The one-pager is the last page of every study report: where you stand, and where to act first. This one shows the structure with an invented brand so there are no confidentiality questions.
Sample: Contapro, an ERP for Mexican small and mid-sized businesses.
The study opens on the committee’s questions — “What is the best ERP for SMEs?”, “ERP for invoicing in Mexico” — asked the way a buying committee asks them, months before the first vendor call, in a tab the vendor never sees. None of those questions arrive in the CRM; the absence happens there.
Invented brand, invented competitors, invented figures. The structure is real; the numbers belong to nobody.
Study base: 720 responses read (3 assistants x 3 repetitions). Every rate below prints its denominator.
The four outcome rates
| Rate | Result |
|---|---|
| Mentioned | 29% (209 of 720 responses) |
| Recommended | 9% (65 of 720 responses) |
| Named as the best | 2% (14 of 720 responses) |
| With a purchase path | 5% (36 of 720 responses) |
Reading: Contapro appears in three of every ten answers, but presence is not preference. Fewer than one in ten answers recommends it, almost none call it the best, and only one in twenty hands the committee a route to a demo. In a considered purchase the shortlist is built inside the answers, before any vendor is contacted. The gap between mentioned and recommended is the work.
Share of category conversation
Share of all brand mentions recorded across the study:
| Brand | Share |
|---|---|
| Administra | 27% |
| Softek | 23% |
| Contapro | 15% |
| The open-source ERP | 19% |
| The spreadsheet stack (“Excel and a shared drive”) | 16% |
Reading: the models set a free open-source project and a spreadsheet setup at the same table as the paid systems. This is the interchangeability trap, and it is why the competitor set has to include who the models treat as rivals, not who the sales team respects.
Mention rate by assistant type
Same brand, three different machines, reported separately:
| Assistant type | Mention rate |
|---|---|
| Memory-only call | 13% (31 of 240) |
| Search-then-answer | 31% (74 of 240) |
| Google AI Mode | 43% (104 of 240) |
Reading: “Inside the model’s memory it lists categories, not vendors; the open web is what names the brand and what recommends it.” The lever is not the product site.
Where the answers come from
Share of all citations recorded across the study:
| Source | Share | Who controls it |
|---|---|---|
| sat.gob.mx | 28% | The regulator. Cannot be edited. |
| comparador-erp.mx | 24% | The comparison site. Vendor profile: editable. |
| reddit.com/r/erp | 18% | Earned. Community. |
| elcontribuyente.mx | 16% | Earned. Trade press. |
| The open-source ERP’s documentation | 14% | Earned. Controlled by the rival’s community. |
More than half of what assistants say about this category comes from the tax authority’s compliance pages and one comparison profile the brand can edit. The committee meets the regulator before it meets the vendor.
Where to act first
Three actions, three different natures:
- This quarter, without asking permission: claim and rewrite the comparison-site profile. comparador-erp.mx carries a quarter of the citations and its vendor pages belong to the editable column.
- Editorial and PR: the committee’s evaluation questions — migration, invoicing, multi-company, licensing — have the weakest presence and the most available citations. Earn the trade press and the community threads: elcontribuyente.mx, reddit.com/r/erp. Named reference customers count double here.
- Accept and surround: sat.gob.mx cannot be displaced. Publish the CFDI and e-invoicing guides that sit next to it and quote it.
Notes
- All figures are from the invented Contapro sample; the invented-sample note above travels with the asset wherever it is embedded.
- Denominator note: the share tables state their base in words (all brand mentions, all citations recorded); raw mention and citation counts must be added to these tables before publication to satisfy the denominator rule with counts, not labels only.
- Study base line prints 720 responses; the per-assistant split is 240 each. Question count deliberately not stated (not in canon).
Where to go next
The same structure on a consumer brand: the Solaria sample one-pager. Or tell us your category and we will scope the real version.