Fifty Authors, One Website: Workflow Before Chaos

A university publishing audit should begin with a census: how many people can put content on the website? The answer is often far higher than anyone expects, with dozens of active accounts spread across departments.

Ninety-one people with publish access, across departments, with no shared definition of who approves what. Drafts circulated by email attachment, gathering conflicting edits. Approvals happened in hallways and vanished with the conversation. The CMS, a perfectly capable one, faithfully recorded whatever survived the process, which made the website an accurate archive of organizational entropy.

Nothing was broken, exactly. Every individual was doing their job. Nothing was governed, either, and the distinction is the subject of this post. The client’s situation is not an outlier. It is the default state of institutional publishing at scale, and the fix is a design discipline most organizations have never been told exists: at multi-author scale, workflow is the product.

Why scale breaks the informal system

Small teams publish on trust. Three people who sit near each other need no formal workflow, because the workflow is conversation. The system fails quietly as authorship grows, and the failure has a predictable anatomy.

Review becomes ambiguous. Who approves a program page: the department, communications, both, neither? So content ships unreviewed or stalls unowned. Standards drift, because ninety-one people have ninety-one recollections of the style guidance. Timing becomes accidental. The news item publishes a week late. The outdated policy lingers a year.

And accountability flips. When everyone can publish, no one is responsible for what was published. That surfaces at the worst moments (the factual error, the compliance miss, the tone problem) as a hunt for who touched it last.

None of this is a people problem. It is what happens when an organization scales its authorship without scaling the system authorship lives in. It resolves the way all such problems resolve: by designing the system deliberately.

The workflow layer, component by component

• Roles that mean something. The vocabulary comes first. Author drafts. Editor shapes and verifies. Approver owns the risk for a domain. Publisher pushes live. Small sites collapse several roles into one person. The design act is deciding, per content type, which roles exist and who holds them. Write the grid down. Most organizations discover mid-exercise that nobody currently holds the approver role for their riskiest content, which is the finding that justifies the meeting.

• Permissions that mirror the grid. The platforms are ready. WordPress’s role system. Drupal’s permissions and the Workbench suite built precisely for editorial states: draft, in review, approved, published. The principle from our governance series applies word for word. Permissions should mirror actual authority. Most of the ninety-one accounts become authors within their sections rather than publishers of everything, and the change is experienced not as demotion but as relief. People mostly never wanted the keys. They wanted their edits to go somewhere reliable.

• Review service levels, so drafts don’t rot. The commonest post-rollout failure: workflow installed, queue created, queue ignored, and the informal system returns wearing the formal one as decoration. The fix is service-level agreements on review. Forty-eight hours for routine content, an expedited lane for the time-sensitive, per our editorial-velocity coverage. The queue stays visible and the misses get reported. A workflow without clocks is a suggestion.

• The editorial calendar as operations, not aspiration. Who publishes what, when, where it runs, and who confirmed it. The calendar converts publishing from a series of surprises into a schedule. It surfaces the conflicts, three departments launching the same week, while they are cheap. And it gives the owner the instrument every operation runs on: a view of what is supposed to happen next.

• Taxonomy governance, before the soup. Fifty authors without a shared vocabulary will invent fifty category schemes. Our WordPress coverage documented the soup in 2007, and the recipe has not changed. The tag policy, the category structure, and the warden who tends them are workflow components. Classification is part of publishing, and the archive’s findability depends on it.

Rolling it out without an uprising

Process changes fail on imposition and succeed on service, so here is the sequencing we use. Start with the pain: the department whose publishing is most chaotic and knows it. Design their lane with them, and let their before-and-after recruit the next section.

Frame the permissions change honestly. Fewer keys, faster publishing, because the review lanes have clocks now and the queue has an owner. Train in role terms, here is what authors do, here is what happens next to your draft, rather than software terms.

And measure the operation like one. Time from draft to publish. Review-queue depth. Misses against the service levels. Published-content error rates.

Those numbers, six months in, are the argument that survives leadership turnover. At our census client they currently read: publish access down from ninety-one to a governed structure, draft-to-live time cut by more than half for routine content, and the hallway approval extinct. The entropy was never necessary. It was just undesigned.

Two objections, answered from the rollouts

The pushback arrives in two reliable forms, so the answers travel with the model now.

First: ‘workflow will slow us down.’ The evidence runs opposite, and the census client’s numbers above are typical rather than exceptional. Informal systems feel fast to the person publishing and are slow in aggregate, because the delays hide in the gaps. The draft awaiting a hallway that never happens. The review nobody knew they owed. Formal lanes with clocks are slower per step on paper and faster end to end in practice. The draft-to-live metric settles the argument within a quarter.

Second: ‘our culture won’t accept process.’ Usually true of process imposed and false of process co-designed. The start-with-the-pain sequencing exists precisely because the department drowning in its own chaos is not culturally opposed to a lifeguard.

The deeper answer to both objections is the reframe the whole post rests on. This is not process for its own sake. It is the difference between the institution knowing what it publishes and hoping. Cultures accept knowing, once they have experienced hoping’s invoice. Every institution over a certain size has, whether or not the invoice was ever itemized. Itemize it (the stale policy that misled someone, the error hunt that ate a week) and the workshop schedules itself.

Publication, not pile

This is where content strategy meets operations, and it is the difference between a publication and a pile. The tooling has been ready for years. The missing ingredient is almost always the decision to treat internal publishing as a system worth designing, with the modest ceremony that implies. A workshop, a grid, some clocks. A calendar and a warden.

That decision costs an afternoon and repays it weekly forever. Ninety-one people with keys is not a capability. It is a liability. A governed lane for each of them is the capability, and the institutions that build it stop experiencing their own website as weather. Design the system. The authors, it turns out, were waiting for one.