English
Quality scoring
Ad and site scores are computed from recent impressions and clicks, not entered by hand. These two settings decide how far back to look and when a sample is too small to trust.
Scores feed matching position, platform take, and premium-ad selection. A change does not recompute everything instantly; rating jobs catch up over time.
CTR Minimum Sample Size
| Field | Value |
|---|---|
| Admin name | CTR Minimum Sample Size |
| Key | ctr.minimum.sample.size |
| Control | Text |
| Format | Positive integer (at least 1) |
| Example | 500 |
What it does
When scoring an ad or a site, if impressions in the window are below this number, the system treats the sample as too small and sets the quality score to 0.
That stops “3 impressions and 1 click” from looking like a high-CTR object.
How to fill it
Enter an impression floor, for example 500.
Impact
- Raise: new ads and sites stay at 0 longer (slower cold start, stabler scores).
- Lower: scores appear sooner, and a few accidental or fraudulent clicks can inflate them.
- Read with Score Analysis Range: a short window plus a high floor leaves many objects at 0.
Score Analysis Range
| Field | Value |
|---|---|
| Admin name | Score Analysis Range |
| Key | score.analysis.range |
| Control | Text |
| Format | Positive integer, unit: days |
| Example | 30 |
What it does
When scoring ads, sites, and per-category performance on a site, the system looks at this many recent days of impressions and clicks.
Site category recommendation weights use the same window.
How to fill it
Days only, for example 30. Do not type 30 days.
Impact
- Longer: smoother scores, slower reaction to sudden quality changes.
- Shorter: follows recent performance, more volatile.
- Shortening the window can push objects under the sample floor, so scores drop to 0. Check recent volume first.