Skip to content

Risk and matching weights ​

These numbers control how strict visitor and site checks are, and whether ad picking prefers bid, CTR, or random exploration.

Weights are compared with each other. Do not set one to 100 and the rest to 0 unless you want matching to collapse onto a single factor.

Traffic Risk Level ​

FieldValue
Admin nameTraffic Risk Level
Keytraffic.risk.level
ControlSlider
FormatInteger 0–100
Example10

What it does ​

Detection strength when scoring a visit’s environment. Higher values are more sensitive to odd browsers, emulators, and rapid environment changes, so visitor risk scores rise more easily.

That score feeds later blocking and site quality.

Impact ​

  • Raise: more fraud is caught; ordinary users on privacy browsers or corporate networks may also be hit, so fill and impressions can fall.
  • Lower: more edge traffic is allowed; invalid clicks may rise.
  • Move in small steps (5–10) and watch the block mix on Realtime.

Site Fraud Risk Weight ​

FieldValue
Admin nameSite Fraud Risk Weight
Keysite.fraud.weight
ControlSlider
FormatInteger 0–100
Example50

What it does ​

How much of the site quality score comes from visitor risk (anti-fraud). Lower average visitor risk means a better site.

Together with Site CTR Weight it builds the site score. That score sets the site’s matching position and where it sits in the platform take range.

Impact ​

  • Raise: punishes sites with poor visitor quality even if CTR looks high.
  • Lower: more tolerant of risky traffic; useful when you need volume first.
  • The two weights do not have to sum to 100, but both should stay in a sensible range. Avoid leaving one at 0.

Site CTR Weight ​

FieldValue
Admin nameSite CTR Weight
Keysite.ctr.weight
ControlSlider
FormatInteger 0–100
Example50

What it does ​

How much historical CTR counts in the site score. CTR uses impressions and clicks in the analysis window. If impressions are below the sample floor, the site score is 0. See CTR Minimum Sample Size.

Impact ​

  • Raise: high-CTR sites match better ads.
  • If click fraud is common, raise fraud weight as well, or you will promote farms.

Ad Bid Weight ​

FieldValue
Admin nameAd Bid Weight
Keyad.bid.weight
ControlSlider
FormatInteger 0–100
Example60

What it does ​

Two layers:

  1. Which band to match: if bid weight is greater than CTR weight, the system matches the site against bid percentiles; otherwise against ad CTR percentiles.
  2. Final ranking: higher bids score higher; a larger weight leans more on price.

Impact ​

  • Well above CTR weight: expensive ads win, eCPM up, cheaper high-performing ads may be ignored.
  • Below CTR weight: matching uses the performance band instead.
  • Combined with CTR weight and the random factor in the pick score.

Ad CTR Weight ​

FieldValue
Admin nameAd CTR Weight
Keyad.ctr.weight
ControlSlider
FormatInteger 0–100
Example40

What it does ​

How much historical CTR matters when picking ads. New ads with too little sample have CTR score 0 and need the random factor or exploration to get a chance.

If this is greater than bid weight, the matching window is cut on CTR percentiles, not bid percentiles.

Impact ​

  • Too high: head high-CTR ads monopolize, cold start gets harder.
  • Too low: almost price-only; weak creatives with high bids fill the slot.

Ad Random Factor Weight ​

FieldValue
Admin nameAd Random Factor Weight
Keyad.random.weight
ControlSlider
FormatInteger 0–100
Example20

What it does ​

Adds jitter on top of bid and CTR so ranking is not fully locked by head ads, and other ads still get some impressions.

Impact ​

  • Raise: more spread, head eCPM may fall, better for testing new ads.
  • Set to 0: almost always the current top scorers.
  • Keep it as a helper; do not let it dwarf bid and CTR weights.

Risk Chain Strength ​

FieldValue
Admin nameRisk Chain Strength
Keyrisk.chain.strength
ControlSlider
FormatInteger 0–100
Example5

What it does ​

After a suspicious visit, the system walks related devices on the same IP and IPs previously used by the same device. This value is how many hops to walk.

Deeper chains mark clusters together. That helps against organized fraud.

Impact ​

  • Raise: wider coverage, more collateral damage (NAT, campus networks).
  • Lower: only the immediate hop; device farms that rotate IPs slip through.
  • The example value 5 is a small neighborhood. Do not crank it high without a reason; the graph grows quickly.

Operator guide for AdChainX Admin.