Attribution Analysis
Attribution Analysis (Attribution) measures the contribution of each touchpoint (channel/source page) to conversion, answering "who drove the conversion". It corresponds to the backend AnalyticsService.Attribution.
1. What Questions It Answers
- Which channel brings the most paying users?
- Which source page delivered the "final strike" leading to conversion?
- How much do channel contributions differ between first touch and last touch?
2. Attribution Models
| Model | Meaning | Use Case |
|---|---|---|
Last touch last_touch (default) | Credits the conversion to the user's last touchpoint before converting | Short decision cycles, emphasis on the final push |
First touch first_touch | Credits the conversion to the user's first touchpoint | Looking at acquisition sources, brand reach |
Attribution dimensions: channel (default) / referer (source page).
3. Backend API
gRPC: AnalyticsService.Attribution
| Field | Type | Description |
|---|---|---|
time_range | TimeRange | Analysis time range |
conversion_event | string | Conversion event name (e.g. pay_success) |
dimension | string (optional) | Attribution dimension: channel (default) / referer |
model | string (optional) | Attribution model: last_touch (default) / first_touch |
app_id | uint32 (optional) | Filter by app |
Response AttributionResponse: buckets[] (AttributionBucket contains label touchpoint value / converter_uv converting users / percentage share), model, dimension, total_converters total converting users.
HTTP (admin proxy)
POST /admin/v1/analytics/attribution
Content-Type: application/json
{
"timeRange": { "startMs": 1718169600000, "endMs": 1718774399000 },
"conversionEvent": "pay_success",
"dimension": "channel",
"model": "last_touch"
}
4. Typical Scenarios
Channel ROI attribution
conversionEvent=pay_success, dimension=channel: see which channel has the highest share of converting users, compute ROI against acquisition costs, and optimize budget allocation.
Landing page effectiveness
dimension=referer: see which page users last came from before converting, evaluating the conversion pull of each landing/campaign page.
First vs last touch comparison
Run it once with first_touch and once with last_touch, then compare how channel rankings shift: some channels are good at acquisition (high first-touch share) but weak at closing (low last-touch share), and vice versa.
5. Notes and Caveats
- A conversion event is required:
conversion_eventis mandatory — without it, "conversion" cannot be defined. - Last touch favors closing channels: direct entries (e.g. organic search) tend to be overestimated under last touch.
- First touch depends on the first record: it requires that a user's behavior has been collected since their very first visit; otherwise first touch is inaccurate.
- Attribution ≠ causation: attribution is a correlation measure; a high-share channel is not necessarily the true cause of conversion.