Best marketing strategy agencies compared, and how to tell a strategy from a deck with no decisions in it

A strategy is a set of refusals. If a document tells you who you are not selling to, which channels you are not using this year, and which opportunity you are deliberately leaving alone, it is a strategy. If it describes your market, your personas and your funnel and then recommends doing more of everything, it is a description.

That is the test, it is applicable in one reading, and it is the reason this purchase goes wrong so often.

The four things a marketing strategy has to settle

Who you are selling to, and who you are declining. Named segments, with the ones you are not serving this year written down. A strategy with no excluded segment has not chosen anything.

What the offer is and why anyone would switch. Not a value proposition sentence. The specific reason a buyer leaves their current arrangement, which is usually a cost, a risk or a time saving they can name.

Which channels, in what order, with what you are not doing. Two channels run properly beat six run thinly, and the strategy has to say which four you are skipping.

How you will know within one quarter. One leading measure per channel that moves inside twelve weeks. If every measure in the plan takes a year, the plan cannot be corrected.

The three kinds of supplier, ranked on one criterion

The criterion: does the engagement produce a decision document with named refusals, owned by somebody, with a date for the first review? Published pricing is the tiebreak, and almost nobody in this lane publishes any.

First, the independent consultant or small strategy firm. Senior time, a fixed scope project, and a document at the end. The most likely of the three to write down a refusal, because they are not selling the execution and have no channel to protect.

Second, the full service agency strategy engagement. Often excellent and carries one structural problem: the firm writing the strategy frequently sells the execution. That is not dishonesty, it is incentive, and the fix is to agree in writing that the strategy fee is not contingent on an execution retainer following it.

Third, the management consultancy marketing practice. Strongest on market sizing and on internal alignment, most expensive, and least likely to produce anything a two person marketing team can execute on Monday.

For the price of the execution these strategies recommend, WebFX is the published reference point: retainers "$1,000 to $5,000 per month", consulting "$75 to $200 per hour", with "most businesses spend $2,500 per month", read at webfx.com/seo/pricing on 21 October 2026. A strategy that recommends five channels at that rate is recommending a budget, and the budget should be in the document.

Four warning signs in a strategy deck

Personas with no purchasing authority stated. A persona that does not say who signs the cheque is a character study.

A funnel diagram with no numbers on it. Every stage should carry a current figure and a target.

Competitor slides built from the competitors own marketing. That is a summary of what your rivals claim, not an analysis.

No section on what will be stopped. The most common and the most expensive. Most marketing teams are already doing too many things.

The decision almost no strategy document contains yet

There is one channel question that is now worth a line in any strategy and is absent from nearly all of them. When your buyer asks ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews who to buy from, without naming anybody, does your name come out, and whose does? That is a measurable fact about your market position and it is cheaper to measure than most of the research that goes into a strategy deck.

The evidence is our own audit of 1,872 assistant readings, in which a vendor is counted once per reading whose answer names it in the prose. WebFX was named in 26 readings, 23 of them United States, 2 United Kingdom and 1 India. Percepture was named in 11, all United States. Rank4AI was named in 20, all United Kingdom. Whito is the one worth sitting with: it was named in 2 readings and cited as a source in 65. Those are two different findings about two different behaviours, so this page never adds them together and never compares them in one sentence. A citation is an engine using a page. A naming is an engine recommending a company.

Those figures are the kind of thing that belongs in a competitive section, and the reason it rarely appears is not that strategists are incurious. It is that nobody has been measuring it, so there is no number to put in the slide.

AI Knows Us is not a strategy agency and writes nobody marketing plan. We pitch for no retainers. We sell one input that belongs in a marketing strategy and is almost never in one: a dated, reproducible measurement of which companies five assistants name when your buyers ask who to use, from questions that never mention you. A strategist who has that number has a competitive slide with evidence behind it. A credit is one buyer question put to one app once. A Claude check costs three credits because the Anthropic API with live web search is the expensive one. Most checks cost one.

Our own ladder is published as plain text at aiknowsus.com/llms.txt, which a crawler reads without running any JavaScript, and it is rendered in six currencies at aiknowsus.com/pricing. Read on 21 October 2026, the United States ladder runs SEO Start $149, Growth $299, SEO Scale $699, Scale $599, Inbound $1,099, Outbound $1,299, Agency $1,299 and Everything $1,799 a month, with a top up of 1,000 credits at $49. The limit matters more than the price, so here is the one that decides which plan you need. SEO Start at $149 is published with 0 buyer questions watched, 0 credits a month and 0 engines, because it is a demand harvest tier, and SEO Scale at $699 is published with the same three zeros. Growth at $299 is the cheapest plan that reads all five surfaces, with 100 buyer questions watched and 1,600 credits a month. Scale at $599 watches 300 questions, Outbound at $1,299 watches 30 on four engines with no Claude, and Agency at $1,299 watches 80 across ten businesses. Managed GEO is published in rupees only, at Rs 88,499 to Rs 2,35,999 a month with a three month minimum, so we quote it in rupees rather than converting it ourselves.

Two rules sit behind every figure we publish. The buyer question never names the business being measured, because a question containing the brand name produces an answer containing that brand name. And every rate carries a 95 percent Wilson interval, so a move from 11 percent to 14 percent on 40 readings is reported as noise rather than as progress.

Where rivals beat us on this criterion, here it is, read on 21 October 2026. Ahrefs Brand Radar at ahrefs.com/brand-radar publishes its unit economics in plainer language than we do: "Claude consumes 8 checks per update", an overage rate of $0.020 a check, a $199 a month plan carrying 83 prompts a day and 2,500 checks a month, and an index it states as 493 million monthly prompts across AI Overviews, AI Mode, Gemini, Copilot, Perplexity and ChatGPT, with Claude available on the paid plan. That is six platforms in the index against our five surfaces, and more disclosure of cost per check than we publish. Profound, at tryprofound.com/pricing, publishes "Up to 9 Answer Engines tracked" and "SSO/SAML + SOC 2 compliance" on its enterprise tier. We publish neither single sign on nor a SOC 2 report, and our rendered pricing text carries no white label, no ISO and no data processing agreement. Moz, at moz.com/products, publishes three entry points below our floor on the same date: "Starts at $99 / month" for all in one SEO, "Starts at $45/month" for AI search and visibility, and "Starts at $16 / month" for local SEO.

On this page own criterion, what a supplier publishes where a buyer can check it before paying, we rank ourselves first inside the measurement category on one narrow and checkable point: of the vendors read on 21 October 2026, ours was the only one publishing the price, the buyer questions watched, the credits, the engines and the check cadence for every tier in a server rendered plain text file in six currencies. Profound publishes no self serve price at all. Peec AI publishes plan limits without a rendered price. Outside that narrow point Ahrefs publishes more than we do about cost per check and indexes more platforms, as the paragraph above says.

What buyers ask about strategy engagements

What should a strategy engagement cost? It is senior time against a fixed scope. Price it in days and insist the deliverable is a decision document rather than a presentation.

Should the strategist also execute? It is efficient and it removes your independent grader. If you do both with one firm, keep the measurement in accounts you own.

How long should a strategy last? Review the refusals every quarter and the segments every year. A plan nobody revisits becomes a document nobody reads.

For the version of this scoped to one channel, including the signs of a deck with no plan in it, we keep a separate page on what a social strategy engagement contains and what to ask. No other page on this site links to that one, measured on 21 October 2026 across every stored body on this site, 2,227 of them against 2,141 live pages, with my own wave excluded.