Your Org Chart Is Not a Sitemap

Watch a real visitor try to find a program page on a typical university website. Watch them pause at the navigation and begin to translate. Which school owns that department? Is continuing education under academics or its own thing? Where do they keep the deadlines?

They are reverse-engineering your org chart, live, because your org chart is what the navigation encodes. It is there because internal structure is what approved the design. Every division got its menu item, every hierarchy its submenu. And it fails because visitors do not have an org chart. They have a task. The site makes them earn it.

How navigation gets this way

Nobody chooses confusing architecture. It precipitates out of governance. Every department needs visibility, though to whom is never asked. Every stakeholder reviews the sitemap, though against what is never stated. The safest resolution of every dispute is inclusion: one more top-level item, one more submenu.

The result reflects the organization admirably and serves it terribly. The audiences who matter think in tasks and their own vocabulary. Applicants, patients, donors, buyers: none of them think in org charts. They want ‘apply,’ ‘find a doctor,’ ‘give,’ ‘get pricing.’ They do not want ‘Office of Enrollment Services’ as a riddle to solve first.

The analytics show what they do instead. They bounce through wrong sections, give up to the search box, or leave. That last group is the campaign budget leaking away.

The good news: this is testable now

Information architecture has matured from taste into discipline. Rosenfeld and Morville’s field-defining book is in its third edition, the methods have settled, and best of all, they are cheap. Three earn a place in any redesign, in sequence.

• Card sorting reveals mental models. Write your content areas on cards, real ones or one of the online sorting tools, and ask fifteen or so actual audience members to group them and name the groups. The findings are reliably humbling. Things you would never shelve together get sorted together confidently. Your carefully chosen labels get renamed in plainer words. Distinctions central to internal politics are invisible to every participant. This is a days-scale exercise, not a research grant.

• Top-task analysis restores proportion. Your analytics and search logs already know the truth: a small set of destinations, a couple dozen typically, accounts for the overwhelming share of visits. List them, rank them, and let the ranking discipline the architecture. Top tasks get direct, obvious paths from the homepage. The long tail gets organized honestly but not top billing. Most navigation disputes dissolve under this data, because ‘my department deserves prominence’ has to argue with ‘four people a month visit it.’

• Tree testing validates before pixels. Strip your proposed structure to a bare text hierarchy: no design, no homepage, just the tree. Then ask users to find things in it. ‘Where would you look for application deadlines?’ Success rates and wrong turns tell you whether the structure works as a structure, before visual design can flatter or rescue it. Testing the naked tree costs days and prevents the expensive post-launch discovery that the beautiful redesign performs exactly like its predecessor.

The sequence is the argument

Notice what all three methods share: they happen before design. That ordering is the reform we push hardest, and the one projects resist. Everyone wants to see homepage comps, because comps feel like progress and card sorts feel like homework.

But we have watched too many redesigns spend their budget making the wrong structure gorgeous. Structure determines whether people find things. Design determines how it feels while they try. Both matter. One is structural.

We now decline to start visual design until the IA work is done and tested. We put that in proposals, and occasionally it costs us a project to a shop that will happily comp the org chart. The sites those projects produce keep validating the policy.

Keeping the architecture from regressing

A warning from every redesign we have revisited a year later: information architecture decays by default, because the forces that produced the org-chart site never disband. New programs launch and ‘need’ top-level placement. A dean’s pet initiative arrives with executive sponsorship. Each request is small, reasonable, and cumulative. Eighteen months post-launch, the tested structure has silently reverted toward the political one.

The defense is governance, boring and effective. A written rule for what earns navigation placement: task data, not the title of the requester. A named owner empowered to apply it. And an open offer that converts most disputes: we’ll test it. ‘Let’s see whether users look for this where you propose putting it’ is a sentence that ends arguments without making enemies, because it moves the disagreement from opinion to evidence. Evidence is cheap now.

Your site search, meanwhile, is the architecture’s ongoing telemetry. Searches spiking for a term means users are voting that browsing failed them. Zero-result searches are content gaps self-reporting. Searches for things prominently in the navigation mean the label is not speaking the audience’s language. Fifteen minutes with the search log each month is the cheapest IA review that exists.

The labels deserve their own confession

One sub-craft merits emphasis: naming. The words in your navigation should be your audience’s words, and your audience’s words are discoverable. They are in card-sort group names, in site-search logs, in the language of their emails and calls.

Internal vocabulary is astonishingly durable. Units named for donors. Programs named by acronym. Services named in the grant proposal that funded them. Every such label outsources translation work to the visitor. The search logs are the cheapest label research there is, sitting unread in nearly every analytics account we open.

Where to start, modestly

A full IA project is not always on the table, so here is the minimum viable version. Pull the top fifty pages and top hundred search terms from analytics. Run one afternoon of card sorting with a dozen genuine audience members. Note where their groupings and words diverge from your navigation.

That divergence list is your architecture backlog, ranked by how often each divergence gets hit. Some items will be label swaps you can ship this quarter. Some will wait for the redesign. All of them beat guessing.

Structure around user intent first. Decorate second. The org chart is a fine document for the intranet, where, come to think of it, people also cannot find anything. But that is another post.

One sentence to carry into your next redesign kickoff: every navigation decision is a bet about how strangers think, and strangers are available to ask. Fifteen of them, an afternoon, a stack of cards. The meeting where their answers contradict the sitemap is uncomfortable exactly once.