Evidence over confidence. Every time.

For twenty years the job was to rank. Someone typed a phrase into a box, ten blue links appeared, and the winner was the site that had done the most to deserve the position. That is no longer the whole picture, and pretending otherwise is how marketing budgets quietly stop working.

A growing share of decisions now begin with a question put to an assistant, which answers in a paragraph rather than a list. The customer does not compare ten results. They read one answer, believe it, and ask a follow-up. If your business is not in that answer, you were not in the consideration at all, and no ranking report will tell you so.

That shift is the reason this page exists. Intelligence work here is about being present at the moment of decision, whether that moment happens on a results page or inside a conversation with a machine.

Answer engines do not read a page the way a person does. They look for an entity they can identify, claims they can attribute, structure they can parse, and consistency across the places your business appears. A business that describes itself three different ways in three different places is harder to recommend than one that says the same true thing everywhere.

So the work starts with the plain facts: what the business is, what it sells, where it operates, who it serves, what it costs, what it refuses to do. Written once, in language a customer would use, and repeated consistently across the site, the profiles, the directories and the review platforms that already describe you.

Then the structure: real headings, a clear page hierarchy, schema markup that states the entity and the offer, pages that answer one question properly instead of ten questions vaguely. An assistant can quote a page that answers a question. It cannot quote a page that circles one.

Then the evidence: the citations, the mentions, the third-party pages that reference the business without being asked. That part cannot be manufactured quickly, which is exactly why it is worth doing.

We call this GEO, for generative engine optimisation, because the industry has settled on the term. In a meeting we would simply say: making sure the machine can find you, understand you and repeat you accurately.

There is a particular kind of writing that gets picked up by both search engines and people: specific, well-structured, and unafraid of a plain answer. A page that opens by defining the thing, then answers the three questions a buyer actually asks, then stops, will outperform a page that warms up for six hundred words.

That means writing the answer into the page rather than hiding it behind a form, naming prices where prices can be named, and stating limits where limits exist. A restaurant that says what it does not serve, a studio that says what it does not promise, a shop that says how long delivery takes: all of it is quotable, and all of it builds the kind of trust that survives a recommendation from a machine.

Structure carries the same weight as the sentence. One idea per section, a heading that says what the section contains, a short summary near the top for the reader who will not scroll. These are old habits dressed in new language, and they work for the same reason they always did: clarity is cheaper to consume than ambiguity.

Rankings still matter, and the discipline behind them has not changed as much as the noise suggests. Crawlable pages, fast pages, internal links that make sense, content that answers a real query, and a site that does not break when it is updated.

We work in that order, because the order is the reason the work holds. Technical foundations first: indexing, canonicals, redirects, the sitemap, the pages that return errors. Then the content: the pages that should exist, the pages that should be merged, the pages that should be deleted. Then the links and mentions that earn authority rather than buying it.

Local search gets its own attention, because it is where many small businesses actually win. A correct business profile, consistent opening hours, photographs that show the room, reviews answered by a person, a map pin that lands on the door rather than the street. None of it is glamorous. All of it decides whether a nearby customer finds you on a phone at seven in the evening.

What we will not do is produce forty articles a month and call it a strategy. Volume without intent produces pages that rank for questions nobody asks and convert nobody who reads them. A smaller set of pages that answer the questions your customers actually type is worth more, and it survives the next algorithm change because it was written for a person.

Paid social is the fastest way to find out whether a promise works. It is also the fastest way to spend money proving nothing, because the platform will happily optimise toward whatever you tell it to, including the wrong thing.

We start from the one number that decides whether the campaign is worth running: what a customer is worth, and what you are willing to pay to acquire one. Without that, a good return on ad spend can still be a loss, and a bad one can still be a bargain on the right margin.

Then the structure: a small number of campaigns, each with a clear job. Cold audiences get the argument. Warm audiences get the proof. Existing customers get the reason to return. Every campaign carries creative that was made for the feed rather than resized from a website banner, because the first frame is the whole pitch.

We test one variable at a time. Same audience, same budget, different opening line, and enough time for the result to mean something. Testing five things at once produces a winner nobody can explain and a lesson nobody can repeat.

Measurement is set up before the first pound is spent, not reconstructed afterwards. Pixel and conversions checked against real orders, a landing page that matches the ad it follows, and a weekly look at what the spend actually bought. Attribution is imperfect everywhere; the answer is not to pretend otherwise but to keep the gap visible and small.

Good decisions are usually the result of somebody having read more than the room. Research here is not a report for its own sake. It is the specific piece of context that changes what we build: what the category says, what customers complain about, what the competitor’s site actually does when you try to buy something, what the pricing page hides.

We do the reading, the checking and the summarising, and then we write a brief short enough to act on. Two pages, with the finding, the implication and the recommendation. If a research document cannot end in a decision, it was a reading list.

Most businesses find out what a competitor did when a customer mentions it. By then the price has been matched, the campaign has run and the advantage is gone.

We set up monitoring that watches the things that matter and reports only when something changes: competitor pricing on named products, new pages appearing on a rival site, a listing moving position, a review score slipping, a stock status flipping to out of stock. The report is short by design. A weekly email nobody reads is not intelligence, it is noise with a subject line.

The same discipline applies inward. Rank tracking, traffic by page, which queries bring people who stay, which pages get visited once and never again. The useful question is rarely how many people came. It is what they did when they arrived.

It begins with an audit, and the audit is honest about what it finds. Where the site currently appears, which pages bring traffic, which queries convert, how the business is described across the web, what competitors are doing that works, and what is broken in a way that no amount of content will fix.

From that we write a short plan with a sequence. Foundations, then content, then authority, then paid support if paid is the right tool. Each stage has a date and a measure, and each stage can be stopped without wasting the one before it.

The first month is usually unglamorous: fixing what is broken, capturing the baseline, writing the pages that should have existed already. Clients who expect a launch moment are often surprised. Clients who have done this before are usually relieved.

Every engagement starts with a small set of measures agreed in writing, and a baseline captured before anything changes. Without a baseline, a month of work can be presented as a success by choosing the right chart.

Reporting is monthly and plain: what changed, what we did, what worked, what did not, and what we are doing next. We include the numbers that went the wrong way. A report that only moves in one direction is a report nobody should trust.

Where a result cannot be attributed honestly, we say so. Search and social both benefit from work done elsewhere: a product people want, a price that undercuts the market, a founder with a following. Taking credit for all of it is the oldest trick in the discipline.