Data measurement definitions
In the event of any ambiguity among the Chinese, English and Korean versions, the Chinese version prevails.
We explain every number supplied externally: how it is calculated, how long it is measured, how it is deduplicated and what is excluded. This page is updated as the service changes.
Core metric definitions
| Metric | Definition |
|---|---|
| Page view (PV) | One successful page load received as a view event by the server. |
| Unique visitor (UV) | Visitors deduplicated by a one-time anonymous device identifier. The identifier is used for on-site counting only. |
| Valid try-on user | A deduplicated user who submitted at least one generation successfully and received a result image during the reporting period. |
| Try-on submission | A click on “Start try-on” that enters the generation queue successfully. |
| Result view | A result image that finishes loading and enters the visible area. |
| Impression | Count 1 impression when a partner product card, showcase or banner completes an agreed display. Count each display event once; deduplicate repeated reports by event identifier. |
| Click | Count 1 click when a user clicks partner content or its entry point. Count each click event once; deduplicate repeated reports by event identifier. |
| Click-through rate (CTR) | Clicks divided by impressions; the numerator and denominator use the same partner placement, reporting period and deduplication rules. |
| Successful generation | Count 1 successful generation when a try-on task returns a complete, viewable result. Deduplicate by task ID; refreshing the page or viewing history does not add another count. |
| Outbound event | Count 1 outbound event when a partner link triggers an exit event and UTM alignment is complete. Deduplicate by link and task event. External sales are not automatically determined on-site unless reporting back is agreed in the contract. |
When the denominator for the same placement, period and deduplication rules is 0, display the click-through rate as “—”, not 0% or 100%.
General measurement rules
- Reporting period: summaries use calendar day, week or month by default, with start and end dates shown; UTC+8 is used across time zones.
- Deduplication: unique metrics use the anonymous on-site device identifier; refreshing on the same device does not add a visitor.
- Exclusions: health checks and internal test traffic are excluded from external data. Clearly abnormal high-frequency automated requests are marked by risk controls and excluded or listed separately.
- CTR: the numerator is click events and the denominator is impression events under the same placement, period and deduplication rules; people are counted separately and are not used for CTR.
- Product event fields: product events carry eventId, UTC+8 time, pageViewId, slot, productId, campaignId, sourceClass and destinationChannel. They do not carry photos, email addresses or buyer identity.
- Organic versus incentivized: invitation, check-in and sponsored try-on-credit use are shown in separate source columns from organic users.
- Phase labels: limited or internal phases are marked clearly as internal-test data; small samples are not presented as a conversion rate.
Initial external reporting covers page and try-on funnel data, product-card clicks and outbound events, with CSV export. Fields that require an advertising SDK or affiliate callback are reported as display requests or clicks until the required capability is ready.
Live public data
The numbers below come straight from the site's own statistics API — the same source the owner's internal dashboard reads — with no key required. Definitions above govern the metrics; data is aggregated by calendar day in UTC+8. Responses are cached for about 10 minutes and contain no user-level detail.
How to verify: ① this page renders the public API's live response — open “Raw JSON” to read the same numbers at the source; ② metric definitions, deduplication and exclusions are published above, before the numbers; ③ sales and commission are reconciled against your own affiliate console (Taobao Affiliate, JD Affiliate, Vipshop Affiliate) order reports — on-site data proves displays and clicks only, never a sale.
Loading…
Partner record reconciliation
Before cooperation, confirm the report template, delivery fields and contact person. After launch, reconcile weekly and closing records for the same product, placement and time range. CSV export is available; sales sharing is reconciled separately against valid orders, refund deductions and the order report confirmed by both sides. The attachment describes deduplication, exclusions and the organic, merchant-brought, advertising and incentivized source categories. Missing records are checked first. UTM is used to align traffic sources and cannot, on its own, prove a sale.
Credits and failure definitions
- A successful try-on submission uses one credit; system, model and timeout failures return the credit automatically.
- The task status shows the failure reason, such as upload validation, queue timeout or generation failure.
- Check-ins, invitation rewards and code-based credits are recorded with their real use and campaign batch.
Advertising and recommendation transparency
- Paid product cards and advertising placements are labelled “Advertising” card by card and shown separately from organic recommendations.
- The try-on result placement and home single-product placement are fixed and do not change with makeup direction or skin analysis.
- Each placement is reconciled separately; shade and price follow the brand or sales platform page and are marked as reference prices.
Privacy boundaries
- Photos are used only to generate the try-on result requested by the user; originals and results are automatically deleted after 7 days.
- User photos and identifiable face data are not uploaded to, provided to or sold to merchants or other third parties.
- On-site statistics use one-time anonymous identifiers and do not use cross-site tracking cookies.