Engagements stall on inputs more often than on output. The agency starts, requests a set of things, and half of them take three weeks to gather because nobody internally knew they would be needed. That delay is not visible in any timeline and it consistently costs the first month.
Most of what gets requested can be prepared before you sign anything. Here is what a competent GEO Agency will need, roughly in the order it becomes urgent, along with what tends to go wrong when it arrives late or incomplete.
Week One Requests: Access and Baseline
Analytics and search console access. Standard, and usually the fastest to arrange. What slows it down is that the person who controls it has left, or the property was set up by an agency you no longer work with.
Site and hosting access, or a named contact who has it. Not necessarily for them to make changes, but to verify what is technically possible and how long deployments take. This determines the entire schedule, so it comes first.
A list of your commercial priorities. Which services actually matter, ranked, with revenue attached if you can. Agencies routinely receive an undifferentiated list of everything the company offers, then optimize proportionally rather than commercially. You know which three lines pay for the business. Say so.
Your competitor list, and the honest version. Not the aspirational one. The companies you actually lose deals to, including the cheap local outfit nobody wants to name in a meeting. Aspirational competitor lists produce analysis of companies your buyers are not comparing you against.
Where it goes wrong: access requests bounce between IT and marketing for two weeks. Assign one internal owner for access before day one and the first month behaves very differently.
Week Two Requests: Language and Positioning
This is where internal disagreement surfaces, usually for the first time in years.
A plain statement of what you do. What you sell, to whom, in what geography, at what scope. It sounds trivial and it frequently takes three internal drafts, because different departments describe the business differently and nobody had noticed.
How customers describe it, which is not the same thing. Buyers use the wrong word for your service consistently and predictably, and that wrong word is what gets typed into an assistant. Ask two salespeople what customers call your service when they get it wrong.
Your genuine differentiators, stated as facts rather than qualities. Not that you are experienced, but that you have held a specific accreditation since a specific year. Not that you are responsive, but that you quote within a stated timeframe. Adjectives cannot be extracted or verified. Facts can.
Anything you cannot claim. Compliance restrictions, regulatory language limits, claims legal has previously rejected. Supply this early or you will spend week five unwinding content that cannot be published.
Where it goes wrong: the agency writes the positioning because nobody internally would commit to a version. Whatever they produce will be generic, and it becomes the foundation of everything downstream.
Week Two to Three: The Question Set
The single most valuable input, and the one most often delegated to the wrong source.
What is needed is what buyers ask before they contact you, in their own phrasing, including the confused and repetitive versions. What is usually supplied is a keyword export, which contains the vocabulary of your industry rather than the vocabulary of your customers.
Get this from conversations, not tools:
- Sales call recordings or notes from the past quarter
- Support tickets from the first week of a customer relationship
- The questions your team is tired of answering because they come up every single time
- Whatever your website chat log contains, which is usually the least filtered source you have
Where it goes wrong: somebody assembles it from a search volume tool in an afternoon because it is faster. The resulting content targets phrasing nobody uses, and this failure is invisible until month four.
Week Three: External Presence Records
The unglamorous request that gets ignored, and the one with the highest ratio of impact to effort.
Every place your company is listed. Directories, industry associations, professional bodies, partner pages, franchise or dealer networks, review platforms, old press coverage, records from any rebrand or acquisition.
Most companies cannot produce this list, which is itself the finding. Independent sources carry more weight than your own site, which means an outdated external description is harder to overcome than an outdated internal one. If a partner page still describes a service you discontinued, it is actively working against everything else being done.
Login credentials or contact routes for each. This is where the real delay sits. Correcting a listing requires access somebody set up four years ago, and tracking that down takes longer than the correction itself.
Week Three to Four: Approval Routes and Constraints
Boring, and it determines the pace of everything after month one.
Who approves content, and how long they take. A two week legal review cycle is workable if planned around and destructive if discovered in month two.
Who can deploy to the site, and what the queue looks like. Schema work is easy to specify and sits behind a development queue. If that queue is six weeks, the timeline is set by the queue rather than by the strategy.
What internal reporting the programme must feed. Establish early whether this needs to produce numbers for a board deck, because retrofitting reporting to a framework that was never designed for it is painful.
Where it goes wrong: the constraint gets discovered rather than declared. Every engagement has approval bottlenecks. The ones that stay hidden until month two cost a full month.
What You Should Ask For in Return
Inputs run both ways, and these are reasonable to require in the first month.
The baseline, in a form you can reproduce. Not a composite score. The actual queries tested, who appeared, and what the assistant said. You should be able to run three of those queries yourself and get comparable results.
The specific gaps identified, as a list of questions rather than a summary of themes. Themes are not actionable and cannot be verified later.
What they intend not to do. A plan without exclusions is a task list, and the exclusions tell you whether anyone made real choices.
How position loss will be detected. Positions shift as competitors publish and platforms revise how they select sources. Ask what triggers a response and how quickly, since this is the part most likely to be absent.
On the last point, insist on measurement you can check independently rather than accepting a supplied dashboard as the only source of truth. Reviewing what any best chatgpt seo rank tracker actually counts is worth doing before you rely on someone else’s version of your numbers.
How Preparation Changes the Timeline
The difference is not marginal.
Extension Architecture, a London firm handling extensions, loft conversions, and renovations, ran roughly three and a half months of content clustering, internal linking, and schema expansion against an existing foundation. EC Council, the global certification body behind CEH and CND, needed six months across a much larger content estate.
Neither timeline includes the time a company spends assembling inputs before real work starts. That period is invisible in every case study and it is frequently four to six weeks of the engagement you are paying for.
Prepare the positioning statement, the honest competitor list, the customer phrasing, and the external listings inventory before the first call. Those four take about a day internally and they can compress the first month into the first fortnight.
