The help centre is a growth channel: SEO that deflects tickets

Somewhere in your analytics there's a line item nobody reads properly. It's the help centre, and it's quietly doing three jobs: answering customers, ranking for searches your marketing pages can't touch, and telling you exactly what to build next. Most companies staff it with whoever drew the short straw, write it in the passive voice of an apology, and let it rot two releases behind the product.

That's a waste of the highest-intent audience you'll ever get.

Why support content wins searches marketing content can't

Search intent is the whole game. For "best budgeting software," you're competing with listicles written by people who've tested forty tools for twenty minutes each. For "why did my direct debit fail," the searcher has one tool in mind — yours — and the only competition is a forum thread from 2022 full of guesses.

Problem-aware queries have properties that make them a gift:

• Low competition. Nobody else can write authoritative docs about your error states, your export format, your billing engine. You're the canonical source by default.
• Long tail with volume. Each query is small; the aggregation is enormous. A mature product accrues thousands of "how do I X in [product]" searches.
• Visitors arrive pre-qualified. Someone troubleshooting your product already bought, or is mid-evaluation checking whether breaking things is survivable. Both audiences matter.
• Forgiving format. Docs don't need cinematic heroes. They need to be right, fast and findable — which is cheaper to produce than the marketing site by an order of magnitude.

The deflection math on top of that: if a solid doc page seen by 4,000 people a month convinces even 3% of them to not open a ticket, and a ticket costs $8–15 to handle well, one good article pays for a writer's week. We ran this calc with a fintech client — nineteen rewritten articles, roughly $11k a month in avoided ticket cost at their volumes, and organic sessions to the help centre up 240% in two quarters because Google started trusting the section.

Start with the plumbing from our technical SEO checklist — indexable, canonical, fast — because a help centre locked behind a JS-rendered widget ranks for nothing.

The anatomy of a doc that ranks and resolves

Most help articles fail one of two ways: written for search (throat-clearing intro, keyword-stuffed headings, answer buried at paragraph six) or written for the already-logged-in (screenshots of a UI that changed in March, jargon nobody searched for). The fix is a structure that serves both masters:

1. Title = the query. "How do I add a second accountant to Northwind Ledger?" beats "Multi-user access management." Write the question the way a human types it, verbatim where possible.
2. The answer in the first viewport. One or two sentences: yes/no, where to click, any prerequisites. Respecting the reader's time is also what earns the featured snippet.
3. The steps, numbered, present tense. One action per step. Screenshots only where a UI is genuinely ambiguous, and every screenshot gets alt text that makes it redundant during the next redesign — because alt text doubles as your freshness alarm.
4. The why and the edge cases. Limits, permissions gotchas, "if you see X it means Y." This is where authority lives and where the word count that signals depth comes from honestly.
5. A graceful exit. Link to the two or three adjacent articles, and offer support contact for the case the doc can't solve. A dead-end doc manufactures tickets.

One more trick we've standardised: structured data for FAQs and HowTos on doc pages that genuinely qualify. Schema won't rescue a thin article, but on a real one it buys SERP real estate you don't otherwise get for a support page.

Your search box is the editorial calendar

Every help centre has a search box, and every search box is a focus group that never stops. Zero-result searches are the single most under-used dataset in SaaS. They're customers telling you, in their own words, exactly what's missing.

The discipline is simple: log queries, review monthly when volumes allow — weekly if you're small and every query is precious — and bucket them:

• Missing doc. They searched "GST on invoices," you have nothing. Write it. This is the backlog, pre-prioritised by demand.
• Wrong language. They searched "cancel," your doc is called "Subscription termination." Rename the doc to the customer's dialect; this is voice-of-customer mining with the volume knob your support queue already listens through.
• Product smell. Hundreds of searches for "undo delete" is not a content problem. It's a product gap, and the query log is your cheapest evidence for the product roadmap argument you'll need to have.

The same search-mining logic applies to your marketing site, by the way — we wrote up the general practice in your search box is a research department. The help-centre version just has a closed loop you can actually measure: doc shipped → query resolves → ticket volume on that topic trends down.

Content ops: docs that don't rot

The failure mode of every help centre is not bad writing — it's drift. The product ships weekly; the docs describe a product from February. Fixing drift is an operating-model problem, not a writer problem:

• One owner, not a committee. A single person (or one per product area) with the authority and the calendar time to ship doc changes. "Everyone owns docs" is how docs get owned by no one.
• Docs in the definition of done. No feature merges without the doc update drafted. This sounds heavy; in practice it's a 20-line template in the PR description and it kills 80% of drift at the source.
• A stale sweep on a calendar. Sort every article by last-reviewed date, oldest first, and review the tail on a fixed cadence. Screenshots and UI labels expire; refresh them or replace them with text that ages better.
• Retire aggressively. A doc for a sunset feature doesn't just clutter — it actively misleads and it dilutes the domain's trust. Redirect it, don't delete it into a 404.
• Write for the AI assistant too. If you're running any AI triage or support routing, the help centre is its retrieval corpus. Well-structured docs make the assistant good; rotting docs make it confidently wrong.

This is genuinely unglamorous work, which is precisely why it's defensible. Anyone can write a clever landing page. Keeping five hundred pages true, week after week, is a moat built out of diligence.

Measuring it honestly

Two dashboards, kept separate. Deflection: ticket topics before/after doc launches, ratio of doc views to tickets on tracked topics, and the self-report in your contact form ("did you check the help centre?"). Be sceptical of any "deflection rate" your tooling computes by mere pageviews — a view is not an avoided ticket, and we've covered how to keep attribution honest about what it doesn't know. Growth: organic sessions to the docs subdomain or section, rankings for problem-aware terms, and — the one that surprises CFOs — doc engagement as a leading indicator of retention. Customers successfully self-serving are customers staying.

And when you need help at the point of need inside the product itself, the help centre is the corpus again. One investment, three surfaces, compounding returns.

Key takeaways

• The help centre ranks for problem-aware searches marketing can't win — you're the canonical source for your own product's problems, with near-zero competition.
• Title docs as the query, answer in the first viewport, number the steps, handle edge cases, exit gracefully to related articles or support.
• Zero-result help searches are your editorial calendar: missing docs, wrong language, and product gaps, pre-prioritised by demand.
• Drift is an operating model problem: one owner, docs in the definition of done, a stale-content sweep, and aggressive retirement.
• Docs now feed three surfaces — SEO, deflection, and any AI assistant — so quality compounds across all of them.
• Measure deflection and growth separately, and stay sceptical of any "deflection rate" computed from pageviews alone.

FAQ

Should the help centre live on a subdomain or a subfolder? Subfolder if you can (yoursite.com/help). Subdomains traditionally split authority; subfolders consolidate it, and the help centre's rankings lift the whole domain slightly while the root domain lends freshness to new docs. If your help tool forces a subdomain, fine — hundreds of successful operations do — but make the cross-linking deliberate and check that the subdomain is properly verified and monitored, not orphaned.

Won't public docs make it easier for competitors to copy us? They'll copy your features by using the product itself. What public docs buy you — search demand capture, customer self-sufficiency, visible competence for evaluators — dwarfs the theoretical risk. Exception: anything touching security internals, fraud logic, or abuse-prevention rules. Those stay behind auth, and honestly some shouldn't be written down anywhere.

How long should a help article be? As long as the problem and no longer. A settings toggle is 120 words. Billing reconciliation is 900. The mistake is applying a uniform template and padding short answers to look substantial — padding buries the answer, and the answer is the product. Write to the question, then stop.

Can AI generate our help docs? AI can draft from a spec or a recording of the feature; humans must verify every step against the real UI, because a wrong doc is worse than none — it trains customers to distrust the whole centre. Where we do use it heavily: clustering zero-result searches, detecting stale screenshots against new UI captures, and drafting first-pass answers for the assistant that the team then edits. Acceleration, yes; autopilot, no.

When does it make sense to get outside help on docs? Two moments. At the starting line, when you need the structure — taxonomy, templates, the ops model — set up properly in weeks instead of discovering it by pain over a year. And at a major migration or rebrand, when the doc corpus needs a coordinated sweep your support team can't absorb. That's a bounded engagement, which is exactly the sort of thing a growth engagement covers before handing you the keys.