How to Choose an SEO Agency for Your Ecommerce Store
Anybody can write a blog post. Very few can diagnose a faceted navigation problem across forty thousand crawlable pages. That is what an ecommerce engagement actually needs. One question separates the two.
What An Ecommerce SEO Agency Actually Does
Six things. Most suppliers do two of them well. Knowing the full range is what lets you notice which two a proposal covers.
Technical diagnosis. Establishing what is preventing the shop being found.
Indexing, crawl, duplication and redirects. On a catalogue this is skilled work rather than a checklist, because the same symptom has several possible causes.
Structure. The category tree, product depth and internal linking.
Category and product work. Content on the pages that carry the commercial search, plus prioritisation across the ones that cannot all be worked on.
Content beyond the catalogue. Guides, comparisons and the questions buyers ask before choosing.
Measurement. Separating brand from non-brand, reporting at category level, holding a baseline.
Working with whoever maintains the shop. The one nobody lists and everybody needs.
Most of this work has to be implemented by somebody with access to the platform. An agency's ability to get changes made, by their own hands or somebody else's, decides how much of the rest ever happens.
What this page assumes. That you have read our pricing guide.
General agency vetting, meaning contracts, references and how to run a selection process, sits in our guides on choosing an agency. Everything here is specific to shops.
What A Service Should Include
Eight workstreams, each with the reason it matters at catalogue scale. The purpose of the list is to make a missing item visible, since a proposal rarely announces what it leaves out.
Technical health and crawl. Because on a large shop, attention spent on filter combinations is attention not spent on products.
Faceted navigation specifically. Because it is the largest single technical issue on most big catalogues and the one block three is about.
Site structure. Because it sets the ceiling. A tree inherited from a supplier targets terms nobody searches.
Category page content. Because those pages carry the commercial search and are usually empty.
Product page prioritisation. Because five thousand products cannot be treated equally and deciding which few hundred matter is the skill being bought.
Content beyond the catalogue. Because a shop publishing only products has a ceiling it cannot pass.
Structured data. Because it describes price and availability accurately, which on a shop is a consumer protection matter as well as a search one.
Measurement separating brand from non-brand. Because without it a report can show growth the work did not produce.
What a missing item usually means. That the supplier cannot do it.
A proposal covering content thoroughly with nothing on technical work is describing what that supplier produces rather than what your shop needs. The gap is rarely mentioned.
The Question That Sorts Them
Ask how they would approach faceted navigation on your shop. That single question separates suppliers who can do this work from suppliers who produce content, faster and more reliably than any other part of a selection process.
Why this question specifically. Because it cannot be answered from a template.
It requires understanding your catalogue, having an opinion about which combinations deserve to exist. Knowing what the consequences of each option are. None of that can be produced by somebody who has not done it.
What a competent answer sounds like. Questions before answers.
They ask how many products you have and how many filters. They ask which filter combinations people actually search for, because that is the deciding question. They want to know how many pages exist against how many are indexed before saying anything definite. And they describe options with consequences rather than announcing a fix.
The specific thing to listen for. Whether they distinguish between combinations.
A competent answer says some facet combinations deserve proper pages because people search for them, most should not be findable at all. The middle group should be usable without being treated as pages. Anybody making that distinction unprompted has done this work.
What evasion sounds like. Four versions, all common.
Reassurance without content. Something about optimising your faceted navigation, said confidently, containing no decision.
A universal answer. Naming one technique as the solution without asking anything about your shop. There is no single right answer here. Offering one demonstrates that.
Changing the subject. Moving to content, keywords or reporting, which is what a supplier does when the ground is unfamiliar.
Deferring to a tool. Saying their software handles it. Software finds these pages. It does not decide which should exist.
What a good answer does not include. A promise to fix it before they have looked.
Who Does The Technical Work
An agency that recommends without implementing is half a service on a shop. That is not a criticism of advice. It is a description of what happens to advice nobody executes.
Why it matters more here than elsewhere. The changes are not editorial.
On a brochure site most recommendations are content somebody can paste in. On a shop they are changes to how a platform behaves. A store owner cannot action a recommendation about faceted navigation by editing a page.
What actually happens to unimplemented recommendations. They wait.
An audit arrives, the shop reads it, the important items need a developer, the developer is busy. A year later the same findings appear in the next audit. We have seen shops buy three audits over four years and implement none of them.
The three questions to ask. Directly, before signing.
Do you implement or merely recommend? A supplier who only recommends is fine if you have somebody to act, half a service if you do not.
Who touches the platform? Whether they have people who work in it. Or whether everything goes through your developer.
What happens when a fix needs development work? Who pays, who schedules it, then what happens when it does not get scheduled.
The arrangement that fails quietly. Nobody owning implementation.
The agency assumes the developer will action it. The developer assumes the agency will say when it is urgent. Nothing happens for months and both parties are working as agreed.
Questions To Ask
Beyond block three, five questions that reveal how an engagement will actually run.
What access do you need? The most revealing of the five.
A supplier who does not want platform access is not planning to change anything on the platform. What they need tells you what they intend to do.
What happens to the work if we part company? Ownership of content, documentation. Anything built in their accounts rather than yours.
Analytics configuration, category groupings and tracking set up in a supplier's own property leave with them. Establish that everything lives in accounts you own.
How do you report? Is brand separated from non-brand? Per our measurement guide.
A supplier who does not separate them is either unable to demonstrate their own contribution or has chosen not to. Ask before signing rather than discovering it in month six.
What would you do first? The answer that reveals their sequence.
Diagnosis first is the right answer. Content production in month one on a shop nobody has crawled tells you what you are buying.
How many products do we have? Ask whether they asked.
Per our pricing guide, catalogue size is the main driver of the work. A supplier who quoted without establishing it has priced a website rather than a shop. The figure is a guess whatever it says.
Warning Signs
Guaranteed rankings. The clearest of the five.
Nobody controls search results, so a guarantee is a claim about something outside the supplier's control. It signals either a misunderstanding of the work or a willingness to say what wins the contract.
A fixed price with no audit. A number produced without looking.
Two shops in the same market can need entirely different work depending on catalogue size and what previous suppliers left behind. A price arriving before anybody examined the shop is a guess or a template.
No interest in your margins. The one shops never think to notice.
A supplier who never asks what you make on a sale cannot prioritise properly, because deciding which categories and products deserve effort requires knowing which are worth having. Anybody working purely from search volume is optimising for traffic rather than for your business.
All content, no technical work. A proposal of articles and descriptions with nothing on crawl, structure or duplication.
Per block two, that describes what the supplier produces rather than what your shop needs.
Nobody asked how many products you have. The simplest signal available.
It takes one question and it changes almost everything about the engagement. A supplier who did not ask has not thought about your shop specifically, whatever the proposal says.
What none of these means. That the supplier is dishonest.
Most are describing a service that works well for ordinary websites. The gap is that a shop is not one. Nothing in their process surfaced the difference.
Judging Their Own Evidence
Every supplier presents case studies. Every case study shows something improving. The question is whether the work caused it, which four questions establish.
Was the growth brand or non-brand? The first thing to ask, always.
A shop that increased its advertising during the engagement will show organic growth that the search work did not produce. A case study that does not separate them has not answered the question it appears to answer.
What else was happening? The context a chart omits.
New products, a rebrand, a television campaign, a competitor closing, a seasonal peak. Ask what else changed in that period, then note whether the answer is ready.
What did the shop look like before? The one that reveals the most.
A shop with a technical fault that was suppressing everything will show dramatic improvement from one fix. That is real. It says nothing about whether the supplier can help a shop that is already technically sound.
Was it revenue or traffic? Per our measurement guide.
A case study reporting sessions rather than money on a shop, where money is available, has chosen the more flattering figure.
What a strong case study contains. Things that are inconvenient.
What did not work, what took longer than expected, then what the shop had to do itself. A study where everything went to plan describes a presentation rather than a project.
The question worth asking last. Can we speak to them?
Pricing Models
Four arrangements. Each incentivises something. Knowing what tells you how an engagement will behave when nobody is watching.
Retainer. A monthly fee for ongoing work.
Suits a shop that changes continuously, which is every shop. It incentivises retention, so a supplier has reason to keep producing visible activity, which is why block five asks what happens in month one and our reporting guidance matters.
Audit plus implementation. Diagnosis bought once, then work commissioned against it.
Suits a shop with hands available. It incentivises a thorough audit, which is good. A long list of findings, which is not always the same thing.
Project. A defined piece of work with an end.
Suits a specific problem such as a migration or a restructure. It incentivises finishing, which is clean. It leaves nobody responsible for the shop afterwards.
Performance based. Payment tied to results.
Appealing and worth understanding before agreeing. It incentivises whatever is being measured, so if the measure is traffic you will get traffic. If it includes brand searches you will be paying for demand you generated. It also gives the supplier reason to avoid slow structural work in favour of anything that moves the number quickly.
What we use, with the reason. A retainer at a fixed price with everything included.
It removes the argument about whether something is in scope, which is the argument that damages these relationships most.
Comparing Two Proposals
Two proposals are rarely the same service at different prices. Four steps make them comparable.
One. Map both against block two. Mark what each covers and what neither mentions.
The gaps are more informative than the contents. They are what a proposal is designed not to draw attention to.
Two. Establish what sits outside each fee. Per our pricing guide.
Development time, applications, photography and product data work are charged differently by different suppliers. A cheaper fee with development billed separately can cost more within a year.
Three. Ask both the same question from block three. Then compare the answers rather than the documents.
This is the most useful thing in this guide. A proposal is written by whoever is best at writing proposals. An answer to an unexpected technical question is not.
Four. Compare what happens in month one. Diagnosis or production.
What not to compare. The number of deliverables.
A proposal offering more items is offering smaller ones. Twelve blog posts and a monthly report is more deliverables than a technical audit and a structural fix, plus considerably less work.
The question to end on. Which of these has understood our shop?
After block three, that is usually obvious. It matters more than the price difference between them.
Ask them about
your filters.
How would you approach faceted navigation on our shop? A competent answer asks how many products and filters you have, plus which combinations people actually search for. Evasion names a technique without asking anything.
What is included every month:
£350 per month, one target area. No setup fee, nothing billed separately.
Twenty-two guides.
One subject.
This guide covers choosing a supplier. The rest of the series covers the sequence, structure, category and product pages, technical health, measurement and everything a store owner has to decide.
