You have built something that works. The people who see it demonstrated understand it within about ninety seconds. And the site produces almost nothing from search, month after month, while a competitor selling something thinner sits above you on every query you care about.
We see this a lot, and the cause is usually not the one the team assumes. The instinct is to blame the content, so the company commissions more of it, and the new pages perform exactly as badly as the old ones. Content is rarely the first problem on a technical product’s site. It is normally the third.
This guide walks through the three layers in the order they actually bite: whether a search engine can read your site at all, whether your product has a name anybody searches for, and whether the people writing understand the product well enough to describe it. Work them out of order and you will pay for something that cannot produce a result.
Start With What a Crawler Can Actually Reach
Before content is the problem, reachability usually is. These faults are unglamorous, they are invisible from an office on fast hardware, and they are the reason a good page earns nothing.
- Compare what your site claims exists against what a crawler reaches. Run a crawl, export the URLs, and diff them against your sitemap in both directions. Anything in the sitemap that the crawl never reached is an orphan. Anything the crawl reached that is not in the sitemap is usually a surprise, and the surprise is usually large.
- Check whether the commercial pages render on the server. Product and category pages that build themselves client-side hand a crawler an empty container and fill it afterwards. That is sometimes rendered and indexed. Depending on it is a strategy that works right up until it does not, and the pages it fails on are the ones carrying the buying intent.
- Count your URLs. Filters, sort orders, and parameters can turn a few dozen products into tens of thousands of near-identical addresses. A crawler has a finite budget for your site. If most of it is spent on color and price permutations, your new pages get discovered late and revisited rarely.
- Read your own robots.txt and your canonical tags together. Blocking a URL from being crawled does not remove it from an index. It removes the crawler’s ability to see the directive that would have removed it. Canonicals that point at redirects, or at pages that canonicalize somewhere else again, produce an index in a shape nobody intended. Google’s own Search Central documentation is the authority here and it is worth reading before acting on a tool’s recommendation.
- Measure speed on the device your buyer actually holds. A synthetic run on a desktop connection will tell you everything is fine. A mid-range handset on a mobile network will tell you the truth. In our experience the culprit is almost never the framework. It is an oversized hero image and a stack of third-party tags loading in the head, each one defended by somebody and half of them with no owner at all.
We took a retailer through exactly this sequence in our work on organic search growth for a fragrance catalog, and not a single word of content was commissioned until the storefront could be read, indexed, and measured.
Your Product Probably Does Not Have a Keyword Yet
Here is the harder problem, and it is specific to genuinely novel products. If your product created a new category, or sits between two established ones, there is no established term for it. Keyword tools will return either nothing or a list of adjacent terms that belong to somebody else’s product.
Chasing the generic head term in that situation is the most expensive mistake available. You will compete against incumbents for a phrase that does not describe what you sell, and the traffic that does arrive will bounce.
- Search for the problem, not for the product. Somebody with no name for the solution can still describe the symptom perfectly. They search for what is going wrong, for the manual workaround they are running today, and for the product they are actively trying to replace. Those queries are yours, they are usually uncontested, and the intent behind them is far stronger than anything at the top of the funnel.
- Take the comparison and alternative queries seriously. People evaluating a purchase search for one named product against another, and for alternatives to the thing they already regret buying. If a competitor is the established name in the space, the honest comparison page is one of the highest-intent pages you will ever publish.
- Harvest the language your buyers already use. Search Console shows you the queries that reach you today, including the ones you never targeted. Your site search log shows what people could not find. Your support inbox shows what they call things. This is real language from real buyers, and it beats a tool’s list of terms sorted by volume.
- Accept that the best pages will have modest volume. A page that thirty qualified evaluators find every month is worth more than one that three thousand students find. Volume is the wrong measure for a considered technical purchase, and optimizing for it is how a company ends up with traffic and no pipeline.
Writing About the Product Accurately Is the Whole Advantage
This is the part most technical companies are best placed to win and most reliably give away. Your competitors are publishing pages about digital transformation. You can publish the specification.
Accuracy is the moat. An evaluator can tell within a paragraph whether the person writing understood the product, and so, increasingly, can a search engine, because the specific technical language is the only thing a specific technical query can match against. This is the whole reason we run technical marketing out of an engineering studio rather than alongside one.
- Publish the numbers you would give an engineer. Tolerances, operating ranges, supported protocols, latency under load, what happens when the network drops. Marketing departments strip these out to keep pages clean. Those figures are the reason a buyer chooses you, and they are the exact strings a serious evaluator types into a search box.
- Answer the evaluation questions in public. Your engineers already answer the same six questions by email every week. Each of those emails is a page. It is already written in the right register, by somebody who knows the answer, for somebody who was qualified enough to ask.
- Let the specification set the vocabulary. Where an industry has a settled term for a component or a standard, use it, even when it is uglier than the marketing alternative. Renaming a known thing removes you from every query that uses its real name.
- Mark up what is true, and only what is true. Structured data is oversold and still worth implementing properly. Schema.org types for products, FAQs, and organizations let a listing present as more than a blue link. Emitting ratings you do not have is both a policy violation and trivially detectable.
- Link the writing into the pages that sell. Editorial content that never links to the commercial pages it supports is a blog. Linked properly in both directions, it is an argument for why those pages deserve to rank.
If the writing is going to be done by an outside team, the quality of what comes back is set almost entirely by what you hand over. We wrote a separate guide on briefing an agency on a technical product for that reason.
The Order to Fix Things In
The sequence matters more than any individual item on it, because each stage determines whether the next one can be measured.
- Establish a baseline in the first week. Crawl health, indexation, current ranking positions, and real-world page speed, recorded before anything changes. Most arguments about search performance are actually arguments about where things stood at the start, and that is not resolvable afterwards.
- Fix crawlability. Orphans, facet explosions, contradictory canonicals, server rendering on the commercial pages. Nothing downstream works until this does.
- Fix performance. Images, blocking scripts, third-party tags with no owner. This is usually the cheapest measurable win on the list.
- Fix structure. Information architecture, internal linking, and structured data, so that the site’s shape tells a crawler which pages are important.
- Then publish, against measured demand. Problem-shaped queries, comparison pages, and the questions your engineers answer by email.
- Report against the baseline, monthly, on one page. What changed, what it produced, and what happens next. Longer reporting formats do not get read.
A Channel That Compounds
The argument for doing this properly is not that search is cheap. It is that search is the one acquisition channel where the work accumulates. A paid campaign stops producing on the day the card stops. A page that is crawlable, fast, accurate, and internally linked keeps earning, and a page published in the second month can still be generating qualified inquiries in the twentieth.
That is also why the order is worth defending against the pressure to skip to the visible half. Content is what everybody can see, so content is what everybody wants to buy first. The technical foundation underneath it is what decides whether any of it was worth writing. If you want to talk through where your own site sits in that sequence, our technical marketing practice starts most engagements with an audit and a baseline before anything is recommended, and you can see how we structure that in our process.