Almost every unhappy SEO engagement we have been asked to look at went wrong in the same place. The agency delivered what the contract described. The client expected something the contract never mentioned. Both were behaving reasonably.
Retainers are sold as a list of activities and bought as a business outcome, and the gap between those two is where the arguments live. A better scope document closes most of it before anyone has signed.
We sell retainers, so take the self-interest into account and judge the argument on its merits. Several of the recommendations below reduce an agency’s room to manoeuvre, including ours.
Start by writing down what success means
Not a ranking position. Not traffic. The business outcome the client is actually buying, stated in one sentence, with a number and a timeframe if one exists.
More qualified enquiries for a specific service. Reduced cost per acquisition compared with paid media. Visibility for a product line entering a new city. Recovery from a decline.
If the two parties cannot agree on that sentence, no amount of deliverable detail will help. If they can, it becomes the thing every quarterly review measures against, and it stops the conversation drifting towards whichever metric happens to look good.
The measurement model underneath it should be agreed at the same time, and we set out a straightforward one in calculating SEO and GEO ROI.
What belongs in the scope
- Named deliverables with a frequency. Not ongoing optimisation. A technical audit in month one, a defined number of pages written per month, a monthly report, a quarterly strategy review. Countable things.
- What is explicitly out of scope. This matters more than the inclusions. Paid media, development work, design, content translation, link outreach, whatever you do not do. Unstated exclusions become expectations.
- Decision rights. Who approves content, who approves technical changes, and how long approval is expected to take. An approval queue is the most common hidden cause of a stalled engagement.
- Dependencies. Plainly stated: this work requires changes to be implemented by the client’s development team, within an agreed time. More on this below.
- Access required. Analytics, Search Console, the content system, server logs, the staging environment. List what is needed, at what level, from when.
- Reporting cadence and format. Monthly report with named metrics, quarterly review meeting, and who attends.
- Review points. A scheduled point at which the plan can change without renegotiating the whole contract.
- Escalation. Who is called when something is wrong, before anybody drafts a termination email.
The dependency clause is the one nobody writes
Most SEO engagements that fail did not fail because the recommendations were wrong. They failed because the recommendations were never implemented.
The agency delivers an audit. It goes into a development backlog behind twelve feature tickets. Six months later nothing has shipped, the results are flat, and the agency is blamed for the results while the client is quietly frustrated about the invoice. Nobody is lying and everybody is unhappy.
Write it into the contract. State that the agency’s recommendations require implementation, name who is responsible, agree a reasonable turnaround for priority items, and specify what happens if that does not occur. At minimum, that the engagement is reviewed and the scope adjusted, so money stops being spent on recommendations nobody will act on.
This protects the client as much as the agency. It converts a vague dissatisfaction into a scheduled conversation with a decision attached.
Ownership and access, written down before anyone needs it
This is the section that matters most at the end of a relationship and is least considered at the start.
Every account must be created under the client’s own ownership, with the agency added as a user. Analytics, Search Console, the business profile, the tag manager, any tool the client is paying for. An agency that owns the client’s analytics property is holding the client’s data hostage, usually without intending to.
State who owns the content produced, which should be the client on payment. State who owns the data and the historical reporting. State that documentation, process notes and any scripts or configurations built for the client are handed over.
Then write the exit clause. Notice period on both sides, what gets handed over, in what format, and within how long. A clean handover costs an agency a day and is the difference between a client who speaks well of them and one who does not.
If you are currently in an engagement without these terms, check your account ownership today rather than at the end.
What should never be in a retainer
A ranking guarantee. Nobody controls the results page. An agency offering one is either inexperienced or planning to rank you for a phrase nobody searches.
A traffic guarantee. Same problem, one step removed, and it incentivises chasing volume over relevance.
A fixed number of backlinks per month. This is the single most damaging clause in common use. It converts link acquisition from a judgement about quality into a quota, and quotas get filled with whatever is cheapest. If links are in scope, describe the kind of placement and the standard, not the count.
A fixed number of blog posts with no quality definition. Volume without a standard produces exactly the thin content that creates the problem you were hired to solve.
Unlimited anything. Unlimited revisions, unlimited requests, unlimited support. It is not real, and both sides know it, which starts the relationship with a fiction in it.
Vague ongoing optimisation. If it cannot be counted or described, it cannot be reviewed, and in a difficult month it is what the invoice will be questioned about.
Pricing models, and which to avoid
A fixed monthly retainer against a defined scope is the workable default for most engagements. It is predictable for the client and plannable for the agency.
Hourly billing suits advisory and troubleshooting work and suits ongoing programmes badly, because it rewards time rather than outcome and it makes the client hesitate to ask questions.
Project pricing is right for bounded work, a migration, an audit, a site build, and wrong for anything continuous.
Performance-only pricing sounds attractive and is usually a poor arrangement. The agency does not control implementation, conversion, pricing, stock or sales follow-up, yet carries the risk for all of them. The predictable result is that the agency optimises for whatever metric is in the contract rather than for the business, and that incentives quietly diverge. A modest performance component on top of a base retainer can work. Pure performance rarely does.
Review it quarterly, properly
A scope agreed in January describes what was known in January. By April the site has changed, the market has moved and some assumptions were wrong.
Build in a quarterly review where the plan can be rewritten rather than defended. Three questions are enough. What did we say we would do and did we do it. What did it produce against the success sentence. What should the next quarter contain, given what we now know.
An agency that resists changing the plan is protecting its own comfort. A client who treats every change as a failure makes honest revision impossible. The review only works if both sides treat being wrong as information.
How to present that upwards is its own problem, and reporting AI search results to leadership covers it. The annual version sits in annual SEO and GEO planning.
Before you sign anything
If you are the client, the selection process matters as much as the contract, and the questions worth asking are in briefing an agency for GEO work. Whether you should be hiring an agency at all is a prior question, covered in in-house or agency.
Three things to check in any proposal. Does it state what success means in business terms. Does it say what is out of scope. Does it name who implements the recommendations. A proposal missing all three is describing activity rather than a result, and activity is easy to deliver and impossible to evaluate.
A sensible order of work
- Agree the success sentence first, in writing, before discussing deliverables.
- List deliverables with frequency, and list exclusions with equal care.
- Write the dependency clause naming who implements and within what time.
- Confirm every account is owned by the client, with the agency as a user.
- Write the ownership and handover terms while everyone is still friendly.
- Remove any guarantee, quota or unlimited clause.
- Schedule the quarterly review as a contractual event, not a courtesy.
If you want an existing proposal or retainer read through before you sign it, get in touch and we will tell you what is missing, including if the answer is that you do not need a retainer at all.