Technical SEO has a reputation problem: half the industry treats it as arcane, the other half as an upsell. It is neither. It is maintenance engineering for a pipeline with three stages — the crawler must find the page, the renderer must understand it, the index must keep it — and every common failure is a blockage at one of those stages. This guide walks the pipeline in order.

Crawl: the budget is real, if you are big enough

Engines allocate each site a crawl budget — how many URLs they will fetch per unit of time. Small sites rarely exhaust it; large ones routinely do, in which case waste becomes the enemy. The classic wastes: faceted URLs generating infinite combinations, calendar pages without end, parameter duplicates, soft 404s returning “200 OK” on dead ends. The instrument that tells you the truth is the server log file — not a crawler simulation, the actual record of what the search engine fetched. An hour of log reading answers questions no tool can: what share of crawl goes to pages that matter, what share to duplicates, whether the new section is being discovered at all.

Render: the page the engine actually sees

Modern engines render pages like browsers do, executing JavaScript to see the final content. The catch is queueing: rendering happens on a second pass, which means a page whose meaning arrives only via client-side script is always a step behind — and occasionally a step missing. The doctrine is progressive: HTML carries the meaning, enhancement arrives on top. Test what the engine sees with its own rendering tools, not your monitor. Rendering overlaps performance, where the Core Web Vitals — loading, interactivity, visual stability — are the honest scoreboard; they are a ranking factor at the margin and a user experience factor everywhere. The practical wins are unglamorous: compress and size images (ours are gradients and CSS, partly for exactly this reason), defer what is below the fold, prune third-party scripts with each season’s enthusiasm.

Index: one URL per page, deliberately

An engine will happily index five versions of one page — http, https, trailing slash, parameter, print — and split the signals across them. The standing defences: canonical tags that say which version is the real one, redirects that fold old paths into current ones, sitemaps that list exactly the pages you want indexed and nothing else, and internal links that agree with all of the above. This site’s own hygiene is the worked example: canonicals to a single host, historic paths such as the old /uk address redirected rather than left to rot, and a sitemap matching the real URL set. When consolidating or restructuring, change one thing at a time and watch the indexation count before changing the next — the debugging method that saves entire migrations.

Technical SEO is the discipline of removing reasons for a page not to rank. It buys no rankings by itself; it unfreezes the rankings your content has already earned.

The standing rhythm

  • Monthly: read the logs, scan the indexation report, fix the crawl errors that recur.
  • Quarterly: full-site crawl against a threshold list — response codes, redirect chains, canonical disagreements, orphans, vitals regressions.
  • On every release: the five-minute check — canonical, robots directives, sitemap entry, one rendered fetch. Most catastrophic technical failures are five minutes old.

How much does any of this matter?

More than nothing, less than the conference talks imply. Our honest ordering for most sites: content quality first, intent match second, technical soundness third — because it is the multiplier, not the engine. But note the asymmetry: technical failures are silent and total. A broken canonical or a stray noindex does not reduce your traffic by ten per cent; it can remove a page from consideration entirely. That is why the maintenance rhythm, however dull, is non-negotiable — and why the economics page counts tools as a standing cost of the whole operation, alongside everything else traffic costs.