Before building a 1,000-company Japan Buyer Evidence Index, I tested the method on 50 English-operating B2B technology companies exhibiting at SusHi Tech Tokyo 2026.
The pilot was deliberately varied rather than statistically representative. Its purpose was to test whether public websites could be collected responsibly and whether the proposed fields measured anything more useful than technical configuration.
Translation makes a product readable. Buyer evidence makes it verifiable.
WHAT WAS STUDIED
A maximum-variation feasibility sample
Every selected company appeared in the official exhibitor directory, supplied a public website, operated from an English-speaking or English-operating market, and offered a B2B technology or technology-enabled product or service.
| Region | Companies |
|---|---|
| North America | 20 |
| United Kingdom and Ireland | 15 |
| Singapore | 10 |
| Australia | 5 |
Population source: SusHi Tech Tokyo 2026 exhibitor directory ↗
WHAT COULD BE COUNTED
The language layer was visible. The decision layer was not.
| Technical outcome | Companies |
|---|---|
| Analysable HTTP 200 HTML homepage | 42 |
| Stopped under applicable robots rules | 5 |
| Non-200 homepage | 2 |
| Fetch error after retry | 1 |
Among the 42 analysed homepages, 11 exposed at least one Japanese-site signal: Japanese text, Japanese hreflang, a Japanese-language route or a Japanese language declaration.
A structured review of ten deliberately contrasting companies then looked beyond route presence.
| Manual observation | Selected cases |
|---|---|
| Japanese product/use-case explanation observed | 8 / 10 |
| Complete Japanese purchase path observed | 3 / 10 |
| Any Japan-local operations evidence, including partial | 2 / 10 |
| Any Japan-specific customer/partner evidence, including partial | 1 / 10 |
| Mixed-language or awkward Japanese flag | 6 / 10 |
The unresolved buyer questions
- Who is responsible for implementation and support in Japan?
- Is there a Japanese entity, office, distributor or reseller?
- Which Japanese customer has used the product, and with what result?
- Can the local buying team find security, privacy, compliance or data-location evidence?
- Does the Japanese purchase path match the global offer?
INTERPRETATION LIMIT
What this study does not claim
The 50-company sample was not random. The ten manual cases were selected to stress the rubric, and their coding has not yet passed a timed reliability check. These counts are hypothesis-generating observations, not prevalence estimates for all exhibitors or all foreign companies entering Japan.
NEXT GATE
Reliability before scale
The next unit of work is a timed manual review, a blind re-code and a revised field dictionary. Only after that gate should the larger population frame and public study window be locked.
See the research method