Most early-stage positioning is decided by whoever spoke last in the meeting. A founder writes about the thing they find interesting, an advisor suggests a topic, a competitor publishes something and the queue is rearranged. Six months later there are thirty pages, no pattern, and no way to say which of them did anything.
There is a cheaper method, and it does not require a budget or an agency. Decide what to publish with an explicit scoring rule, publish in that order, measure what each item produces, and let the measurements reorder the queue. That is all "algorithmic positioning" means. It is not artificial intelligence writing your copy, and anyone selling it as that is selling you a volume problem you do not have.
This article is for the founder of a two-person software startup, the partner in a 12-person engineering consultancy, or anyone running a new project without a marketing team. It covers the scoring rule, where the numbers come from for free, why order matters more than volume, what AI assistants change, what to measure, where the method fails, and what it costs to run.
What does an algorithmic approach to positioning actually mean?
It means the decision about what to publish next is made by a written rule applied to measured inputs, not by whoever feels strongest about it. You list every question a buyer might type, attach a demand figure and a difficulty figure to each, score them, and work down the list. When a published item produces enquiries, its neighbours move up. When it produces nothing after a fair window, its neighbours move down.
The rule is deliberately simple, because a rule nobody can compute in their head gets abandoned in the first busy week.
The scoring rule
Four inputs, all obtainable without paid tooling in a first pass.
Demand. How many people search this phrasing each month. Approximate is fine; what matters is the order of magnitude, not the decimal.
Answer quality of what already ranks. Read the top five results. If they are thin, generic, or written for a different company size than your buyer, this is an opening. If three of them are excellent and specific, it is not, whatever the volume says.
Distance from what you sell. A page can rank beautifully and attract nobody who could ever buy. Score honestly: can a reader of this page become a customer in one step, two, or never.
Cost to write it well. Some pages need a specialist's afternoon. Some need a specialist's week. In a small company this is the real constraint and it is almost never in the spreadsheet.
Score is demand, multiplied by the opening, divided by distance and cost. Any consistent weighting works. The value is in applying the same one every time, not in the arithmetic being clever.
A worked example. Take a 4-person startup selling scheduling software to physiotherapy practices, with a first list of 60 questions. Half are about features nobody searches for. Of the rest, "how to reduce no-shows in a physio practice" has modest demand, weak results written for large hospitals, and a reader one step from a trial. It goes to position one. "What is practice management software" has higher demand, ten strong results and readers who are years from buying. It goes near the bottom, whatever the volume says.
Where do the input numbers come from without paid tools?
From the search engine itself, your own site data and your sales conversations. Search suggestions and related questions show phrasing. Once a site exists, Google's free Search Console reports how often each page appeared in results and how often it was clicked. Your inbox and call notes show the words buyers actually use. Approximate numbers applied consistently beat precise numbers applied once.
Google's Search Console Performance report defines the two numbers that matter most: impressions, the number of times your site appeared in results, and clicks, the number of times someone clicked through. Rising impressions on a page with few clicks tells you the demand exists and the title or snippet is not answering it. That is a rewrite, not a new page.
It is also worth remembering that search is where research happens. In the Federal Statistical Office's 2025 survey of internet use, 85 percent of residents aged 15 to 88 had searched online for information about goods or services in the previous three months. Your buyer will look you up. The question is what they find.
Why the ordering matters more than the volume
A new domain has no history, and the pages you publish first shape what search engines and AI assistants conclude the site is about. Twelve pages tightly clustered around one problem establish a subject; forty scattered pages establish nothing and compete with each other for the same weak signal.
Google's own guidance on helpful, reliable content lists warning signs that read like a description of the scattered approach: producing lots of content on many different topics in the hope that some of it performs, and using extensive automation to produce content on many topics. It also says its systems give more weight to experience, expertise and trust for topics that can affect health or finances, which is why the effect is stronger for a project in either field. The sector-specific versions are in positioning a new fintech brand and positioning a new medical practice.
This is also why the method survives contact with a two-person team. Twelve items, in order, with a rule for what comes next, is a plan a founder can execute between other work.
It is worth remembering who else is competing for that attention. The Federal Statistical Office counted 561,952 micro-enterprises with fewer than ten employees in 2023, roughly nine in ten market enterprises in Switzerland, and more than 99 percent are SMEs. Almost every company you compete with for a search result is also small, also without a marketing team, and also publishing whatever somebody felt like. Consistency of method is an advantage precisely because it is rare.
Do AI assistants change the method?
They raise the stakes without changing the rule. A growing share of people use AI assistants, and an assistant can only quote a page that answers one question clearly and specifically. That rewards a small set of narrow pages chosen by measured demand, and it punishes the shortcut of generating hundreds of thin pages, which Google's spam policies now name explicitly.
The adoption is real. The Federal Statistical Office reported that in spring 2025, 43 percent of residents aged 15 to 88 had used a generative AI system, and 79 percent of those aged 15 to 24. That does not mean search has gone. It means your pages are increasingly read by a machine on behalf of a person, and a page that answers a question in its first paragraph is the one that gets quoted.
The search engines say the same thing in plain words. Google's documentation on AI features states that there are no additional requirements to appear in AI Overviews or AI Mode: a page must be indexed and eligible to show with a snippet, and the usual SEO practices apply. Its spam policies name "using generative AI tools or other similar tools to generate many pages without adding value for users" as an example of scaled content abuse. The scoring rule keeps you on the right side of both: few pages, each worth reading.
What to measure, and what most people measure instead
The default dashboard measures the things that are easy to collect, which are almost never the things that decide anything.
| Commonly tracked | What it actually tells you | Track instead |
|---|---|---|
| Page views | How many people arrived | Enquiries per page, per month |
| Bounce rate | Very little on an informational page | Whether the next page was the service page |
| Followers | Nothing about pipeline | Enquiries mentioning a specific page |
| Keyword rank | Position, not demand captured | Impressions and clicks for that phrasing |
| Time on page | Reading speed, mostly | Whether an enquiry followed within the session |
The right-hand column requires that conversions are actually instrumented. If a form submits and fires no event, none of it is measurable and the whole method degrades to guessing with extra steps. That is worth fixing before anything else; we describe the general version of the problem in measuring the ROI of digital transformation.
Where the algorithm stops working
Three places, and being honest about them is what separates a method from a pitch.
It cannot invent expertise. The rule tells you which question to answer. It cannot answer it. If nobody in the company knows something a generic article does not, the correct output of the process is to publish nothing and go and learn something instead.
It is blind to demand that does not exist yet. Scoring by search volume finds problems people already know they have. A genuinely new category has no volume by definition, and for those the channel is direct conversation, not publishing. Most early-stage companies are not in this position, but the ones that are will waste a year if they follow the rule anyway.
It has a floor on time. A new domain publishing well typically needs several months before search results respond, and no amount of optimisation compresses that to weeks. Google's guidance on hiring an SEO is blunt that no one can guarantee a first-place ranking. Any plan that depends on revenue from search inside a quarter will be abandoned in month two.
How long before an early-stage project sees results from this?
Direct outreach produces conversations within weeks. Published pages take longer: in the projects we have observed, a new domain publishing consistently starts to see meaningful search-driven enquiries somewhere between three and nine months, depending on how competitive the subject is and how specific the pages are. Founders who plan both timelines at once stay funded through the gap.
Google's description of how Search works explains part of the floor. Pages have to be crawled, then indexed, then judged relevant to a query, and Google states that it does not guarantee any of the three and does not accept payment to crawl a site more often or rank it higher. The time is spent earning a place in the index, not waiting for a queue.
What this costs to run
A first pass takes about a week: enumerate the questions, score them, write the twelve-item queue. After that it is one afternoon a month to re-score against what the measurements say, plus whatever the writing itself takes. These are typical observations from the projects we have run, not a quote.
There is no software to buy. A spreadsheet is sufficient and is easier to argue with than a tool, which matters, because the argument about whether an item really belongs at position three is where most of the thinking happens.
What changes with size is who does the writing. In a two-person startup the founder writes, and the queue should be short enough to finish. In a 20-person company a product or delivery lead can draft, but one person still owns the queue and the re-scoring, or the method quietly reverts to whoever spoke last. Whether that person should be a founder, an adviser or a first hire is the question in fractional CTO, consultant or first technical hire.
If any part of this ends up automated, it should be the scoring and the reporting, not the judgement, the same boundary we draw for AI adoption beyond pilots. Where machine learning genuinely helps is in clustering thousands of query variants into the dozen real questions underneath them, which is our AI and machine learning work applied to a small problem, and it is worth doing only once the manual version is already running. The candidate uses of generative AI in drafting are weighed in which generative AI use cases pay back for a small company. For technology and startups specifically, the manual version is usually enough for the first two years.
Sources
- Federal Statistical Office (BFS) — Kleine und mittlere Unternehmen, enterprises by size class, 2023
- Federal Statistical Office (BFS) — Internetnutzung in den Schweizer Haushalten 2025, press release of 5 December 2025 on generative AI use
- Federal Statistical Office (BFS) — Individuals and online activities, STAT-TAB, Omnibus ICT survey
- Google Search Central — Creating helpful, reliable, people-first content, warning signs and E-E-A-T
- Google Search Central — AI features and your website, eligibility for AI Overviews and AI Mode
- Google Search Central — Spam policies for Google web search, scaled content abuse
- Google Search Central — How Search works, crawling, indexing and serving
- Google Search Central — Do you need an SEO?, no guaranteed rankings
- Google Search Console Help — Performance report, definitions of clicks and impressions
Related reading
- How to measure ROI on a digital transformation before you fund it
- How does a new fintech brand get taken seriously in Switzerland?
- How does a new medical practice get found in Switzerland?
- Why AI pilots stall in Swiss companies, and what gets past that
- AI and machine learning consulting
Working on this right now?
Tell us where you are in two questions. A consulting partner reviews every enquiry within 2 business days.
