An SEO proposal should show what the agency believes is wrong, what evidence supports that diagnosis and why the proposed work deserves priority. If it jumps straight from logo to deliverables, you are not looking at a strategy. You are looking at a shopping list with a monthly fee attached.
That distinction matters before the contract is signed. Twenty articles, ten links and a technical audit can all be delivered perfectly while leaving the commercial problem untouched.
My rule is that the deliverables must be the consequence of the diagnosis. They cannot serve as its substitute. A proposal does not need complete access or perfect certainty before appointment, but it must show where observation ends, where judgement begins and what new evidence could change the plan.
That final condition matters. If every possible discovery leads back to the same package, the discovery has no authority. It is sales choreography wearing an audit badge.
TL;DR
- Hide the logo, client name and price. If the document could go to any business in your sector, the package came first.
- Screenshots are observations. A diagnosis separates observed, supplied, inferred and unknown.
- Follow every deliverable backwards through evidence, interpretation, consequence, action, dependency, measure and review decision.
- The order of work and the exclusions page reveal the real strategy. Choose the proposal whose reasoning is easiest to inspect.
Remove the agency logo and read the proposal again
The fastest proposal test is brutally simple: hide the cover, company name and price, then ask whether the document could have been sent to any business in your sector. If the answer is yes, the proposal probably describes the agency’s production capacity rather than your situation. A useful proposal should make several things recognisable: the market and search behaviour being examined; the pages, queries or journeys creating the problem; the evidence already observed; the assumptions that still need testing; the commercial consequence of doing nothing; and the reason one intervention comes before another.
Brand polish is not the enemy. Boostbery designs brands and websites; presentation matters. But design should make judgement easier to inspect. It should not turn an untested assumption into a confident-looking plan.
I call this the logo-cover test, but the logo is not really the target. The test removes borrowed authority. Once the agency identity, client name and polished cover are hidden, the reasoning must still reveal why this work belongs to this business.
Specificity is not achieved by repeating the company name. It appears when an observation, dependency or decision would stop making sense if copied into another account. If the proposal still fits the next business after a find-and-replace operation, the package came first.
A diagnosis is not a pile of screenshots
Search Console charts, audit scores and competitor screenshots are observations. They become a diagnosis only when the proposal explains what those observations mean for this business. Google describes Search Console’s Performance report as a way to inspect clicks, impressions, click-through rate and average position across dimensions such as queries, pages, countries, devices and dates. Those dimensions provide useful scope, but they do not decide what a business should do next (Google Search Console Performance report). A defensible proposal separates four categories:
| Category | What it means | How it should appear |
|---|---|---|
| Observed | The supplied data or review directly shows it | State the source, date and scope |
| Supplied | The client reports it, but the agency has not independently tested it | Attribute it to the client |
| Inferred | The evidence suggests an explanation | Name the inference and what would test it |
| Unknown | The information needed for a decision is missing | Turn it into discovery work, not fake certainty |
That discipline stops a crawl warning becoming “the reason traffic fell” and a competitor’s higher visibility becoming “proof we need more backlinks”. Both may justify investigation. Neither conclusion arrives automatically.
I do not expect a proposal to know what the agency has not yet been allowed to inspect. I do expect it to price uncertainty honestly. An unknown may justify a bounded discovery phase. It should not be silently converted into six months of predetermined production.
Follow every deliverable backwards
For each proposed activity, ask the agency to complete the chain from evidence to review decision. Suppose the proposal recommends rebuilding service pages. The chain might be:
- Evidence
Important service pages attract impressions but few qualified visits, while enquiries repeatedly misunderstand the offer.
- Interpretation
The search-result promise and page proposition may not match the buyer's decision.
- Business consequence
Visibility is being spent on people who are unlikely to buy, while suitable buyers receive weak qualification.
- Action
Revise the page architecture, proposition, supporting proof and search-result copy.
- Dependency
Confirm which services the company actually wants to sell, their scope and the available evidence.
- Measure
Inspect relevant query groups, qualified visits, useful conversion events and sales disposition.
- Review decision
Retain, refine or reverse the change based on the bounded evidence.
Now compare that with “optimise ten service pages”. The second version is easier to price. The first is easier to hold accountable.
This chain is my main deliverable test because it gives the buyer several places to challenge the work. The evidence may be weak. The interpretation may not follow. The consequence may be commercially trivial. The action may depend on access nobody has approved. The proposed measure may be unable to answer the question. Any one of those breaks can change the scope.
The chain does not need to appear as seven labelled boxes in the proposal. That would turn a thinking tool into another template. It must be reconstructable from the document and explainable by the person selling the work.
The order of work reveals the real strategy
A proposal earns its keep by choosing what not to do yet. If measurement is unreliable, producing more landing pages may only create more unmeasured activity. If the offer is unclear, keyword research cannot decide what the company is willing and able to sell. If the site has severe rendering or indexation problems, a content calendar may be premature. If existing pages already compete for the same decision, another article can make the estate worse. Look for explicit sequencing:
- What must be learned or repaired first?
- What becomes possible after that?
- Which work is conditional on the result?
- What has deliberately been deferred?
A ninety-day calendar is not automatically a strategy. It can be three months of coloured rectangles concealing the absence of a decision. My preference is for decision gates rather than calendar theatre. A useful first phase should produce knowledge capable of cancelling, narrowing or reprioritising later work. That may feel less reassuring than a fully populated roadmap, but it protects the buyer from paying for certainty the provider does not possess.
The strongest line in a proposal may be the work the agency refuses to recommend yet. Restraint is not automatically virtuous; it earns credibility only when the provider explains the evidence and threshold behind it. But a proposal that cannot name anything it would postpone is unlikely to be prioritising.
Ask what the agency needs from you
SEO work rarely happens inside a sealed agency box. It can depend on developer access, analytics consent configuration, subject-matter review, legal approval, photography, pricing decisions, CRM data or somebody being willing to say which enquiries were commercially useful. A credible proposal names those dependencies before they become excuses. It should identify:
This is not administrative padding. It tells you whether the plan has been designed for your organisation or merely sold to it. I would add an acceptance point to every important dependency. “Client to provide access” is incomplete. Which access, by when, in what usable state, and what happens to the sequence if it does not arrive? “Agency to provide recommendations” is equally vague. Is the work complete when a document is sent, a ticket is accepted, a change is deployed or the result is reviewed?
Those states carry different costs. A fast diagnosis attached to a six-week development queue is not a fast implementation. The proposal should show the operating reality before delay becomes somebody else’s surprise.
Comparing proposals and can’t tell which one actually diagnosed your business? We’ll read them with you and show you where each chain breaks.
Treat measurement language as part of the contract
“Increase rankings and traffic” sounds measurable until somebody asks which rankings, which traffic and why either outcome matters. Google explains that average position is the average topmost position recorded across impressions in the selected view. Search results also vary by factors including time, place, device and recent history (how Search Console counts position). A proposal that promises a naked ranking number without its query set, page scope, market, device or period is promising something less stable than the slide suggests. The measurement section should specify:
This last point is often missing. Reporting should not merely celebrate activity. It should create permission to continue, alter or stop it. That is the measurement standard I would sign: every major workstream needs a result capable of changing the next decision. A metric included only because it usually rises after activity is dashboard decoration. The proposal should explain why the signal matters, what else could affect it and what response follows from movement or non-movement.
Read the exclusions before the deliverables
The exclusions page is where a proposal becomes commercially real. Does “content production” include interviews, expert review, images and CMS upload, or only a document? Does “technical SEO” include implementation? Are design changes included? Who pays for development? Does digital PR mean campaign ideation, outreach or guaranteed coverage? What happens when the evidence contradicts the original plan? Precise exclusions do not make an offer weaker. They stop both parties discovering different versions of the deal after work has begun.
Be cautious when the proposal guarantees rankings, traffic, links, coverage or revenue it cannot control. Google’s own guidance on hiring an SEO advises asking for a technical and search audit and notes that nobody can guarantee a number-one ranking (Google Search Central: choosing an SEO).
The exclusions also reveal whether apparently comparable proposals describe comparable jobs. One fee may include interviews, writing, design, implementation and validation. Another may buy recommendations in a document while the client funds every dependency separately. Neither model is inherently wrong. They become misleading when the buyer assumes the same word (content, technical SEO or digital PR) contains the same work.
Run the proposal through one meeting
Before signing, ask the agency to explain the document without presenting it. Use five questions:
- What is the most important problem you believe we have?
- Which evidence supports that belief, and what remains uncertain?
- Why does the first proposed action come before the others?
- What do you need from us for the work to succeed?
- What result would make you change the plan?
Good answers need not be dramatic. They should be specific enough to disagree with. Then perform one final piece of adversarial reading. Circle every deliverable and draw a line back to the diagnosed constraint that permits it. Mark every unsupported jump. Identify which recommendation would disappear if its underlying assumption proved false.
If no recommendation can disappear, the proposal is not responding to evidence. If every activity is equally urgent, the provider has exported a backlog rather than made a decision. If success can only lead to renewal and failure also leads to renewal, measurement has been detached from accountability.
I would not choose the proposal with the most outputs. I would choose the one whose reasoning is easiest to inspect, whose limits are hardest to miss and whose first action has clearly earned the first invoice.
That is the standard. A proposal is not valuable because it contains more deliverables than the next agency’s PDF. It is valuable when the diagnosis, priorities and limits are clear enough for a buyer to make an informed decision before the production line starts.
Good to know
FAQs
A diagnosis before deliverables: what the agency believes is wrong, the evidence for it, what remains uncertain, why the first action comes before the others, what it needs from you, how results will be measured and what would change the plan. Deliverables should be the consequence of that diagnosis.
Hide the cover, company name and price, then ask whether the document could have been sent to any business in your sector. If it still fits after a find-and-replace, the package came before your situation. The test removes borrowed authority so the reasoning has to stand on its own.
Yes. Google’s own guidance on hiring an SEO says nobody can guarantee a number-one ranking, and average position varies by query set, device, place and time. A naked ranking promise without its scope is less stable than the slide suggests.
Not automatically. It can be three months of coloured rectangles concealing the absence of a decision. Look for decision gates instead: a first phase that produces knowledge capable of cancelling, narrowing or reprioritising later work.
Next step
Want a proposal you can actually interrogate?
Our proposals show the diagnosis, the evidence behind it, what we would defer and what would change the plan, so the first action has to earn the first invoice.