Oracle® Fusion
Applications Sales Implementation Guide 11g Release 6 (11.1.6) Part Number E20373-06 |
![]() Home |
![]() Contents |
![]() Book List |
![]() Contact Us |
![]() Previous |
![]() Next |
This chapter contains the following:
Define Sales Prediction Configuration
Manage Recommendations Configuration Parameters: Explained
Select Model Entities and Attributes
Predictive models analyze sales data to evaluate buying patterns and selling win or loss rates. The model can be used to identify customer profiles with a greater likelihood to buy certain target products. After evaluation of model results, lead generation can be scheduled in order to disseminate lead recommendations to users who can benefit from the insight gathered. Each lead recommendation includes win likelihood in addition to the average expected revenue and sales cycle duration from past similar deals for the same product to similar customers.
The Oracle Fusion Sales Predictor Engine (SPE) provides configuration parameters that you can edit to influence the behavior of the application. These configuration parameters control how recommendations are generated and displayed in other applications.
The configuration parameters are listed in the following table.
Configuration Parameter |
Default Value |
Description |
---|---|---|
displayModelBasedRecommendations |
True |
When this is set to True, model-based recommendations are displayed in consuming applications. |
eligibilityMinEligibleNum |
0 |
Sets the minimum threshold for number of recommendations eligible for each customer. If eligible recommendations are lesser than the threshold defined here, SPE requests more recommendations from InLine Service (ILS). This parameter applies only when leads are not eligible. |
eligibilityNumRecommendationsILS |
50 |
Indicates the number of recommendations to request from ILS when the number of eligible recommendations is lesser than eligibilityMinEligibleNum. This parameter applies only to the Order Capture eligibility rules. |
enableAvgWinRateRecommendations |
True |
When set to True, recommendations based on average win rates are displayed in consuming applications. This parameter applies only to the Order Capture eligibility rules. |
enableEligibilityRules |
True |
When set to True, marks leads in the staging table. Indicates whether to apply Order Capture eligibility rules. When set to False, it disables the parameters eligibilityMinEligibleNum and eligibilityNumRecommendationsILS. |
leadSalesChannel |
Null |
The selected value is updated in the SALES_CHANNEL in the lead staging table for leads customization. During implementation, you can customize the ILS with new values. |
leadSourceCode |
Null |
The selected value is updated in the SOURCE_CODE in the lead staging table for leads customization. During implementation, you can customize the ILS with new values. |
numberOfTopDrivers |
3 |
Defines the number of top drivers to return for each recommended product. The values can range from 0-5. |
Oracle Fusion Sales Prediction Engine leverages the power of predictive analytical models to identify patterns and correlation of data for the purpose of identifying what products to consider positioning next to your customers. The application leverages multiple mathematical models in order to formulate the likelihood a given customer will purchase a specific product, the estimated revenue which can be expected, and the duration of the estimated sales effort.
After the statistical model generates against the historical sales data based on the selected entities and attributes, summary and detail reports show critical insights as to what customers buy and can be used to predict the right products for the right customers. You can then use these insights to refine the process of generating leads based on what the customers are more likely to buy. Moreover, the same model can be used to predict the win likelihood of current opportunity revenue based on analysis of similar opportunities in the past. Also, product domain or market experts can write prediction rules to recommend products based on a set of rules conditions, utilizing all available customer profile attributes as well as other metrics.
The statistical analysis will identify which data has an influence in determining a likelihood to buy. After the statistical model generates against the historical sales data based on the selected entities and attributes, summary and detail reports show critical insights as to what customers buy and can be used to predict the right products for the right customers. As such, the decision regarding the selection of entities and attributes is critical. While the selection of certain entities and attributes may seem logical based on the awareness of sales behavior (customers in certain industries have a stronger affinity for certain products, for example), where there is uncertainty, the statistical model analysis will provide the necessary insight into whether patterns and correlation emerge. As such, as much available data as possible should be leveraged for the purpose of evaluation. Some factors which weight into the decision include the availability and accuracy of the data.
While selecting attributes, you must not select similar attributes for model training. For example, Annual Revenue and Annual Revenue Category. Your customer's annual revenue might range from 250,000 dollars to 50 million dollars. However, for efficient management, you decide to target only five customer types based on the Annual Revenue Category such as, Nano (250,000 dollars to 1 million dollar), Small (1-5 million dollars), and so on. The Annual Revenue Category uses Annual Revenue for this classification. Therefore, you must use either Annual Revenue or Annual Revenue Category but not both as they are redundant. Similar duplicate attributes could manifest in multiple areas such as Number of Employees and Company Size, Location and Zip Code, and so on.
Additionally, the application allows for the inclusion of expert insight from product management and sales and marketing operations. These expert insights can be captured via prediction rules. The same data made available to and leveraged by the predictive model, is also available for rule authoring.
Your company sells a service that appeals mostly to larger companies, and another service that targets smaller customers. If a customer purchased one of your product packages, then the customer already has all service needs covered by the package. You want to know, given a product recommendation, if credit score, asset, and customer size are important predictors when it comes to recommending this particular product.
You select the following entities and attributes:
Customer Profile
Annual Revenue
Credit Score
Customer Size Code
Past Purchased Products or Services
Assets and Service Contracts
These entities and associated attributes entail the data available to the enterprise which the predictive models can evaluate for identifying correlation, and which you can use to create prediction rules. Over time, you can further refine the selections based on availability of data and the cost to integrate that data for evaluation.