Why most SEO recommendations never get implemented
The common failure in B2B SEO is not bad strategy. It is a good strategy that sits in a document.
An agency runs an audit and produces a list: fix the redirect chains, restructure the URLs, add schema, resolve the duplicate title tags, sort out pagination, improve the Core Web Vitals. The list is usually correct. It then goes to a development team who did not commission it, do not own the outcome, and have a roadmap already full of things the business asked for.
Six months later maybe a fifth of it is done. The agency reports that rankings have not moved because the recommendations were not implemented. The client concludes SEO does not work. Both are describing the same structural problem: the people who identified the work could not do it.
This is the gap we were built for. We are engineers, so a technical audit produces a pull request rather than a slide. Redirect logic, rendering, schema, internal linking, performance, and CMS structure are all things we change in the codebase and ship. There is no handoff, because there is nobody to hand off to.
If you already have an SEO firm, keep them
This is a genuine offer rather than a competitive tactic, and it is one of our more common engagements.
If you have a strategist you trust, the strategy is not the constraint. Their recommendations arriving in your backlog and dying there is the constraint. We will take their audit, their priorities, and their reporting cadence, and implement against it. They keep the relationship and the strategic work. We do the part that requires touching the code.
We do this today for a client whose SEO firm produces the plan and whose development capacity we are. Nobody has to be replaced for the work to start moving.
What is actually different about B2B SEO
Consumer SEO advice dominates the internet and most of it misleads B2B teams in the same three ways.
Volume is the wrong success metric. A B2B category term might get 200 searches a month. Ranking for it can be worth more than ranking for something with 20,000, because the 200 are people evaluating a purchase with a five or six figure annual value. Traffic reports that celebrate total sessions are usually measuring the wrong audience, and the fastest way to grow that number is to attract people who will never buy.
You are writing for a committee, not a searcher. The person who first finds you is rarely the person who signs. They research, then they have to persuade a technical evaluator, a finance owner, and often a procurement process. Content that only serves the initial searcher stops working exactly when the deal gets serious. The pages that quietly matter most are the ones a champion forwards internally: comparisons, implementation detail, security and compliance answers, pricing logic. In technical categories such as life sciences and manufacturing this is most of the buying process.
The cycle outlasts attribution. If the sale takes nine to eighteen months, a page read in January contributes to revenue booked the following year, by which time nobody remembers the page existed. Without first-touch data wired to the CRM, the SEO programme cannot be defended in a budget conversation, no matter how well it worked. This is why we treat tracking as part of the SEO engagement rather than as a separate purchase.
Technical SEO that requires a developer
Worth separating from the technical SEO that does not, because the second kind is what most audits deliver.
Editing title tags, writing meta descriptions, and fixing headings can be done in any CMS by anyone. Useful, rarely the bottleneck.
The work that actually moves a stuck B2B site tends to require someone who can deploy:
Rendering. If content only exists after JavaScript executes, you are betting on how a crawler handles your particular framework. Server rendering or static generation removes the bet. This is a build decision, not a plugin.
URL and redirect architecture. Consolidating duplicate URL forms, removing chains, and handling trailing slashes and parameters correctly is edge or server configuration. On this site it is handled in the worker, because the platform’s default behaviour was not the behaviour we wanted.
Structured data generated from content, not hand-written per page. Schema that is maintained by hand goes stale within a quarter. Schema emitted by the template stays correct.
Internal linking as a system. Related content, hub pages, and contextual links generated from the content model rather than added manually by whoever remembers.
Performance under real conditions, which usually means auditing third-party scripts rather than optimising images. Most sites are fast until the tag manager, the chat widget, and the heatmap tool load, and keeping that in check is ongoing maintenance rather than a one-off fix.
Crawl budget and index control on large sites: what is indexable, what is canonical, what should never be crawled.
An audit that stops before this line is describing symptoms.
AI answer engines, without the hype
Search results increasingly include an AI-generated answer, and assistants are becoming a research channel in their own right. Two claims are circulating, and they cannot both be true: that this changes nothing, and that it makes SEO obsolete.
What has actually changed is the shape of the query and the shape of the payoff. Someone asking an assistant writes a sentence describing a situation, including constraints and what they have already ruled out, rather than typing three words. And being cited in an answer can produce no click at all while still influencing a shortlist.
What has not changed is that the systems doing the citing are reading the web, and they favour content that states things clearly, attributes claims, and is structured well enough to extract. The practical work overlaps almost entirely with good SEO: clear headings that match real questions, direct answers near the top of a section, factual precision, structured data, and a site that renders without JavaScript.
Our current position is that this is worth deliberate attention and not a separate budget line. The teams selling GEO as a distinct discipline with distinct pricing are mostly repackaging what they already do. Where it does change our work: we write answers to be quotable in isolation, we keep FAQ and definitional content structured, and we monitor citation appearance alongside rankings.
How to tell whether SEO is working before rankings move
Rankings are a lagging indicator and the first three months of a serious programme can look like nothing is happening. There are earlier signals, and asking for them is a reasonable way to hold an agency to account.
Crawl and index coverage. Pages discovered, pages indexed, and the gap between them. If the site was being crawled badly, this moves first, often within weeks.
Impressions before clicks. A page appearing in more searches at position 30 is not producing traffic and is a real signal that it has begun to register.
Movement in the 8 to 25 band. Positions 11 to 20 produce almost no clicks and are one push from producing many. Watching that band tells you more than watching the top three.
Query breadth. The number of distinct queries a page surfaces for, which grows before position does.
Lead quality, once traffic arrives. The check that matters most and gets run least. More traffic with a worse conversion rate usually means the content is attracting the wrong reader.
Where we are not the right fit
We are not a link building agency and we do not buy links. We earn them with content and digital PR where that is worth doing, which sits inside organic growth. If a link volume programme is what you want, someone else should run it.
We do not do local SEO at multi-location scale, and we do not do e-commerce SEO for large catalogues. Both are specialisms with their own tooling.
And if the site has almost no content and the budget only covers technical work, technical SEO alone will not produce rankings. It removes obstacles; it does not create reasons to rank. We would rather say that at the start than bill for six months of it.