Automatic performance-triggered campaign adjustment
Automatic performance triggered campaign adjustment. A hierarchical feature tree is generated. Each child node's feature is more specific than its respective parent node's feature. The discovery system creates a behavioral model comprising features of the feature tree which is used in the operation of an advertising campaign. A degraded model feature is detected at the discovery system by comparing a performance metric of a model feature from two different time windows. The discovery system matches a node of the feature tree with the degraded feature and selects a prospective model feature from an ancestor node of the matching feature's node. An estimated performance metric for the prospective model feature is determined and the results are used to decide if the prospective model feature should be incorporated into an updated model or not. The model can be updated with a new model feature selected from one or more prospective model features.
1. A computer-implemented method comprising:
generating a feature tree comprising a root node and a plurality of hierarchically arranged nodes, each node corresponding to a feature of browser histories, each child node corresponding to a feature of increased specificity over a respective parent node's feature;
creating a model, the model configured to assess a browser's suitability for receiving advertising content of an advertising campaign based on features of the browser's history, the model comprising a plurality of features represented in the feature tree;
operating the advertising campaign using the model, the operating comprising;
selecting a plurality of browsers suitable for the advertising campaign according to a result of applying the model to the respective browser's history;
sending advertising content of the advertising campaign to the first set of browsers;
detecting a degraded feature, the detecting comprising:
comparing a campaign performance metric of a feature of the model from a first time window to a campaign performance metric of the feature of the model from a second time window that is more recent than the first time window, the campaign performance metric an indication of the predictive value of the feature for the success of the campaign; and
responsive to a decrease in the campaign performance metric of the feature of the model from the first time window to the second time window, identifying the feature of the model as a degraded feature; and
matching a feature of the feature tree with the degraded feature;
responsive to detecting a degraded feature, automatically updating the model, at the discovery system, comprising:
discovering nodes of decreased specificity over the matching feature's node by discovering nodes from the feature tree which are closer to the root node of the feature tree than the matching feature's node; and
adding a feature of a node selected from the discovered nodes, to the model; and
deleting the degraded feature from the model;
operating an updated advertising campaign, the operating comprising:
responsive to receiving a notification, from a real-time advertising exchange,
of an opportunity to send advertising content to an available browser, applying the updated model to the available browser's history; and
sending advertising content of the advertising campaign to the available browser according to a result of applying the updated model, before the opportunity expires.
2. The method of claim 1 wherein operating an updated advertising campaign comprises:
selecting a bid price based on the result of applying the updated model.
3. The method of claim 1 wherein operating an updated advertising campaign comprises:
selecting advertising content based on the result of applying the updated model.
4. The method of claim 1 wherein operating an updated advertising campaign comprises:
customizing advertising content based on the result of applying the updated model.
5. The method of claim 1 wherein:
the new feature corresponds to an ancestor node of the matching feature's node.
6. The method of claim 1 wherein:
the feature tree is generated before the model is created; and
creating the model comprises creating the model with features of the feature tree comprising at least one feature which corresponds to a child node in the feature tree.
7. The method of claim 1 wherein:
the model is created before the feature tree is generated; and
generating the feature tree comprises generating nodes corresponding to features of the feature tree wherein at least one child node in the feature tree corresponds to a feature of the model.
8. The method of claim 1 wherein updating further comprises:
selecting a prospective model feature from the features corresponding to the nodes of the feature tree;
creating a prospective model comprising the prospective model feature;
estimating a performance metric of the prospective model feature by simulating campaign operation with the prospective model.
9. The method of claim 8 further comprising:
selecting the new feature from one or more prospective model features based on the results of estimating.
10. The method of claim 1 wherein a second new feature comprises a feature of a descendant of an ancestor node of the degraded model feature.
11. The method of claim 1 wherein a second new feature comprises a feature of a descendant of the degraded model feature's node.
12. The method of claim 1 further comprising:
searching for an updated feature.
13. The method of claim 12 wherein:
searching is initiated responsive to detecting the degraded feature.
14. The method of claim 12 wherein:
searching is initiated responsive to detecting a change in the performance of the advertising campaign.
15. The method of claim 12 wherein:
searching is initiated responsive to detecting a change in the performance of at least one model feature.
16. The method of claim 12 wherein:
the updated feature comprises a distinctive feature found in post-conversion histories of converters with a frequency which is different from a frequency of the distinctive feature's occurrence in the histories of a standard population.
17. The method of claim 12 wherein searching comprises:
searching for a feature which is increasing in popularity in a population.
18. The method of claim 12 further comprising:
adding a node representing the updated feature to the feature tree.
19. A nontransitory computer readable storage medium including computer program instructions that, when executed, cause a computer processor to perform operations comprising:
generating a feature tree comprising a root node and a plurality of hierarchically arranged nodes, each node corresponding to a feature of browser histories, each child node corresponding to a feature of increased specificity over a respective parent node's feature;
creating a model, the model configured to assess a browser's suitability for receiving advertising content of an advertising campaign based on features of the browser's history, the model comprising a plurality of features represented in the feature tree;
operating the advertising campaign using the model, the operating comprising;
selecting a plurality of browsers suitable for the advertising campaign according to a result of applying the model to the respective browser's history;
sending advertising content of the advertising campaign to the first set of browsers;
detecting a degraded feature, the detecting comprising:
comparing a campaign performance metric of a feature of the model from a first time window to a campaign performance metric of the feature of the model from a second time window that is more recent than the first time window, the campaign performance metric an indication of the predictive value of the feature for the success of the campaign; and
responsive to a decrease in the campaign performance metric of the feature of the model from the first time window to the second time window, identifying the feature of the model as a degraded feature; and
matching a feature of the feature tree with the degraded feature;
responsive to detecting a degraded feature, automatically updating the model, at the discovery system, comprising:
discovering nodes of decreased specificity over the matching feature's node by discovering nodes from the feature tree which are closer to the root node of the feature tree than the matching feature's node; and
adding a feature of a node selected from the discovered nodes, to the model; and
deleting the degraded feature from the model;
operating an updated advertising campaign, the operating comprising:
responsive to receiving a notification, from a real-time advertising exchange, of an opportunity to send advertising content to an available browser, applying the updated model to the available browser's history; and
sending advertising content of the advertising campaign to the available browser according to a result of applying the updated model, before the opportunity expires.
20. A system comprising:
a computer processor; and
a computer readable storage medium soring processor-executable computer program instructions, the computer program instructions comprising instructions for:
generating a feature tree comprising a root node and a plurality of hierarchically arranged nodes, each node corresponding to a feature of browser histories, each child node corresponding to a feature of increased specificity over a respective parent node's feature;
creating a model, the model configured to assess a browser's suitability for receiving advertising content of an advertising campaign based on features of the browser's history, the model comprising a plurality of features represented in the feature tree;
operating the advertising campaign using the model, the operating comprising;
selecting a plurality of browsers suitable for the advertising campaign according to a result of applying the model to the respective browser's history;
sending advertising content of the advertising campaign to the first set of browsers;
detecting a degraded feature, the detecting comprising:
comparing a campaign performance metric of a feature of the model from a first time window to a campaign performance metric of the feature of the model from a second time window that is more recent than the first time window, the campaign performance metric an indication of the predictive value of the feature for the success of the campaign; and
responsive to a decrease in the campaign performance metric of the feature of the model from the first time window to the second time window, identifying the feature of the model as a degraded feature; and
matching a feature of the feature tree with the degraded feature;
responsive to detecting a degraded feature, automatically updating the model, at the discovery system, comprising:
discovering nodes of decreased specificity over the matching feature's node by discovering nodes from the feature tree which are closer to the root node of the feature tree than the matching feature's node; and
adding a feature of a node selected from the discovered nodes, to the model; and
deleting the degraded feature from the model;
operating an updated advertising campaign, the operating comprising:
responsive to receiving a notification, from a real-time advertising exchange, of an opportunity to send advertising content to an available browser, applying the updated model to the available browser's history; and
sending advertising content of the advertising campaign to the available browser according to a result of applying the updated model, before the opportunity expires.