One buyer, three different questions
The same person asks “why is this happening”, then “what kind of help is this”, then “what exactly am I buying”. Those are three different pages, and a single landing page answers none of them well.
Skip to content
Buying journey
Five stages, and what has to exist at each one
A GEO programme does not fail everywhere at once. It fails at one specific point, and which point depends on where the buyer is. This page walks the buying journey: what the buyer types at each stage, what has to be on your site for the answer to be possible, and what most companies publish instead.
Overseas reference price starts at US$1,800–4,000 per month, minimum term 3 months. Scope and fees are set per market, language and content volume in a written proposal.
Why the stage decides the work
Assets do not transfer between stages
A baseline record is useful in stage one and irrelevant in stage three. A price table is decisive in stage three and noise in stage one. Publishing more of the wrong stage's material does not close the gap.
The absence is usually earlier than expected
Companies that feel stuck at stage four often have a stage one gap: no saved baseline, so no month can be compared with any other, so no result can be demonstrated either way.
The five stages, one at a time
Each stage lists the questions buyers actually ask, the assets that have to exist to answer them, the more common substitute that does not, and the record you end up with.
01Something is wrong and nobody knows why
The buyer has noticed an absence, not a solution
What the buyer is asking
- Why does ChatGPT leave my company out when someone asks about our category?
- Where can I see what the assistants actually say about us, today?
- Is this something we can fix internally, or does it need an outside supplier?
What has to already exist
- A written record of what the answers say right now: the same 10–20 questions, asked identically, saved with the date and the platform.
- Pages that state plainly what you make, for whom, and where you ship — the baseline any later change gets measured against.
- One named person on your side who owns the question set, so the second round is comparable with the first.
What gets published instead
Essays about the future of search, published on a site that never states its own product specifications. The buyer's question is about them, and the page is about the industry.
What you end up with
The baseline. Without it every later claim is unfalsifiable — including ours.
02Comparing approaches
The buyer is deciding what kind of help this is
What the buyer is asking
- Which of these suppliers does what I need, and what do they actually hand over?
- What does this do that the SEO agency we already pay does not — and what is the overlap?
- Has anyone like us done this, and what did it involve week by week?
What has to already exist
- Pages that draw the comparison honestly, including where the two disciplines share groundwork. A comparison that claims no overlap reads as sales material.
- Deliverable descriptions specific enough to be checked: named artifacts with a stated acceptance check, not outcomes.
- Method pages a technical reviewer can read without sitting through a call.
What gets published instead
A “why choose us” page listing founding year and awards. It answers none of the three questions above, and the model answering on your behalf has nothing to quote.
What you end up with
Which of your pages the answers cite at this stage — and which competitor pages they cite instead.
03De-risking the decision
The buyer has chosen the approach and is now buying it
What the buyer is asking
- What exactly happens in the first month, and what do I receive at the end of it?
- What is the minimum term, what does it cost, and what happens if it does not work?
- Who does the work, and how do I confirm it was done rather than described?
What has to already exist
- A published basis for the quote — unit prices or a stated formula — so the buyer can compute before the call rather than after it.
- Terms in writing: minimum term, what is included, what is billed separately, how it ends.
- A verification route for each deliverable: the thing the client opens or queries to confirm it exists.
What gets published instead
“Contact us for a custom quote” plus a calendar link. That moves the buyer's uncertainty into the sales call instead of removing it, and models cannot quote a price that has never been published.
What you end up with
That each deliverable was accepted against its check, with the date it was accepted.
04Using what was bought
The buyer now has to judge whether it is working
What the buyer is asking
- How do I read this month's report, and what changed since last month?
- Which of the recommended changes is worth doing first?
- Is the work actually moving anything, or are we paying a monthly fee for activity?
What has to already exist
- A report format fixed before work starts — same question set, same layout — so two months can be laid side by side.
- The original answers saved verbatim, so the client reads the source rather than the summary.
- A decision cadence with dates, and the authority to stop something that is not producing.
What gets published instead
A dashboard of composite “visibility scores” that changes colour but cannot be recomputed by the client. If only the supplier can produce the number, it is not evidence.
What you end up with
Round-over-round differences on the same questions, with a log of what changed in between.
05New markets and new languages
The buyer wants the same thing somewhere else
What the buyer is asking
- We are entering another market — what changes, and what carries over?
- Can our distributor reuse this material, and under what attribution?
- We have a result we can talk about. Where should it be published?
What has to already exist
- A separate question set per market and language, because buyer questions do not translate literally — the constraint that matters in one market is often irrelevant in another.
- Language declarations that match which pages actually exist, and translations that someone has read.
- Third-party corroboration in the new market, which is slower than publishing and cannot be compressed by paying more.
What gets published instead
Machine-translated pages for six countries, live within a week, with no local question set and no local source. The site now exists in six languages and is a candidate for none of them.
What you end up with
Whether the new market's questions mention you at all. It is the first useful number and it is usually zero.
Where you probably are
The four columns below are deliberately about assets and records rather than about tiers or packages. Find the row that matches what you can produce today.
| Stage | Buyer's question | Asset that answers it | Record produced |
|---|---|---|---|
| 01 Problem | Why is this happening to us? | Baseline record + plain description of what you sell | What the answers say today |
| 02 Comparison | What kind of help is this? | Honest comparison pages, named deliverables, method pages | Which pages get cited |
| 03 Purchase | What am I buying, and what if it fails? | Unit prices or a stated formula, written terms, per-item checks | Acceptance dates |
| 04 Usage | Is it working? | Fixed report format, saved original answers, review cadence | Round-over-round differences |
| 05 Expansion | What changes in the next market? | Per-market question set, reviewed translations, local sources | Whether the new market mentions you |
Three layers of measurement
Measurement in this category gets complicated mostly because three different questions are answered with one number. Separated, each layer is answerable on its own.
Layer 1
Can it be reached at all
Whether the pages exist, are crawlable, render without waiting for client-side scripts, and are readable as text. This is the only layer you can settle in an afternoon, and it is binary enough to argue about with evidence.
Layer 2
Does anything answer the question
Take your question set and match each question to a page that answers it in a sentence. The output is a list of questions with no home, which is a work plan rather than a score.
Layer 3
Does the answer change
The same questions, asked the same way, with the original answers saved. This is the only layer that shows effect, and the only one that cannot be produced without a baseline taken earlier.
No thresholds are published on this page, on purpose. A figure like “indexed above 80%” reads as an industry standard and is really somebody's target; against your own previous round it means something, and against nothing at all it does not. The useful comparison is your baseline, not a benchmark we chose.
If you have to put a number on it
Finance will ask what the spend returned, and “it depends how you count” is not an answer. The workable approach is to build the estimate in the open, so the assumption is visible and can be argued with.
The four inputs
- 1Incremental enquiries, counted the same way each month, from a source you can identify.
- 2Close rate on those enquiries, from your own CRM rather than an industry average.
- 3Average order value or contract value for the segment being counted.
- 4An attribution weight — the share of a closed deal you are willing to credit to this channel.
Three rules that keep it honest
- Write the weight down before you compute. A weight chosen after seeing the result is not a weight, it is a conclusion.
- Test the sensitivity. If the weighting halves, what happens to the answer? If the conclusion flips, say so.
- Show the range, not the point estimate. A single figure implies precision that survey-based work does not have.
Weights for a position inside an answer (first mention versus later ones, say) are a modelling choice. Useful for internal comparison; not evidence, and not a measurement.
What we will not claim
- Thresholds presented as standards. A target such as “indexed above 80%” is somebody's target, not an industry benchmark; it is only meaningful against your own previous round.
- Case studies as results. The scenarios on this site are illustrative; they are not customer outcomes and do not imply that a similar client got a similar number.
- Attribution weights as measurements. Any weighting you apply to estimate value is an assumption. It can be useful and it is still an assumption.
- That stage four assets create stage one demand. A better report does not make a category start being asked about.
- A date by which answers will change. Model refreshes are not on a supplier's schedule.
What we do commit to is the record: the same question set asked the same way each round, the original answers saved, and a log of what changed in between. You can audit that without our help — the artifacts and their checks are on the deliverables page, and the mechanism behind each one is on the method page.
Which stage are you stuck at?
Send your website, the market you sell into, and one sentence about what you have already tried. We will tell you which stage we think is actually blocking you — including when the answer is that the block is earlier than the one you called about.
We answer in English. Invoicing currency and payment terms are specified in the proposal.
Operated by 南宁超级月亮跨境电商有限责任公司 (registered company name in Chinese).