How to choose an SEO agency

Choosing an SEO agency is a decision about how work gets done. A proposal can sound comprehensive while leaving content, development and measurement responsibilities unresolved. Start with the business problem and ask each provider to explain how its scope addresses it.

Define the outcome your team actually needs

“More traffic” is rarely a complete brief. A professional-services business may need inquiries for a specific service. A retailer may need customers to find an available product. A multi-location operator may need people to reach the correct branch with accurate hours and directions. Those needs point to different pages and measurements.

Write down the priority audience, the desired action and the systems used to observe it. Then identify the constraint: limited content, a difficult platform, weak location information or uncertainty about the data. This gives an agency something concrete to investigate.

Ask for the baseline behind the recommendation

A useful proposal distinguishes what the provider has observed from what it intends to examine after access is granted. Ask which properties, periods and tools support the initial diagnosis. Search Console clicks, analytics sessions and third-party traffic estimates are different measures and should be labelled accordingly.

Watch for comparisons with a tiny or zero baseline. Moving from no recorded traffic to a positive number is an absolute gain, not a defined growth multiple. Portfolio averages also need a clear project set and methodology. A single headline should not hide whether the result is typical, estimated or driven by one unusually large project.

Inspect the work behind a case study

A case should explain the starting problem, the work delivered, the period and the evidence for any result. A customer review can be useful evidence of that customer's experience. An analytics export can support a particular metric. Neither automatically proves that the agency alone caused every business change.

Ask for a relevant example rather than only the largest number. A provider experienced with content-led ecommerce may approach a multi-location service business differently. The useful question is whether the team understands your operating constraints and can explain its responsibilities.

Make implementation responsibilities explicit

An audit is valuable only if someone can act on it. Clarify whether the agency writes content, publishes it, changes templates, fixes technical issues and coordinates with your other suppliers. If your developer has to implement every recommendation, their capacity belongs in the plan.

The same applies to subject-matter review. Your team may need to approve product claims, service descriptions, regulated information and customer examples. Agree a review process that can keep up with the proposed publishing cadence.

A proposal comparison worksheet

• Problem: what evidence shows this is the right priority?
• Deliverables: which pages, repairs and reports are included?
• Responsibilities: who writes, approves, implements and accepts each item?
• Measurement: what does each metric mean and where does it come from?
• Dependencies: what access, input or third-party work is needed?
• Ownership: who controls the domain, accounts, content and source files?
• Commercial terms: what is recurring, what is a project and what costs extra?

Evaluate how the agency reports uncertainty

A good working relationship needs room for evidence to change the plan. Ask what the provider does when a page fails to gain visibility, a migration affects traffic or the expected conversion data is unavailable. The answer should include investigation and a decision process, rather than an unqualified guarantee.

Look for reports that connect changes to the next action. A long ranking list can be informative, but you should also understand which work was completed, what remains blocked and what the business needs to decide. The report should be usable by the person accountable for the budget.

Choose an engagement shape that matches the problem

A diagnostic suits an uncertain starting point. A defined implementation suits a known set of repairs or a rebuild. An ongoing program suits a continuing need for content, maintenance and iteration. You do not need to buy all three before the first useful decision.

For a team with strong internal delivery capacity, a specialist review may be enough. For a team without that capacity, a proposal that stops at recommendations may create work without resolving the problem. Our agency versus in-house comparison helps map those responsibilities.

Before you sign

Confirm the scope, exclusions, access requirements, reporting cadence and exit arrangements. Make sure the proposal uses the same definitions as the sales discussion. You should know what the first deliverable is, who accepts it and what information the agency needs to begin.

A ten-question interview for the shortlist

Use the same questions with each shortlisted provider, and retain the answers beside the proposal. You are looking for a clear relationship between the diagnosis, the work and the people responsible for delivering it. A provider does not need to know every answer before receiving access, but it should explain how it will investigate the unknowns.

1. What would you investigate first, and why?

Ask the provider to identify a specific uncertainty in your current search performance. It might be whether important pages are being discovered, whether location information is consistent or whether visitors can complete the inquiry journey. A useful answer explains what evidence would confirm or change the diagnosis. A fixed list of monthly tasks, offered before anyone understands the business, tells you less about the quality of the proposed strategy.

2. Which comparable project can you explain in detail?

Choose an example with relevant operating constraints: a similar platform, customer journey, location network or content-review process. Ask what the agency actually delivered and what the client team contributed. For each performance figure, ask for the metric definition, source and period. A smaller, well-documented example may be more informative than an impressive number whose baseline or relationship to the work is unclear.

3. Who will be working with us?

Ask for the delivery lead and the responsibilities of any specialists. Find out who writes, who implements technical changes and who reviews factual claims. Understand when subcontractors are involved and how continuity is handled. You do not need to dictate the provider's staffing model, but the proposal should make accountability clear enough that an unresolved issue has somewhere to go.

4. What do you need from our team?

List the approvals, subject-matter interviews, development capacity and system access required. Ask how delays in those dependencies affect delivery. If the agency plans to publish several pages while your internal reviewer can approve only one, the schedule needs adjustment. A realistic plan includes the client's work as well as the supplier's work, so neither side mistakes a blocked dependency for completed delivery.

5. What does the first review contain?

Request a sample agenda and report with confidential information removed. Look for completed work, clearly defined measures, unresolved decisions and the next priorities. Ask whether the report distinguishes estimated traffic from measured visits and qualified inquiries. A dashboard becomes more useful when somebody can explain what decision follows from it, rather than simply presenting a collection of upward arrows.

6. How do you choose and review content?

Ask how topics connect to actual buyer questions and the business's expertise. Clarify who supplies original information, checks accuracy, approves publication and maintains older pages. Discuss the use of AI or other production tools in terms of review and responsibility. The useful issue is whether the finished content is accurate and valuable to your audience, not how quickly a tool can produce a draft.

7. What happens after a technical issue is found?

An agency may identify a crawl obstacle, broken redirect or measurement problem without having permission to repair it. Ask who will implement the change, how it will be tested and whether that effort is included. For a website migration, require a specific plan for existing URLs and launch checks. An audit backlog without an implementation owner can remain unchanged through several reporting cycles.

8. How do you handle local and AI-search reporting?

If these channels matter to your business, ask which systems and properties the agency can actually observe. For local work, clarify how individual locations are represented. For AI-search work, distinguish a provider's citation report from a small set of manually tested prompts. Ask how changes in coverage will be disclosed. A citation count should not be relabelled as visitors or added to ordinary website traffic.

9. What are the commercial boundaries?

Read the term, renewal, cancellation, payment and change-control provisions. A minimum engagement period can reflect real setup work; a flexible contract can still have unclear exclusions. Evaluate the actual agreement rather than assuming that one billing model proves confidence or quality. Identify what happens to unfinished work, prepaid services and subscriptions if the engagement ends.

10. What do we keep at handover?

Confirm ownership and access for the website, accounts, content, source files and reporting records. Ask for an example of operating documentation and an explanation of how permissions will be removed. If an agency uses its own tools, clarify which outputs you can retain and whether continued reporting depends on its subscription. A useful handover lets the next team understand the work without reconstructing it from old presentations.

Compare proposals without inventing precision

Create a short comparison sheet using the responsibilities that matter to your business. Mark each as included, excluded or unclear, and link to the relevant proposal section. Keep quality observations in plain language. A numerical score can look objective while concealing that one provider includes implementation and another only supplies recommendations.

For cost, compare the same planning period and include your own team's required contribution. Record content production, development, subscriptions, specialist reviews and ongoing maintenance separately when the proposal does. Ask for clarification where the scope is unclear. Do not assume a high fee includes every service or that a low fee is automatically poor value; evaluate what can actually be delivered.

Use a defined first engagement when uncertainty is high

A diagnostic or limited implementation can be a useful first purchase when the underlying problem is uncertain. Agree the questions it must answer, the evidence you receive and the decision it should enable. This gives both teams a way to test the working relationship. The diagnostic should produce something useful even if you choose a different partner for the implementation.

For example, a multi-location business might begin with an inventory of location pages, profile ownership and inquiry routing. That scope can establish whether the next priority is a website repair, information cleanup or a content programme. It is more concrete than purchasing a large publishing commitment before the team knows whether customers are reaching the correct branch.

Read reviews and references in context

Reviews can help you understand communication, reliability and a customer's view of the work. Read the detailed account, date and service described, rather than relying only on the average rating. A website redesign review may say little about a long-term search programme, while a technically detailed case study may say little about day-to-day collaboration. Use several kinds of evidence to answer different questions.

Keep platform ratings separate when they represent different review sets. Ask permission before contacting a private reference, and respect any confidentiality limits. A provider should be able to discuss its delivery approach without exposing another client's private information. If a result is supported only by a client's recollection, it should be presented as that person's account rather than as an independently measured analytics finding.

What to evaluate after work begins

Review delivery and performance as related but distinct topics. Early evidence can include a documented baseline, repaired technical issues, approved content and working measurement. Search outcomes need an appropriate observation period and can be affected by demand, seasonality, competition and other changes. Agree the review approach at the start so neither team moves the goalposts after seeing the data.

If progress stalls, identify the constraint before changing suppliers. The issue may be the quality of the recommendation, a missing implementation owner or a slow approval process. Ask for a concrete recovery plan with responsibilities and checkpoints. A good review should end with a decision about the work, the scope or the relationship, rather than another report nobody can act on.

For Google's own hiring guidance, read Do you need an SEO?. To explore our approach, see SEO services, Vancouver SEO and engagement options.