Scoring recommendations and explanations with a probabilistic user model

• US 7,983,952 B1
• Filed: 01/07/2010
• Issued: 07/19/2011
• Est. Priority Date: 06/03/2005
• Status: Active Grant

• Current Assignee: Versata Development Group, Inc. (ESW Capital, LLC)
• Original Assignee: Versata Development Group, Inc. (ESW Capital, LLC)
• Inventors: Dillon, Thomas H.
• Primary Examiner(s): Airapetian; Mila
• Application Number:
  US12/683,556
• Time in Patent Office: 558 Days
• Field of Search: 705/26, 705/27
• US Class Current: 705/26.7
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/06 Buying, selling or leasing transactions
• CPC Class Codes: G06Q 30/0631 Item recommendations

Abstract

A data processing system generates recommendations for on-line shopping by scoring recommendations matching the customer'"'"'s cart contents using by assessing and ranking each candidate recommendation by the expected incremental margin associated with the recommendation being issued (as compared to the expected margin associated with the recommendation not being issued) by taking into consideration historical associations, knowledge of the layout of the site, the complexity of the product being sold, the user'"'"'s session behavior, the quality of the selling point messages, product life cycle, substitutability, demographics and/or other considerations relating to the customer purchase environment. In an illustrative implementation, scoring inputs for each candidate recommendation (such as relevance, exposure, clarity and/or pitch strength) are included in a probabilistic framework (such as a Bayesian network) to score the effectiveness of the candidate recommendation and/or associated selling point messages by comparing a recommendation outcome (e.g., purchase likelihood or expected margin resulting from a given recommendation) against a non-recommendation outcome (e.g., the purchase likelihood or expected margin if no recommendation is issued). In addition, a probabilistic framework may also be used to select a selling point message for inclusion with a selected candidate recommendation by assessing the relative strength of the selling point messages by factoring in a user profile match factor (e.g., the relative likelihood that the customer matches the various user case profiles).

Claims

1. A recommendation selection system comprising a non-transitory, computer readable medium that includes code that is executable by a processor to implement:

a rule evaluator for producing a set of candidate recommendations; and

a recommendation selector that uses a probabilistic purchase decision model to rank each of the candidate recommendations using one or more purchase decision gating factors to calculate, for each candidate recommendation, a boost in expected margin resulting from the candidate recommendation being issued, as compared to the candidate recommendation not being issued.

2. The system of claim 1, further comprising a message selector that uses a probabilistic use case model to choose, for each candidate recommendation, a selling message, thereby generating a recommendation, message pair for each candidate recommendation.

3. The system of claim 2, where the recommendation selector uses the probabilistic purchase decision model to choose a candidate recommendation based on the recommendation, message pair that maximizes boost in expected margin resulting from the candidate recommendation being issued, as compared to the candidate recommendation not being issued.

4. The system of claim 1, wherein the probabilistic purchase decision model is implemented as a Bayesian network.

5. The system of claim 1, where the purchase decision gating factors for each candidate recommendation comprise a relevance measure, an expected exposure measure and an expected clarity measure.

6. The system of claim 5, where the purchase decision gating factors for each candidate recommendation further comprise a post-recommendation exposure measure and a post-recommendation clarity measure.

7. The system of claim 1, where the purchase decision gating factors for each candidate recommendation comprise a relevance measure, an expected exposure measure, an expected clarity measure, a post-recommendation exposure measure, a post-recommendation clarity measure and a session age measure.

8. The system of claim 2, where the probabilistic purchase decision model and the probabilistic use case model are implemented with a single probabilistic model.

9. An article of manufacture comprising at least one recordable medium having stored thereon executable instructions and data which, when executed by at least one processing device, cause the at least one processing device to:

use a recommendation context associated with a customer to identify a set of candidate recommendations that match a recommendation context associated with a customer;

calculate a score value for each candidate recommendation using a probabilistic model to determine a boost in expected margin value for each candidate recommendation using one or more predetermined scoring criteria;

rank the set of candidate recommendations using the score value for each candidate recommendations; and

issue at least the candidate recommendation rule having the highest score value as a purchase recommendation to the customer.

10. The article of manufacture of claim 9, wherein the executable instructions and data cause the at least one processing device to calculate the score value for each candidate recommendation by identifying a selling point message for each candidate recommendation that maximizes the purchase probability for the candidate recommendation if the candidate recommendation is issued, where the selling point message is identified using a probabilistic use case model to generate a probability distribution for possible use cases for the customer.

11. The article of manufacture of claim 9, wherein the probabilistic model comprises a Bayesian network.

12. A computer-based method for providing purchase recommendations on a web site to one or more customers, comprising:

collecting recommendation context information for a customer;

using the recommendation context to identify a set of candidate recommendations;

using a probabilistic model of a customer purchase decision to score the set of candidate recommendations by calculating a boost in a purchase-related criteria for each candidate recommendation using a plurality of gating factors in the customer purchase decision; and

selecting the candidate recommendation having the highest score for issuance to the customer.

13. The method of claim 12, where the plurality of gating factors may be selected from a group comprising relevance, exposure and clarity.

14. The method of claim 12, where the plurality of gating factors may be selected from a group comprising relevance, prior exposure, expected exposure, post exposure, prior clarity, expected clarity and post clarity.

15. The method of claim 12, where the purchase-related criteria comprises an expected margin, revenue or profit.

16. The method of claim 12, further comprising selecting a pitch for each candidate recommendation by using a probabilistic model of use cases to identify a use case that best matches the recommendation context, and then selecting a pitch associated with the identified use case.

17. The method of claim 12, further comprising selecting a first pitch for each candidate recommendation from a plurality of pitches based on a probabilistic model of use cases and an estimated value of each pitch for each use case.