Is Your SEO Company Producing Evidence You Can Verify?

The client growth owner and the agency strategist should review query performance together. Inputs should include the agreed service priorities, Search Console queries and pages, landing page analytics, lead or sales records, and any location limits.

First, classify each gaining query by intent: research, comparison, brand, service, or local action. Next, connect the query to the page that earned visibility and the action available on that page. Then compare that traffic with qualified inquiries, assisted conversions, or another business measure already used by the company.

A law firm ranking for a historical topic may support education, but it cannot be counted as evidence that priority service demand improved unless the page contributes to a documented path. Likewise, a service page gaining impressions without clicks may indicate better discovery but weak presentation, poor fit, or insufficient position.

The decision criterion is alignment, not raw volume. Continue when priority pages are gaining relevant impressions, clicks, and qualified actions in a sequence that fits the market. Correct the plan when most gains come from unrelated topics, unsupported locations, or visitors who cannot use the service.

The tradeoff is that informational content can build context before commercial pages move, so do not reject it automatically. Require the agency to explain its role, expected next step, and measurement.

The output should be an intent map that labels each tracked query and page as core, supporting, irrelevant, or still inconclusive.

The agency delivery lead should maintain the work log, while the client owner verifies access, approvals, and business accuracy. Begin with the scope in the agreement and list every promised recurring or one-time output.

For each item, request evidence at the level needed to review it: changed URLs, before and after metadata, redirect mappings, internal link additions, briefs, drafts, approvals, published dates, outreach targets, and technical tickets.

A dashboard export can summarize this work, but it should not replace the underlying record. Use three decision criteria. First, completion: was the agreed work delivered? Second, quality control: was it reviewed against the site's technical, editorial, legal, or brand requirements?

Third, rationale: can the agency explain why the task was prioritized and what it is intended to change? Some methods can remain proprietary without making delivery invisible. You do not need private templates or vendor negotiations, but you do need enough evidence to confirm what affected your website and reputation.

The tradeoff is administrative overhead. A useful log should be concise enough to maintain, yet specific enough to support rollback, compliance review, and performance analysis. The monthly output should be a signed-off list of completed work, blocked work, owner, reason, and next measurement date.

This review is owned jointly by the agency strategist and the client person responsible for brand facts. Inputs include the website, Google Business Profile where applicable, major industry listings, public organization and person profiles, and a fixed set of search and AI prompts.

Start with factual consistency. Check the business name, address, phone details, services, locations, credentials, and official profiles against the website. Correct contradictions before interpreting visibility.

Then search the brand with priority services and inspect whether search results describe the entity accurately. For Google AI Overviews or other AI features, use repeatable prompts and save the date, wording, market, result, cited sources, and exact classification.

Do not treat a recorded mention as a customer choice or hiring event. The purpose is to observe whether the business is represented, cited, confused with another entity, or missing. A Knowledge Panel or a particular AI response is not a required success condition for every business.

The decision criterion is whether the agency is improving the quality and consistency of accessible evidence across sources that genuinely apply. The tradeoff is control: the agency can improve first-party facts and relevant public references, but it cannot guarantee how a search feature will summarize them.

The output should be a discrepancy register, correction owner, and repeated observation log rather than a promise of inclusion.

The agency technical lead owns diagnosis and implementation guidance; the client development owner controls releases when the agency cannot deploy changes directly. Inputs should include Search Console, crawl data, performance testing, sitemap records, redirect rules, templates, and a list of priority page groups.

Use a simple sequence for every issue: define the affected URLs, explain the business or discovery risk, assign severity, identify the owner, implement the change, and validate the result after release.

Core Web Vitals, indexing states, internal links, mobile behavior, canonicalization, redirects, and crawl access should be interpreted by page type and importance. A growing index is not automatically good, and a smaller index is not automatically bad.

The useful question is whether intended pages are available while duplicate, broken, obsolete, or blocked states are controlled. For a regulated business, technical review should also confirm that important policy, contact, authorship, and disclosure pages remain reachable where applicable.

This is a publishing and governance check, not a claim that those pages create rankings. The tradeoff is prioritization. Fixing every minor warning can consume resources that should go to high-impact templates or conversion paths.

The output should be a technical register showing open risk, completed fix, validation evidence, residual limitation, and the next review date.

The agency editorial lead should own the brief and production record; the client subject matter owner should verify claims, examples, offers, restrictions, and required disclosures. Inputs include the target user question, search results, first-party expertise, approved sources, service details, and conversion path.

An 800 word article can be useful or empty. Judge it by whether it resolves the reader's actual uncertainty, explains tradeoffs, distinguishes facts from opinion, and provides the next appropriate action.

For high-trust topics, require clear authorship and a review process that matches the risk of the subject. Do not use E-E-A-T as a checklist of decorative elements. Author information, first-hand experience, citations, and expert review matter only when they make the page more accurate, transparent, and useful.

Schema should reflect visible content and real relationships; it is not a substitute for those qualities. Compare the published page with the brief and with the business's actual practice. Flag generic paragraphs that could describe any company, unsupported claims, copied structures, missing limitations, and answers that avoid the decision implied by the query.

The tradeoff is speed versus depth. A faster publishing schedule may cover more questions, while deeper review may be necessary for priority or sensitive pages. The output should be an editorial acceptance record with owner, evidence used, review state, update trigger, and expected role in the user journey.

The agency outreach owner should provide the placement record, while the client owner reviews brand fit and any regulated or reputational concerns. Begin with links acquired during the last 90 days and inspect the actual source pages rather than only summary metrics.

For each link, record the publishing site, page topic, placement context, destination page, anchor text, acquisition method, and reason the source was accepted. Check whether the publication has a coherent audience, real editorial content, topical or local relevance, and a plausible reason to reference your page.

Third-party authority and traffic metrics can help triage, but neither proves quality. A lower-scoring industry association, local organization, partner, or specialist publication may be more relevant than a general site with a stronger headline score.

The decision criteria are legitimacy, relevance, editorial fit, natural placement, and risk. Continue when the agency can explain why the link belongs and how it supports a useful asset. Correct or stop the tactic when placements are repetitive, unrelated, hidden in low-quality pages, or produced through an undisclosed network.

The tradeoff is pace. Legitimate editorial links can be uncertain and slower than purchased volume, so evaluate the agency on the quality of prospecting, assets, outreach, refusals, and completed placements rather than a guaranteed count. The output should be a link register with accepted, rejected, questioned, and remediated items.

The agency account lead owns the report, but the client sponsor owns the decision it informs. The report should combine the delivery log, search performance, technical status, content outcomes, link activity, and qualified business actions for the same period.

Automated exports from Semrush or Ahrefs can provide supporting data, but they need interpretation. A previously reported 50% traffic increase is not decision-useful until the report identifies the pages, queries, countries, devices, and actions behind it.

The same applies to ranking gains that occur on terms outside the reachable market or with no clear role in the strategy. Require three answers in every review: what did the agency complete, what evidence changed afterward, and what decision follows?

Add a confidence statement when attribution is uncertain. Search demand, seasonality, site releases, brand activity, competitors, and measurement changes can all affect the result. Use explicit decision criteria.

Continue when contracted outputs are delivered, leading indicators move in the intended direction, risks are controlled, and the next hypothesis is credible. Correct scope when work is active but misaligned.

Escalate when delivery cannot be verified. Prepare to replace the provider when missing work, unexplained risk, or repeated failure to act persists after a documented correction period. The output is a one-page decision summary supported by appendices, not a data dump. It should name owners, deadlines, blockers, and the evidence expected before the next review.

Most evaluations begin with total traffic and a list of keyword positions. Those measures can be useful, but they do not prove that the agency selected the right work, completed it correctly, or improved demand from people who could become customers.

A more useful operating system compares four layers in sequence: contracted inputs, verified outputs, leading search indicators, and business results. Each layer has a different owner and delay. For example, the agency owns delivery of approved page changes; the client owns timely access and subject matter review; search systems determine discovery; the sales team determines whether inquiries are qualified.

Do not force every change into a revenue claim. Instead, require a clear link between the work and the next measurable stage, record the tradeoff, and define what evidence would cause the plan to continue or change.

I used to place too much weight on technical cleanliness because it was easy to inspect and document. Technical work remains essential, but it does not answer the full client question. A fast, crawlable site can still fail to explain the offer, support important claims, earn relevant references, or attract qualified demand.

I now review an engagement as a chain of evidence: agreed priority, completed change, observable search response, user action, and business interpretation. The client contributes facts, approvals, and sales feedback; the agency contributes research, execution, and analysis.

When either side cannot show its part, the review should identify the missing input rather than assign unsupported causation. This shift makes the partnership easier to govern. It replaces vague confidence with a record that can support continuation, correction, or an orderly change of provider.

Day 1-7

Collect the agreement, priority services, access list, and a dated log of technical, content, internal linking, and outreach changes from the last 90 days.

Outcome: Create a verified baseline showing promised work, completed work, blockers, owners, and missing evidence.

Day 8-14

Map Search Console queries and landing pages to service intent, reachable markets, qualified actions, and the stage each metric represents.

Outcome: Separate useful leading indicators from irrelevant traffic, vanity positions, and outcomes that remain too early or uncertain to judge.

Day 15-21

Inspect the most recent 10 backlinks for source legitimacy, relevance, editorial context, destination fit, anchor use, and acquisition risk.

Outcome: Produce an accepted, questioned, rejected, and remediation list with the reason for each classification.

Day 22-30

Hold a decision meeting that compares delivery, technical risk, content quality, search movement, and qualified business outcomes.

Outcome: Choose and document one path: continue, correct the plan, narrow scope, set a final evidence period, or prepare a provider transition.