Ranking methodology
SEO agency ranking methodology
Follow the evidence behind SEO agency comparisons. Understand source provenance, project context and RankScore so you can build a better-informed shortlist. SEOS.co publishes curated editorial agency comparisons to help buyers assess fit, examine supporting information and identify what remains unverified. This methodology connects the comparison with the practical questions to resolve before an engagement.
Start with the question the comparison answers
An agency can be relevant to a market without being the right choice for every business in that market. A local service company, a software business and a retailer may all need SEO, but their customers, websites and implementation needs differ. A comparison is useful when its context resembles the decision you are making and its evidence addresses the kind of work you need.
Read the page-specific criteria and limits before treating a position as a recommendation for your own project. Ask which business model, location or specialty the page concerns. Serving a location does not necessarily mean maintaining an office there. Experience with one kind of website does not automatically establish experience with a different platform, buying process or operational constraint.
The framework here separates editorial placement, agency-reported information, third-party reviews and measured outcomes. These are related inputs, but they are not interchangeable. Keeping them distinct helps you judge whether a claim supports a narrow observation, a useful hypothesis for a sales conversation or a conclusion strong enough to influence your shortlist.
Read RankScore as an editorial signal
RankScore is an editorial comparison score. It is not a customer star rating, a Google ranking factor, a probability that an engagement will succeed or an estimate of future return on investment. The number belongs to the editorial comparison in which it appears. It should be read with that page’s criteria and supporting information rather than exported as a universal measure of agency quality.
Read numerical precision alongside the evidence behind the comparison. A displayed difference such as 98.8 versus 98.6 does not establish a scientifically measured difference in quality. A fully reproducible numerical calculation would require the underlying observations and component scores. Page-specific criteria explain the editorial assessment; use those criteria and supporting information to understand what the score can establish.
Use the score to orient yourself within the comparison, then inspect fit and evidence. An agency’s implementation model, relevant experience and willingness to explain its assumptions may be more important to your project than a small score difference. Ask what the comparison establishes, what remains unverified and what you need to confirm directly before entering an agreement.
Follow a claim back to its origin
Provenance means knowing where information came from and what that source can establish. An official agency page is an appropriate source for the agency’s stated services, published pricing, team information and account of its work. It is also an interested source. A result described there remains agency-reported unless a relevant part of the claim is corroborated independently.
A third-party review describes a client’s perspective within the platform’s process. Look at the platform, date, sample size, service purchased and what its verification label actually covers. A verified label does not mean SEOS.co interviewed that client. Verification of an identity, verification of a commercial relationship and independent measurement of a result are separate propositions.
Client confirmation or independent evidence can strengthen a specific statement when it concerns the same project, period and outcome. Check that connection. A client confirming a relationship does not necessarily confirm every performance number in a case study. A report naming the agency does not necessarily establish which team delivered a particular intervention or which channel produced the reported revenue.
Other editorial lists and public discussions can help identify candidates and buyer concerns. They do not independently demonstrate delivery quality simply because they repeat a name. Several pages reproducing one press release remain one originating account of the result. The useful question is whether another source adds genuinely independent information about the claim being considered.
Match evidence to the work you need
Local and multi-location SEO
Look for comparable location counts, business profile work and reporting that distinguishes qualified calls or enquiries from general activity. Ask who controls profiles, service areas, duplicate listings and implementation. A business with one location and a business coordinating many locations can face different approval, ownership and reporting problems even when both describe their need as local SEO.
SaaS and business-to-business search
Look for named software clients, relevant buying motions and a clear account of trial, demo or pipeline attribution. Ask how the provider learns the product, obtains expert input, builds pages and connects analytics with the CRM where needed. General traffic growth is not the same evidence as progress among the people who can buy a particular business product.
Ecommerce search
Consider platform, catalog complexity, migrations and the distinction between organic revenue, transactions and total business revenue. Ask who ships changes to the store and how implementation is coordinated. Results combining paid media, marketplaces and organic search should not be presented as if every part came from SEO. The context behind a commercial number matters as much as the number itself.
Specialist and regulated work
Look for relevant sector work and an explicit account of content review and implementation responsibilities. Ask who approves professional claims and who manages the client’s compliance requirements. Experience in a specialist market can be relevant, but it should not be inflated into a promise that the provider assumes responsibilities it has not agreed to carry.
Make performance claims comparable
A useful result identifies the client or business context, baseline, measurement period, metric, attribution and work delivered. Without these elements, an impressive number may answer a different question from yours. A percentage change with no starting value can exaggerate the apparent scale. A revenue figure with no channel definition can obscure whether organic search contributed to it at all.
Traffic, leads, revenue and profit measure different things. More visits do not automatically mean more qualified enquiries. More forms do not necessarily mean more suitable customers. Revenue can change because of pricing, promotions, paid marketing or other activity alongside SEO. Attribution should explain the connection being claimed rather than letting readers assume that one intervention caused the entire result.
Consider a hypothetical comparison between two case studies. One reports a large traffic increase without a period or baseline. The other reports a smaller increase with dated observations, relevant landing pages and an explanation of the work. The smaller figure may be more useful for due diligence because you can understand its scope. This example illustrates evidence quality; it is not a result attributed to an agency.
Case studies are selected examples, not forecasts for a new client. Missing evidence means a claim has not been established from the available material; it does not prove the agency cannot perform the work. Ask for the original report or a relevant reference when the point is material to your decision. Keep the distinction between absent information and negative evidence explicit.
Compare scope before comparing price
Hourly rates, minimum project prices, monthly retainers and per-location fees are different units. Convert proposals into a common view of responsibilities before trying to compare cost. Identify whether strategy, implementation, content, development, software, digital PR and reporting are included. Ask what the client must supply and which activities would require a separate agreement or additional budget.
Published prices can change and may exclude media or other costs. Confirm current terms with the provider. A lower price can describe a narrower service, while a broader proposal can include work another candidate expects your team to deliver. Neither arrangement is inherently suitable without reference to your resources and requirements. The point is to make those differences visible before a decision.
Include account access and ownership in the comparison. Ask who controls analytics, business profiles, content, reports and other relevant assets, and what happens when the engagement ends. An attractive proposal is incomplete if important delivery assumptions remain unspoken. Record exclusions, dependencies and exit terms alongside the price so that the comparison reflects the engagement rather than just its headline cost.
Use confidence without inventing certainty
Confidence should reflect how well the available evidence supports a particular claim. An attributed service description, a relevant client account and independent project evidence can support different degrees of certainty about different facts. They should not be collapsed into a claim that every aspect of an agency has been verified. A strong source for one statement is not automatically a strong source for another.
For your own shortlist, keep a simple evidence note for each material question: the claim, its origin, the relevant date, the scope it covers and what remains unresolved. Connect each question with the source that can answer it. This practical reader tool makes follow-up more precise when proposals sound similar and keeps your assessment separate from the publisher’s editorial score.
Sources can be promotional, incomplete, outdated or in conflict. The comparisons do not constitute audits of private analytics, finances or client accounts. Confirm material facts directly. A visible update date does not prove every source on a page has been rechecked. A count of linked domains does not equal verified client outcomes, and a displayed agency count does not establish how many candidates were initially considered.
Treat AI search evidence as conditional
AI search experiences can vary with the platform, prompt, location, date and session. A mention is different from a citation, and neither alone establishes a qualified lead. An isolated screenshot records one observed answer under particular conditions. It does not establish a dependable exposure level or a commercial outcome for another business.
When an agency presents AI search work, ask for the prompts, platforms, observation dates, cited URLs and limitations behind the claim. Ask what was repeated and what changed between observations. Connect visibility reporting with relevant business outcomes where evidence permits, without assuming that every change in leads came from an answer engine. The measurement should describe what it can actually observe.
Evaluate the proposed work through its implementation requirements and measurement plan. Ask which pages will change, who will make those changes and which observations will be repeated. A useful proposal explains why the work fits your circumstances and how its effects will be assessed. Technical terminology earns its place when it makes those decisions more precise. No comparison guarantees inclusion in an AI answer or a particular Google position.
Build a shortlist you can interrogate
Write one brief covering your market, problem, budget, internal resources and desired outcome. Give each candidate the same relevant pages, locations or examples. Ask for a comparable case, the period it covers and the work the provider delivered. This creates a shared basis for comparison instead of rewarding whichever proposal uses the most persuasive language.
Then ask who implements, who approves and who owns the accounts, content and reporting. Discuss assumptions, exclusions, risks and exit terms. If the provider relies on your development team, make that dependency explicit. If a performance claim is central to the proposal, ask what evidence would let you distinguish the provider’s contribution from seasonality, brand demand or other marketing activity.
Do not treat every unresolved question as a disqualification. Some information may require a direct conversation or a relevant reference. What matters is whether uncertainty is acknowledged and whether the missing information is important to your decision. A shortlist should become more specific as you learn, rather than appearing settled because a page assigns a numerical position.
Keep an evidence trail for your decision
For a material claim, save the source URL and the date you checked it alongside your notes. Record the exact metric and period rather than paraphrasing every result as growth. If a provider clarifies that a case covers paid and organic channels together, retain that qualification when you compare it with another case. Removing the context can make individually accurate statements misleading in combination.
Separate your own assessment from the source’s statement. You might judge a project relevant because it uses the same platform, while still having no evidence about the provider’s current staffing or availability. Both observations can coexist. This discipline makes it easier to revisit a decision when facts change and to explain why one candidate was a better fit for a particular brief without claiming universal superiority.
Corrections, attribution and changing information
For a factual correction, provide the affected URL, exact statement, a primary source and the date of any known change. This makes the issue inspectable. A correction request is different from a request for a preferred editorial position. The corrections policy explains the route, while review standards and the editorial policy provide further context about evidence and attribution.
Services, people, pricing, platforms and public evidence can change. Confirm the facts that materially affect your engagement before signing. Reading a comparison carefully is not the last step in due diligence; it is a way to make the next step better informed. Use editorial guidance to find relevant questions and candidates, then test them against the circumstances of your own project.
Questions about the methodology
Does a high RankScore predict my return?
No. RankScore is an editorial comparison score, not an ROI forecast or probability of success. Your outcome depends on the work, implementation, market and other circumstances. Ask for relevant evidence and a clear scope rather than treating the displayed number as a prediction of a future engagement.
Does a third-party verified review establish the reported result?
Not necessarily. Read what the platform verifies and what the reviewer describes. Identity, commercial relationship and measurement of performance are distinct. An attributed client perspective can be useful without proving every result associated with the project. Keep the label and the underlying claim in their proper scope.
How should I interpret small differences in RankScore?
Read the number alongside the specific comparison’s criteria, supporting information and limits. Small displayed differences do not establish scientifically measured differences in agency quality. A reproducible calculation would require the underlying observations and component scores. Use the comparison to identify relevant candidates and questions, then assess fit against your own implementation requirements.