Status: instrumented; production observation and five moderated sessions remain open.
The public baseline uses consent-gated GA4 events and never attaches a user ID,
email address, organization ID, or professional-workspace route. Browser-local
markers distinguish a first artwork completion from a visit at least 24 hours
later; clearing browser storage resets those markers.
| Event | Outcome | Trigger |
|---|---|---|
| `first_time_artwork_completion` | First-time completion | First consented artwork route reached in a browser. |
| `artwork_to_artwork_continuation` | Artwork continuation | Visitor follows previous, next, related, or curated-connection artwork link. |
| `return_visit` | Return visit | A consented public visit occurs at least 24 hours after the prior stored visit. |
| `public_domain_download` | Rights-safe download | Visitor selects the image download control, which is rendered only for an exact item-level public-domain or CC0 image statement. |
| `share` | Share | Native share completes, or the clipboard fallback successfully copies the canonical artwork URL. Cancellation and errors are not counted. |
The initial report must include event counts, eligible consented sessions, the
measurement window in UTC, and known gaps. Do not calculate conversion rates
when the denominator is unavailable, and do not treat GA4 page views as any of
the five outcome events.
Import real evidence with:
pnpm pd0:evidence -- --input=path/to/pd0-evidence.json --check
The input object requires `measurementWindow` (`startedAt`, `endedAt`,
`eligibleConsentedSessions`, and `source`), integer counts for all five event
names, `usabilitySessions`, and an optional `productOwnerReview`. Each session
uses only an opaque `participantCode`, UTC completion time, `nonSpecialist: true`,
`consented: true`, the measured homepage exposure in milliseconds, two separate
five-second results for offer and primary-next-action comprehension, the four
journey comprehension/completion results (`yes`, `with-help`, or `no`), and a
friction note. Quotes are optional and limited to 20 words. Direct identifier
fields, duplicate codes, exposure above 5,000 ms, negative counts, invalid
windows, and synthetic completion are rejected. `--check` exits unsuccessfully
until five sessions exist, at least four participants independently score `yes`
for both five-second questions, and an approved, attributable product-owner
review exists.
Cross-origin image hosts may cause the browser to open the original image before
the visitor saves it. The download event therefore measures selection of a
rights-qualified download action, not confirmation that a file reached disk.
Non-specialist usability protocol
Recruit five adults who do not work professionally with museum metadata,
collections systems, Linked Art, or cultural-heritage APIs. Record only a
participant code and coarse experience note—no names or contact details.
Ask each person to:
- Show the canonical homepage for no more than five seconds, then hide it.
- Ask, without prompting, “What does this website offer?”
- Ask, without showing the page again, “What is the main thing it asks you to
do next?”
- Restore the page and ask the participant to find an artwork that interests
them and say where it came from.
- Ask whether its image can be reused.
- Ask them to continue to a second artwork.
- Ask them to follow the Flowers connection and describe what is claimed and
not claimed.
For each session, record the actual exposure duration; separate unassisted
offer and primary-action results; journey completion (`yes`, `with-help`, or
`no`); notable friction; rights/source comprehension; connection-boundary
comprehension; and one verbatim phrase of at most 20 words with permission. Do
not convert a `with-help` response into a five-second pass. A product owner must
sign the five-session summary before PD3 expands beyond the first journey.