User Insights

Nefta's core deep learning models ingest partner first party data, understand user value and deliver user level insights back via the AdSDK. The insights are used to dynamically set user-level bid floor prices to optimise all bidding network ad revenue and customise the game or app for the specific user.

User insights provided in the AdSDK:

• Predicted user churn / retention: The probability of this specific user to churn at a specific day. The available periods are: 1d, 3d, 7d, 14d, 30d. The churn probabilities are returned alongside a churn confidence score of low, medium and high.
• Predicted user ad impression value: The predicted market price eCPM value the partner would receive to show a banner, interstitial or rewarded ad to this user. This bid landscape model includes user advertising revenue, device, segment, ad format and other data to produce the prediction.
• Predicted user value: The predicted dollar value of a user in: predicted IAP value, predicted Ad value and predicted total value (pLTV).
• Predicted user bid floors: The predicted bid floor prices for a user for a ad format.

Dynamic user-level bid floors

By combining the insights of user predicted total value and predicted user ad impression value, a user value spread is calculated and used to set a bid floor for bidding networks to bid above. With this method, partners can optimise ad revenue across all their ad networks.

User insight integration

To start receiving the user insights, integrate the AdSDK and then activate the function.