How to assess technical SEO agencies with evidence

What to take away

Decide the shape of the purchase before you read a pitcha fixed-scope audit or a retainer.
Treat sales decks, case studies and award badges as vendor claims. Note them, then set them aside.
Only two kinds of input carry weightevidence you could re-run, and guidance from a regulator or platform.
Score every supplier on one rubric, so the comparison holds when finance or procurement asks how you chose.
Write down what you did not assess. A stated gap is cheaper than one found six months into a contract.

What counts as evidence when assessing technical SEO agencies?

Material falls into three piles. Vendor claims are anything a supplier publishes about itself. Independent evidence comes from platform documentation, regulator guidance or legislation. The third pile is editorial judgement, where both are weighed against your brief and the reasoning is shown.

Evidence piles for reviews

Vendor claims

Source: Supplier self-published
Example: Delivery model
Use: Claim to test

Independent evidence

Source: Platform, regulator, law
Example: Google mobile-first
Use: Test for parity

Editorial judgement

Source: Weighed against brief
Example: Shown reasoning
Use: Decision basis

This review assessed published documentation, scoring rubrics and stated delivery models. It did not assess live campaign results, client retention or staff seniority, none of which can be verified from outside a business.

Google's mobile-first indexing guidance sets the test for any parity claim, because the mobile version of a page is the one crawled and indexed. An agency that still reasons from a desktop crawl has a thin basis for its recommendations.

Metrics need the same treatment. The GOV.UK service manual frames performance measurement around a baseline, a target and a named method. A retainer that cannot describe its baseline is producing decoration, not reporting.

A written source list for technical SEO assessments

Use the same written references for every supplier and record the relevant document or tool output in the rubric. Distinguish authoritative guidance from diagnostic evidence.

Platform documentation:Google Search Central's Search Essentials, Google's mobile-first indexing guidance, Google Search Console documentation, Google's Rich Results Test, Bing Webmaster Guidelines and schema.org's structured data vocabulary.
Audit and diagnostic tools:Screaming Frog SEO Spider, Sitebulb, Lumar, Ahrefs, Semrush and PageSpeed Insights. Use relevant, reproducible tool outputs as evidence, not as platform policy.
Data protection:Information Commissioner's Office (ICO) guidance, the UK GDPR and the Data Protection Act 2018.

Data handling belongs in the review too. Where an agency runs outreach or holds lead data, ask it to identify and document the lawful basis under the UK GDPR, read alongside the Data Protection Act 2018, before the first send. Check the ICO's data-protection guidance as well; the lawful basis should not be chosen after a complaint.

What should you score, and what should you leave out?

Score what you can point at: crawl diagnostics, indexation and canonical handling, mobile parity, structured data coverage, and the clarity of the written method.

Leave out what you cannot check: team churn, private client reasons for leaving, and internal tooling. Park those in a column marked unverified, separate from the score.

Weighting should follow your own risk. A site moving platforms cares more about indexation control than about content briefs. A stable local build cares more about reporting clarity. Same rubric, different weights, agreed before scoring starts.

Our guide to technical SEO reviews and comparison methods for 2027 explains how to fix weights before scoring. Stable weights stop a familiar name drifting to the top of the table.

How do you run a technical SEO agency review?

Write the decision and a budget band on one page. For example, a team paying £2,500 a month for an audit retainer should specify quarterly handovers such as a crawl export, a prioritised issue log with affected URLs and evidence, written recommendations, a performance report stating the baseline, target and method, and a record of retests.
Use the source list above alongside supplier documents.
Set the rubric and its weights before you contact anyone.
Ask every supplier the same written questions about how they audit and what they deliver.
Score each answer twice, once against evidence and once against the brief, and mark unsupported claims.
Record the gaps, set a review date, and file the scored sheets with the contract.

How do you keep the scoring honest after the contract starts?

Re-run the rubric at each renewal, using delivery evidence rather than the original pitch. Dated notes matter more than memory here.

If tooling sits alongside the agency, read technical SEO product comparison in England first, because a platform choice can quietly widen or narrow the retainer's scope.

A short written owner for the review keeps it alive. Without one, the scored sheets become an archive nobody opens.

Common questions

Do paid review directories count as evidence?

No. A directory entry shows that a supplier listed itself or bought a badge. Score it as vendor material, exactly as you would a sales deck.

How many suppliers should the rubric cover?

Enough to compare, few enough to score properly. Six completed entries with dated notes beat twenty half-finished ones.

What if a supplier will not share its audit method?

Then there is nothing to score. Ask once, in writing, and record the refusal in your notes rather than in the score column.

Should data protection questions be in scope?

Yes, where lead data or outreach is involved. The lawful basis question comes first, and a short written answer is enough at review stage.