Prediction of content performance in content delivery based on presentation context

1. A computer-implemented method comprising:

accessing baseline prediction models configured to predict future performance of detectible triggers of presentation opportunities in a first stage in one or more content delivery auctions, without considering presentation opportunity characteristics, using analytics data indicating historical performance of the detectible triggers;

learning an adjustment factor of a presentation context factor model based on a residual between predicted values of the baseline prediction models and actual values associated with a subset of the presentation opportunity characteristics associated with the detectible triggers of presentation opportunities, wherein the presentation context factor model is configured to learn the adjustment factor in a second stage that quantifies a contribution of the subset of the presentation opportunity characteristics to the future performance;

computing a contextual predicted value using the adjustment factor for the subset of the presentation opportunity characteristics; and

transmitting a performance prediction that is based on at least the contextual predicted value to a user device.

2. The method of claim 1 , wherein the detectible triggers of presentation opportunities are selected from a content delivery campaign of a user associated with the user device.

3. The method of claim 1 , further comprising generating context model data of the presentation context factor model from predicted values of a plurality of the baseline prediction models.

4. The method of claim 1 , wherein the subset of the presentation opportunity characteristics comprise one or more of a seasonality content factor, a device context factor, a geographic context factor, and an audience context factor of presenting content.

5. The method of claim 1 , wherein the baseline prediction model predicts the performance metric for the detectible trigger as a function of bid amount.

6. The method of claim 1 , further comprising:

optimizing a plurality of bids on the detectible triggers within a budget defined by a user; and

submitting the optimized plurality of bids to a content delivery auction over one or more network communications, the submitting causing at least one content item to be presented on a client device in the content delivery auction based on one or more of the optimized plurality of bids.

7. The method of claim 1 , wherein the baseline prediction models predict the future performance of the detectible triggers for a user and the analytics data is used in the baseline prediction models based on indicating the historical performance of the user in the one or more content delivery auctions.

8. The method of claim 1 , further comprising constructing the baseline prediction models for the detectible triggers based on identifying the detectible triggers as being part of a content delivery campaign of a user, wherein the historical performance is of the user in the content delivery campaign.

9. The method of claim 1 , further comprising receiving the analytics data from an analytics provider over one or more network communications, wherein the analytics data comprises actual values of a performance metric of the detectible triggers based on historical bids made in the one or more content delivery auctions.

10. The method of claim 1 , further comprising:

tracking performance of keywords of a content delivery campaign of a user in a content delivery auction of the one or more content delivery auctions; and

receiving the analytics data that provides actual values of a performance metric of the keywords based on historical bids made in the content delivery auction, wherein the detectible triggers comprise the keywords.

11. A computer-implemented system for predicting content performance in content delivery auctions, the system comprising:

one or more processors; and

one or more memories comprising program instructions stored thereon that are executable by the one or more processors to cause the one or more processors to perform operations comprising:

accessing baseline prediction models configured to predict future performance of detectible triggers of presentation opportunities in a first stage in one or more content delivery auctions, without considering presentation opportunity characteristics, using analytics data indicating historical performance of the detectible triggers;

learning an adjustment factor of a presentation context factor model based on a residual between predicted values of the baseline prediction models and actual values associated with a subset of the opportunity characteristics associated with the detectible triggers of presentation opportunities, wherein the presentation context factor model is configured to learn the adjustment factor in a second stage that quantifies a contribution of the subset of the presentation opportunity characteristics to the future performance;

computing a contextual predicted value using the adjustment factor for the subset of the presentation opportunity characteristics; and

transmitting a performance prediction that is based on at least the contextual predicted value to a user device.

12. The system of claim 11 , wherein the detectible triggers are from a content delivery campaign of a user associated with the user device.

13. The system of claim 11 , wherein the operations further comprise generating context model data of the presentation context factor model from predicted values of a plurality of the baseline prediction models.

14. The system of claim 11 , wherein the subset of the presentation opportunity characteristics comprise one or more of a seasonality content factor, a device context factor, a geographic context factor, and an audience context factor of presenting content.

15. The system of claim 11 , wherein the baseline prediction model is configured to predict the performance metric for the detectible trigger as a function of bid amount.

16. One or more non-transitory computer-storage media storing computer-useable instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform a method comprising:

accessing baseline prediction models configured to predict future performance of detectible triggers of presentation opportunities in a first stage in one or more content delivery auctions, without considering presentation opportunity characteristics, using analytics data indicating historical performance of the detectible triggers;

learning an adjustment factor of a presentation context factor model based on a residual between predicted values of the baseline prediction models and actual values associated with a subset of the presentation opportunity characteristics associated with the detectible triggers of presentation opportunities, wherein the presentation context factor model is configured to learn the adjustment factor in a second stage that quantifies a contribution of the subset of the presentation opportunity characteristics to the future performance;

computing a contextual predicted value using the adjustment factor for the subset of the presentation opportunity characteristics; and

transmitting a performance prediction that is based on at least the contextual predicted value to a user device.

17. The one or more computer-storage media of claim 16 , wherein the detectible triggers are from a content delivery campaign of a user associated with the user device.

18. The one or more computer-storage media of claim 16 , wherein the method further comprises generating context model data of the presentation context factor model from predicted values of a plurality of the baseline prediction models.

19. The one or more computer-storage media of claim 16 , wherein the subset of the presentation opportunity characteristics comprise one or more of a seasonality content factor, a device context factor, a geographic context factor, and an audience context factor of presenting content.