Founders ask us about search more often than they ask about almost anything else that is not the build itself. The question is usually some version of the same thing: we have a product, nobody is looking for it by name, and we cannot afford to buy every click forever. What should we actually write?
The honest answer is that a startup SEO strategy is mostly a set of decisions about what not to publish. The writing is the easy part. Deciding which twenty posts to skip so that four can be finished properly is the part almost nobody gets right, and it is the part that decides whether any of it produces revenue.
What follows is the advice we give when a founder asks. It assumes you are early, your domain has little authority, and your time is the scarcest thing you have.
Pick fewer clusters and finish them
The standard advice is to build topic clusters: a substantial cornerstone piece on a question your buyers genuinely search, with supporting posts hanging off it and linking back. That advice is correct. The failure mode is not the model, it is the count.
What happens in practice is that a founder plans six clusters, finds the research interesting, and by month three has fifteen loosely related topics with one post each. Every post is individually fine. None of them has anywhere to point. A supporting post with no cornerstone to link to is an orphan whatever its word count, and a cornerstone with nothing pointing at it is doing a cornerstone’s job on an orphan’s link equity.
Three clusters with four supporting posts each will beat eight clusters with one, because the internal linking resolves into a shape rather than a spray. The test is simple. Look at your plan and ask which cornerstone each planned post will link to. If more than a couple of rows cannot answer that, you have a list of article ideas, not a strategy.
Finishing a cluster is also more motivating than it sounds. A cluster that is genuinely complete, say a cornerstone on choosing between a fractional CTO and a product studio plus the rates piece, the three-way comparison and a contrarian piece on when not to hire one at all, hangs together in a way a reader can feel. They arrive on one page and find the rest of the answer.
Commercial intent beats keyword difficulty
This is the mistake with the highest cost, and it is seductive because it looks like discipline.
You export keyword research, sort by difficulty, and start with the easy terms. Sensible on the face of it. The problem is what makes a term easy. Low-difficulty terms are low difficulty because very few people are competing for them, and very often nobody is competing because very few people search them with money in hand. Meanwhile the terms that would actually change your business, the ones around cost and pricing and choosing a supplier, sit high enough that established competitors own the first page and will keep owning it for a while.
The fix is not to abandon the easy terms. It is to stop treating them as a warm-up for the hard ones. A page that ranks in a month for a term nobody buys from is not a stepping stone towards ranking for the term they do. It is a different product with a different purpose, and it should be planned and resourced as one.
The test we would put on every row of your calendar before you write a word:
If this page ranked at position three tomorrow, would an enquiry land in my inbox?
If the honest answer is no, the piece can still be published, but it belongs in the awareness bucket and it does not get cornerstone effort or cornerstone word count. Save that for the questions with both volume and intent. How much an MVP actually costs is a question people ask immediately before they spend money. A surprising amount of the volume in any keyword export has one of those two properties without the other.
Service and location pages are a different job entirely
If you sell in specific places, or to specific verticals, your research will surface a pile of location and service terms with difficulty scores low enough to look like free traffic. City pages, framework pages, industry pages. On paper these are the easiest wins on the whole plan.
In practice they behave nothing like editorial content, and the common error is writing them as though they do. A location page competes on proximity signals, business listings and trust markers. It is not competing to be a better essay than the other results, so writing one to editorial standards with a strong lede and a considered argument is effort spent on an axis the ranking does not reward. What those pages need is specificity about the work, real proof, and consistency with the rest of your local presence.
Put them on a separate track with a separate quality bar and a separate success measure. A service page succeeds if it converts the people who land on it. A cornerstone succeeds if it earns links and answers the question better than whoever is currently ranking. Holding both to one standard produces pages that are over-written for one job and under-equipped for the other.
There is a second-order cost that matters more than the pages themselves. Because service pages sit on the same calendar as editorial, they eat slots that could have gone to the supporting posts inside the clusters running thin. The real price of the mistake is not the location pages, which are usually fine. It is the cornerstones that never got the support they needed because the calendar was busy.
Internal linking is the compounding asset, and it needs a system
This is the piece of advice founders are most likely to skip and most likely to regret skipping.
Every post you publish is a chance to send authority somewhere it matters. In the early months that barely registers. By month six the difference between a set of individually optimised posts and a genuinely connected site is large, and it is almost entirely down to whether anyone did the linking deliberately.
The problem is chronological. A post written in June can only link to what existed in June, so your earliest pieces accumulate links by being early rather than by being the best destination for the reader. Left alone, that skew hardens into your site architecture.
The fix costs an hour a month. After every four weeks of publishing, go back over the previous month’s posts and add links to anything published since. That single hour converts a pile of posts into something with a shape, and it is the highest-return hour in the whole process.
While you are there, be disciplined about anchor text. Early posts tend to link with descriptive phrases that read naturally and carry no signal at all. Using the target page’s keyword as the anchor is better, and it is the kind of habit that takes about six weeks to become automatic. A piece on what an app build actually costs should be linked as exactly that, not as “read more here”.
A rhythm you can hold beats a clever idea
Fix the slots and fix the types. Strategy piece on Monday, tactical how-to on Tuesday, comparison on Thursday, opinion or field notes on Friday, or whatever shape suits your business. The specific arrangement matters far less than the fact that it does not change.
The mechanism is worth understanding, because it is the reason this works when motivation does not. When the slot and the type are both fixed, the weekly decision is no longer “what should we write”, which is an enormous question that consumes strategic energy every single time you ask it. It becomes “what goes in the Thursday comparison slot”, which is small, answerable in five minutes, and almost impossible to procrastinate on.
The opinion slot in particular earns its place. A recurring slot that demands a genuine point of view forces you to have one on a schedule, and those are reliably the posts that sound like a person rather than a content operation. Our piece on why founders arrive with a Figma and no users exists because a slot demanded it, and it has done more for us than several more carefully optimised pages.
The cost of a rhythm is that it will cheerfully march you off a cliff. A fixed calendar does not notice that the last four weeks produced pages nobody will search. Consistency keeps you producing, it does not keep you correct, which is why the quarterly review below is not optional.
Publish real numbers or publish none
The highest-effort posts you will write are the ones with actual money in them: what things cost, what you charge, what a range really looks like. They are also the ones that work, because they answer the question the reader came with instead of circling it.
Publishing figures a competitor can read is a real trade and we would make it again. Two qualifications, though, learned the expensive way.
First, every number in a post is a maintenance liability. Rates change, thresholds change, and a cost breakdown published in June is quietly wrong by the following April. Tag the posts that contain figures with an expiry so your refresh cycle can target them rather than sweeping everything once a year and missing the ones that matter.
Second, never invent one. Every regulatory, tax or pricing figure needs a link to the primary source, and if you cannot find the source you do not publish the number. This applies to your own performance data too. Ninety-day ranking figures on a young domain are noise dressed up as insight, and publishing them as if they meant something undermines every other number on your site. Our post on the £25k MVP myth works precisely because the ranges in it are real.
If you are considering a second market
Do it as a separate editorial line or do not do it.
Search behaviour genuinely diverges across markets in ways that translation cannot bridge. Vocabulary differs, the size and type of business doing the searching differs, and the local infrastructure your content has to cover differs. A payments piece written for one country will be missing the rails that matter in another. A role that is common in one market may barely exist under that name in the next.
The consequence is that the research does not transfer. Every cluster has to be built twice, from separate keyword sets, with separate competitor analysis, and the pieces cannot borrow structure from each other without quietly becoming translations. That is not an argument against a second market. It is an argument against assuming it costs a translation budget when it costs a second content operation.
Where to start
If you take one thing from this, take the subtraction. Almost every early content plan we see is too broad rather than too shallow, and the fix is removing rows rather than adding them.
Concretely, for your next quarter: choose three clusters and refuse to write outside them. Apply the position-three test to every row before it goes on the calendar. Book the monthly linking hour now, while it still feels unnecessary. Separate service pages from editorial onto their own track. And set a rhythm modest enough that you will still be holding it in week eleven, because consistency beats every clever idea you will have about content in your first year.
The strategy work is not the hard part. Holding the line on what you decided not to publish is.
If you are weighing up whether to build this kind of engine in-house or get the product built while you do the writing, send us a project brief and we will give you a straight answer about which side of that line you are on.
.webp)
.webp)
