Mirror of Oracle documentation
Converted for search and offline reading. Authoritative source: Oracle. Diagrams and some complex tables are simplified — check the PDF when in doubt.
3
Demand Transference
This chapter described the use of the Demand Transference Science Cloud Service module.
Introduction
Demand Transference (DT) helps you to compare products based on their similarities in order to determine what, if any, products customers might buy if the product they want to buy is for some reason unavailable. In this way, planning and ordering can be optimized. DT calculates similarities by comparing the attributes of the two products. If you are using CDT in conjunction with DT, you also have available the similarities calculated by CDT, which are based on customer-supplied transaction data.
The DT Cloud Service module consists of three tabs: Overview, Generate Models, and Manage Models. You use the Overview tab to keep track of the status of each stage during the main work you do with the application within the Generate Models tab. You use the Manage Models tab to evaluate the demand elasticity results and override the Maximum Substitutable Demand Percentage value, if that is needed.
Overview of DT Process
When you use the DT Cloud Service module, you follow this general iterative process to create and manage DT models:
-
Data Setup. You define the categories to be used in the DT calculation.
-
Data Filtering. You configure filters that remove input data that might cause errors in the calculation or that can lead to inaccurate or unreliable results.
-
Similarity Calculation. You calculate similarities and assess the results of the calculation.
-
Elasticity Calculation. You calculate the assortment elasticities and assess the results in terms of substitutable demand, which is the percentage of demand of a SKU that is retained when the SKU is deleted from the stores where it is selling.
-
Escalation. When you are satisfied with your results, you can set the escalation path you can set the escalation path to fill in the holes for partitions whose DT models were removed during pruning by setting up a search path through the segment hierarchy and the location hierarchy. Then you can set a version of the DT model as complete.
-
Manage Models. Use this tab to set time intervals for evaluating your results and to override the value for the maximum substitutable demand percentage.
Overview Tab
The Overview tab displays information that you can view and use to monitor the progress of the DT stages as well as to view some aggregate statistics and the DT results from the last successful run.
This tab contains the following sections:
- Generation Stage Status
eC‘
ee
| ~ Aggregate Statistics | |||||
|---|---|---|---|---|---|
| Views wh | Fg a | ||||
| Version | Createdfy | Distinct_= | ——stnet_— | Customer Segments | ——=—smodels generated |
| Coffee Attr | cdmUserd | 1 | 10 | 4 | 10 |
| Coffee Attr Chain | dtUser0 | 1 | 1 | 1 | 1 |
| Coffee Attr2 | cdmUser0 | 1 | 10 | 6 | 60 |
| Coffee Smoke Test | dtUser0 | 1 | 1 | 1 | 1 |
| CoffeeTxn | cdmUserd | 1 | 10 | 4 | 10 |
|
ee
TO
—S—CSSSCSCisa(
eC‘
eC‘
eC‘S
eeee ee ee
Table 3-11 Version Setup: Fields
| Field Name | Field Description |
|---|---|
| Version Name | Assign a name to each version of a DT model calculation. This allows you to create and save more than one version of a DT model. The version name you assign here is used in the Calculation Report, Aggregate Statistics table, and in the Manage DTs tab. Version names can be re-used; however, if the version name in question has active DT models, then you will see a warning that the active DT models will be removed from the version if you do re-use the version name. |
| Select the Source of Similarities | Use this option to define the type of data used in the calculation: transaction data or attribute data. Transaction-based data uses similarities calculated by Customer Decision Tree using transaction- based data. Attribute-based data calculates similarities within DT based on the attribute values associated with every SKU in the category. |
| Process Location Top Level Only | Check this option if you want DT models to be calculated for the Location Chain_only_. You can select this option in order to decrease the amount of time it takes the system to perform the calculation. |
| Process Customer Segment Top Level Only | Check this option if you want DT models to be calculated for the Customer Segment Chain_only_. You can select this option in order to decrease the amount of time it takes the system to perform the calculation. |
Once you select the source for the similarities, you will see either Category Attribute Setup, if you have selected to use attribute-based similarities, or Transaction-based Similarity Availability Per Category, if you have selected to use transaction-based similarities.
Category Attribute Setup
The application, using historical data, determines a specific weight for the category attributes. You can optionally change this weight and assign your own weight to the attributes for a category.
The weights indicate the importance of the attribute to the customers when they are making purchasing decisions. The attribute with the highest weight is the one the customer considers first when making a purchase. The system-generated weights are determined by the application from historical sales data. However, if a user disagrees with those weights, the user can override them. For example, in the case of coffee, the system may assign a weight of 0.7 to brand and 0.2 to size. This indicates that brand is historically more important to the customer than size when purchasing coffee. If the user disagrees with this analysis and thinks that brand and size are actually much closer together, the user can assign a weight of 0.5 to brand and 0.4 to size.
The Category Attribute Setup table displays the following:
Category Attribute Setup Attribute Setup Setup
eeee ee een erence eens
The Category Attributes Setup pop-up lists the categories you are calculating DT models for. The system-assigned weights are also displayed. You can adjust the weight for all the attributes or a subset of the attributes.
The Category Attributes Setup dialog box contains the following fields. For each attribute you want to assign a custom weight to, enter a number between 0.000 and 1.000. For attributes that have no substitutes (such as windshield wipers of a specific length), the Functional Fit check box is checked by the system, so that similarities are not calculated for these attributes. When you are finished configuring the category attributes, click OK .
Table 3-13 Category Attribute Setup Fields
| Field | Description |
|---|---|
| Attribute | The category attribute to assign a weight to. |
| User-Overridden Weight | The user-defined weight for the attribute. |
| System-Generated Weight | The system-generated weight for the attribute. |
| Functional Fit Attribute? | This is checked by the system if the attribute has no substitutes. |
After you have finished configuring the similarity parameters, click Run to calculate the similarities. You see the results via the Similarity display table.
Transaction-Based Similarity Availability Per Category
The Transaction-Based Similarity table displays the following:
Table 3-14 Transaction-Based Similarity Availability Per Category
| Field | Description |
|---|---|
| ID | An external code used to identify the category in other systems such as CMPO. |
| Name | The category name. |
| Description | A description that provides additional information about the category. |
| Available | A flag that indicates that a CDT version that contains data for this category has been made active. |
| Available As Of | Indicates the date that the CDT version was activated. This information can help you identify whether the CDT results are recent, or if they are potentially too old to use. For example, if the CDT data became available two years ago, you may consider that data to be out of date. |
Similarity Display
The Similarity Display table shows the list of SKUs for which similarities have been calculated so that you can sort and analyze the results. You can search through the list of results by Category Name, Location, or Customer Segment.
Click the Refresh icon to update the fields and see the latest information in this table.
eC‘S
ee
ieeeeSCSCSCisN
ee
SCC‘
~ Escalation Report
@
Total number of partitions 25 ’ NumberNumber ofof partitionspartitions sourcedsourced fromfrom escalationcalculation 124 Number of partitions with no source 0
| View~ | By | a | |||
|---|---|---|---|---|---|
| Customer Segment Level | - anti | crat | Percentages | ||
| SEGMENT | REGION | 24 | 96% | = | |
| CHAIN | REGION | : | 4% | ||
| SEGMENT | AREA | 0 | 0% | ||
| CHAIN | AREA | 0 | 0% | ||
| SEGMENT | CHAIN | 0 | 0% | ||
| CHAIN | CHAIN | 0 | 0% | ||
| SEGMENT | COMPANY | 0 | 0% |
OO—“C:SCSC‘CC‘tés
Table 3-21 (Cont.) Escalation Report
| Field Name | Description |
|---|---|
| Number of Partitions Sourced from Escalation | The number of partitions removed during escalation. |
| Number of Partitions Sourced from Calculation | The number of partitions removed during calculation. |
| Number of Partitions with No Source | A partition that does not have a model assigned to it because all models related to the partition have been pruned. |
| Customer Segment Level | Identifies the customer segment level. |
| Location Level | Identifies the location level. |
| Number of Models | The number of partitions that are trying to have a model assigned. This is generally the number of customer segments by the number of locations. |
| Percentages | The percentage of partitions that have been assigned a model from a given escalation level. |
Completion of Process
When a version is complete, the results for the version are activated so that other applications can use the information. The similarity data that has been calculated during the generation process is also activated for use.
After the completion of this step, the intermediate results from each stage are removed from the database and can no longer be used.
Be aware that once a version is completed, it cannot be completed again unless a different version is completed first. Changes made to the version’s data after completing it will not be copied to the relevant output tables.
Manage Models Tab
You can set up various time intervals to use in the evaluation of a version of a DT and configure the value for maximum substitutable demand and see how different maximum values affect the substitutable demand. This allows you to change the maximum value to one you find more suitable. You can see the percentage of demand of a SKU that is retained when the SKU is deleted from the stores where it is selling. This is the substitutable demand percentage. In this way you can evaluate the accuracy and usability of the elasticity calculation.
Substitutable demand is a measure of how much demand is retained by the rest of the assortment when an item is removed. When the item is removed, a portion of its demand is transferred to the remainder of items in the assortment. This portion is considered the retained demand. If the magnitude of the assortment elasticity is larger, then the amount retained will be higher. By examining the retained demand, you can evaluate the assortment elasticity value to see if its magnitude is too large. The key value to examine is the maximum substitutable demand percentage. For a given category, you may decide that this value is too large.
Process
Here is the high-level process for determining suitable substitutable demand values and thus suitable DT models.
- Set up the time intervals you are interested in.
ieeeeSCSCSCisN
ieee—SCSSCi‘(N
~ Escalation Report
@
Total number of partitions 25 Number of partitions sourced from escalation 25 Number of partitions sourced from calculation 0 Number of partitions with no source 0
View & Customer : Segment Per nil hee 1 Percentages jase eve models SEGMENT REGION 0 0% CHAIN REGION 0 0% SEGMENT AREA 0 0% CHAIN AREA 0 0% SEGMENT CHAIN 0 0% CHAIN CHAIN 0 0% SEGMENT COMPANY 0 0%
ieeeeSCSSCisN
Histogram
The Substitutable Demand Distribution histogram displays the distribution of the substitutable demand values for SKU/stores.
In this guide
- Guide: AI Foundation User Guide
- Previous: 2 Customer Decision Trees
- Next: 4 Advanced Clustering