Quality Scoring
About Quality Scores
Quality Score is a measure of relevancy that influences ad rank/position and cost of an ad.
Koddi Ads provides a real-time auction as well as a quality score system to enable publishers to optimize ad results for margin potential, ad quality, seller score, click/conversion propensity and more.
To determine the position of an ad, each ad is ranked using a process which takes into account the bid and the quality score. Ads are then listed in descending order based on the result of that equation.
Quality scoring in ad programs is a well-established industry practice.
Koddi Ads Quality Score
The Koddi Ads quality score is built by Koddi using several machine learning algorithms and our decade of experience as a company in running ads programs. We continually adjust this scoring toward a win win win scenario – a win for end users, advertisers, and publishers.
The general formula is:
Bid Amount x Koddi Ads Quality Score = Bid Weight
Publisher Quality Score
Publishers can also optionally send a Quality Score to the auction engine to influence the results of the auction. The most common use cases are organic sort and propensity to convert, but more are listed below. This attribute (passed in the real time in the POST winning ads call) allows you the publisher control of the results of the auction. We suggest a scale of either 0 – 1 or 0 – 100, as long as is it applied consistently, either will work well. Koddi does require that all possible publisher score values are positive.
When Publisher score is added, it is used as a major input to the predictive model to maximize the intended KPI, such that:
Bid Amount x Quality Score f(Publisher Quality Score) = Bid Weight
Publisher Quality Score Use Cases
| Use Case | Comments |
|---|---|
| Organic sort | The scoring is based on the organic sort rank – which generally includes personalization, propensity to convert, marketplace commission margin and bidder 'quality' all in one. This approach uses your internal competitive advantages and the investments in your own sort to enhance your end user experience. This is generally our recommended option to start if this data is available. |
| Propensity to convert | The scoring is based on the propensity to convert of the specific user, the bidder or (ideally) the user for that specific bidder. This approach ensures the ads are as relevant as possible for the end user. This is generally our #2 recommended option to start if this data is available. |
| Marketplace commission margin | The scoring is based on the potential marketplace commission of the advertiser. Doing this helps optimize the total value of the search to the publisher. Commission is usually a factor of the organic sort, so this item is usually covered there. |
| Bidder ‘quality’ | The scoring is based on the 'quality' of the bidder (star rating/user reviews). Doing this ensures high quality bidders are shown first, generally improving the user experience and conversion rate. |
Publisher Quality Score Recommendations
| 1 | To start, use what is readily available |
| 2 | Update and refine over time – test and learn |
| 3 | Make sure you consider end users, advertisers and your own goals |
Advanced Models
Our data science team is constantly testing and iterating quality score models to maximize auction performance. For more advanced modeling inputs, or specific requests about how an input is utilized or weighted in the quality score, please contact your Koddi team.
Updated over 2 years ago