale · Family businesses handing over to the second generation · Firms that outgrew their finance function this year · Nonprofits facing a first federal single audit

Both are menus. The first asks visitors to classify themselves before reading anything. The second tells them, before a single click, that somebody here has met their situation before.

The hard call

This takes the menu bar away from the practice heads, which is why most firms don’t do it.

Say it out loud in the first meeting rather than presenting the rebuild as a neutral technical exercise. They’ll work it out by the second anyway, and letting them discover it turns a design decision into a political ambush. What you can offer in exchange is real. Each practice keeps its service page, gets named as owner of the situations it drives on the pages where buyers are reading, and gets attribution on the work filed underneath.

We’ve watched this argument run in a lot of rooms, and it lands the same way each time. A firm that won’t spend the political capital is choosing internal peace over the visitor, which is a legitimate choice...it just shouldn’t be filed under strategy.

Run this on your own material

Write down the last five problems real prospects arrived with, in the words they used on the phone. Their words, before anyone translated them into a service line.

Then open the site and try to find where each one lives, in under two clicks, on the clock. Hand the test to somebody who doesn’t work at the firm, because you know where everything is and that knowledge disqualifies you as a subject.

Score each attempt: found it, found something adjacent, gave up. Three “gave up” results out of five means the problem lives in your architecture. Rewriting service-page paragraphs won’t touch it, and most firms spend a year trying anyway.

Signals you’ve cleared it

Advance when inquiries arrive already describing a situation your site names, sometimes in the site’s own phrasing. When a prospect references a specific article rather than the firm in general. When somebody says they weren’t sure a firm like yours handled this until they saw the page.

Hold when the navigation still requires a visitor to know your taxonomy before reading anything. Hold when published thinking is still filed by practice group, because readers and sorting systems both follow the filing. Hold when practice heads negotiate their way back to the top level one exception at a time. One group is genuinely different and needs its own tab, then another, and six months later the menu is the org chart again.

Where this connects

Engagement Success in an AI World · Proximity Is No Longer Measured in Miles, on becoming the firm closest to the problem.

Research Library · Entry 15, AI-mediated discovery.

Next move · Go to How to write a call to action.