Discussion
  • Enterprise SEO

Embedded SEOs in product teams or a central centre of excellence: what failure modes have you seen in each model?

Published 1 reply

Discussion starter: an enterprise SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    Both models fail, and they fail in opposite directions. Embedded SEOs tend to fail through drift and blind spots: they get absorbed into one team’s roadmap, each team solves the same SEO problems differently, and the parts of the site several teams share (robots.txt, sitemaps, redirects, navigation, shared templates) end up with no owner. A central centre of excellence tends to fail through distance and throughput: it hears about changes after designs are approved, becomes a review queue that teams learn to route around, and ends up accountable for organic performance with no authority over engineering time.

    The qualification is that the org chart is the wrong unit of decision. What should drive it is where organic risk and opportunity sit: how many teams ship changes to pages that earn search traffic, how much of the site runs on shared templates, and how many SEOs you have relative to teams. The arrangement I would expect to hold up on a multi-team site is a hybrid: a small central team that runs shared infrastructure, standards and monitoring as an internal product, with SEOs embedded only in the few teams whose releases carry most of the organic exposure. The hybrid has failure modes of its own, covered below.

    No controlled public comparison of the two models exists that I can point to, so read what follows as mechanisms and diagnostics rather than measured outcomes.

    Why the two models break in opposite ways

    An embedded SEO works like a safety officer assigned to one building crew. Problems on that site get caught while the frame goes up, but the officer has no say over the building code, and crews without an officer build unchecked. A central team works like the council’s building inspector: one code for everyone, but it sees each building at inspection, when moving a wall is expensive.

    In delivery terms, the difference is when SEO input arrives in a team’s cycle (discovery and design, or review and QA) and how much of the site each SEO can see. Embedded SEOs get early input and narrow sight. A central team gets wide sight and late input. Most of the failure modes below trace back to one of those two gaps.

    Failure modes in the embedded model

    • Capture. The embedded SEO becomes the team’s generalist: copy edits, QA, analytics tickets, release notes. SEO work competes with sprint goals and loses, and the performance review is written by a product manager who measures feature delivery. A year later the team has a helpful colleague and no SEO program.
    • Standards drift. Each team solves canonicals, pagination, faceted URLs, structured data and internal linking its own way. No single choice is wrong, but a crawl split by template shows several patterns for the same problem, and every site-wide fix now needs several implementations.
    • Orphaned shared assets. robots.txt, XML sitemaps, the redirect map, hreflang, global navigation and shared design-system components sit on nobody’s roadmap. They change as side effects of some team’s feature work.
    • Coverage follows headcount, not risk. Embedded SEOs go to the teams that own landing pages. The changes with the widest blast radius often come from teams that rarely get one: platform, framework upgrades, CDN and edge rules, consent banners, authentication. Google’s JavaScript SEO documentation shows what is at stake: pages with non-200 status codes might not be rendered, and a noindex in the original HTML can stop rendering, so removing it with JavaScript may not work. A framework upgrade that changes error handling or the initial HTML can cause either problem, and no embedded SEO sits in that team’s reviews.
    • Uncoordinated targeting. Two teams launch pages for the same queries because each owns its own acquisition number. The result is cannibalisation, followed by a quarterly argument about whose traffic it was.
    • Isolation. One SEO per team means no peer review, no shared playbook, and knowledge that leaves with the person. Embedded roles filled by mid-level SEOs add a standing problem: they are expected to push back on a senior product manager with no escalation path.

    Failure modes in the central model

    • The queue. Every request becomes a ticket with a service level. Teams on weekly releases stop waiting, ship, and ask for review afterwards, so the centre drifts into post-release incident response.
    • Late involvement. The centre sees designs after approval, so every recommendation arrives as rework. Implementation rates fall, audits go unread, and the team earns a reputation as the department of no.
    • Accountability without authority. The centre owns the organic target but no engineering capacity. Its backlog competes with each product roadmap for sprint slots and tends to lose to features with a named owner.
    • Generic standards. Guidelines written for the whole site ignore a team’s real constraints (inventory rules, pricing logic, legal copy), so teams treat them as optional.
    • Attention politics. Time goes to the largest or loudest team. A small, high-margin section with a fixable problem can wait a year.
    • Context gap. Central SEOs don’t sit in user research or planning, so their recommendations can be technically right and commercially naive, and product teams notice.

    How to tell which failure you actually have

    The evidence is usually already sitting in tickets, crawls and Search Console. I would pull these before proposing any reorganisation:

    Evidence Points to an embedded-model failure Points to a central-model failure
    SEO incident log for the last 12 months: who shipped it, whether SEO saw it before release, days to detect Incidents cluster in teams with no SEO coverage Incidents come from teams that requested review but shipped before it happened
    Lead time from recommendation to release, and share implemented within two quarters Fast in covered teams, near zero elsewhere Slow everywhere, with a low implementation rate
    Crawl compared by template Different canonical, pagination or structured data patterns in each team’s section Consistent patterns, but some page types bypass the standard because it doesn’t fit them
    Query and page data from Search Console Pages owned by two teams share impressions for the same queries Rarely the main symptom
    Two-week time audit of SEO staff Most embedded time goes to non-SEO product work Most central time goes to reviews and escalations, little to shared tooling
    Ownership list for shared assets Blank owner cells for robots.txt, sitemaps, redirects, navigation An owner exists, but nothing stops teams changing the asset

    I would start with the incident log, because it shows where unreviewed changes come from. It is easier to build when releases are logged against performance data: Search Console supports custom chart annotations, which makes it quicker to match a traffic change to the team that shipped just before it.

    The hybrid, and where it breaks

    The Team Topologies vocabulary, from Matthew Skelton and Manuel Pais, helps because it separates kinds of help. It describes stream-aligned teams, aligned to a flow of work from a segment of the business; platform teams, which provide a compelling internal product that accelerates delivery by stream-aligned teams; and enabling teams, which help stream-aligned teams overcome obstacles and detect missing capabilities. The framework isn’t about SEO, but it maps onto it well:

    Central SEO as a platform

    Owns shared assets and offers them as a service: SEO-correct components in the design system, sitemap and redirect services, status-code, canonical and rendering checks in the build pipeline, monitoring and alerts. Standards are enforced by failing builds, not by documents.

    Central SEO as an enabling team

    Works alongside a team for a fixed period when it is about to ship something with organic impact (a new page type, a migration, a framework change), then steps back once the team can run it.

    Embedded SEOs, used sparingly

    Only in teams whose releases carry most of the organic exposure, with a dotted line to the SEO lead for standards, peer review and career progression.

    Ownership needs writing down, because the hybrid fails wherever two groups each assume the other owns something:

    Asset Owner Consulted Enforced by
    robots.txt, XML sitemaps, redirect rules Central SEO with platform engineering Teams whose URLs change Change review plus automated checks on status codes and redirect chains
    Canonical, meta robots and hreflang logic Central SEO sets the rules; shared templates implement them Embedded SEOs Tests that compare rendered output with the rule
    Structured data and link components Design system team Central SEO Acceptance criteria on each component
    Page-type content and query targeting Product team, with its embedded SEO where there is one Central SEO, which keeps a map of which team owns which queries A quarterly overlap review
    Monitoring and incident response Central SEO The team that shipped Alerts routed to the shipping team

    The hybrid’s own failure modes are predictable. The dotted line turns decorative when the product manager writes the embedded SEO’s review alone. The centre becomes a policy office that publishes standards nobody enforces. Platform work goes unfunded because it has no feature to demo. The fixes are unglamorous: the SEO lead contributes to embedded SEOs’ reviews, shared-asset ownership is written down, and platform SEO work is budgeted like any other internal tooling.

    Which way to lean

    Condition Lean towards embedding Lean towards a central platform and enabling team
    Teams shipping to organic landing pages A few, each owning a high-traffic page type end to end Many, each touching a little
    Shared templates and components Page types owned by one team Most templates changed by several teams
    Organic’s role in the team’s goals Primary acquisition channel for that surface Secondary channel
    SEO headcount relative to teams Enough to embed where it matters and still staff the centre Far fewer SEOs than teams
    Engineering tooling Little automation, so review at design stage is the main safeguard Design system and build checks in place, so rules can be automated
    Release cadence Planned releases with design reviews Continuous deployment across many teams

    If I had to weight one row, it would be shared templates. When most organic landing pages render through components that several teams change, embedding alone can’t protect them, because the risk sits between teams rather than inside any one of them.

    Hypothetical example:

    A marketplace has 14 product teams and three SEOs. Most organic landing pages are listing and category pages owned by two teams, but nine teams change the shared header, filter and listing-card components. Embedding all three SEOs leaves eleven teams unreviewed; keeping all three central leaves the two highest-exposure teams in a queue. A workable split: one SEO embedded across the two listing teams, and two central SEOs who own the shared components, redirects, sitemaps and build checks, and run enabling work with whichever team launches a new page type that quarter.

    Pilot before you reorganise

    Org models can’t be split-tested cleanly, but a time-boxed pilot beats a reorganisation on opinion. Pick the team with the most organic exposure, embed or assign a named SEO for two quarters, and keep a comparable team on the current model as a loose comparison. Two quarters, because product teams commonly plan quarterly: the first quarter shows whether the SEO got into planning, the second whether that changed what shipped. Organic outcomes lag further and are confounded by core updates and seasonality, so judge the pilot on process measures first:

    • Share of SEO issues caught before release rather than after
    • Lead time from recommendation to release, and implementation rate
    • Incidents per quarter and days to detect
    • Share of the SEO’s time spent on SEO work

    What would change the decision: if the pre-release catch rate doesn’t rise, or the embedded SEO’s time drifts to non-SEO work, the problem is intake or authority rather than proximity, and a design-review gate plus build checks will do more than headcount. If catch rates rise but incidents keep coming from platform and framework changes, the next hire belongs in platform coverage, not another embed.

    Where I would start

    If this were my organisation, I would stop debating the model and write two lists: the teams whose releases touch pages that earn organic traffic, and the shared assets those pages rely on. Embed where one team owns a high-exposure page type end to end. Centralise, and automate, everything that sits between teams. Then let the incident log decide the next hire, because it shows where unreviewed changes are coming from.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.