| Oracle® Retail Assortment Planning User Guide for the RPAS Fusion Client Release 14.1 E55312-01 |
|
 Previous |
 Next |
| Oracle® Retail Assortment Planning User Guide for the RPAS Fusion Client Release 14.1 E55312-01 |
|
Previous |
Next |
Clustering is the final step in the AP business process.
Table 4-1 Business Role in Cluster Maintenance
|
|
The Buyer Analyst builds and updates clusters in the Clustering task. |
Clustering is a business process where the merchant classifies the (store, site, application, and network) PoC base into multiple groups of PoCs that are similar in performance, space, or other attributes. Each cluster contains similar PoCs according to the criteria chosen by the merchant/trader. All PoCs within a cluster receive an identical assortment.
There are two types of clustering.
Basic Clustering
Basic Clustering enables the merchant/trader to create clusters within the solution without having to depend on external systems. With Basic Clustering, PoCs are clustered using either the Breakpoint or BaNG algorithm. For the Breakpoint algorithm, the merchant/trader uses sales and margin performance to group high-level clusters (cluster parent) and one or two PoC attributes to further define the lower level clusters. The merchant/trader reviews the cluster versions in the Look Maintenance task and assigns a cluster version to a look.
Batch Neural Gas (BaNG) Algorithm
The BaNG algorithm automatically generates optimal clusters based on user-specified number of clusters and clustering criteria. The algorithm provides a means for clustering data based on data distributions. For example, while clustering on weekly PoC sales data, the BaNG algorithm considers the Euclidean distance of the individual PoC/week level data points from a cluster center to determine the clusters. This is different from the Breakpoint method, where clustering is performed based on average sales.
The BaNG algorithm iteratively updates cluster centers while considering the distance of each data vector from the cluster centers and its contribution to each cluster center. For every data point, cluster centers are ranked based on their distance from the data point within each iteration.
Additionally, the cluster centers are guided, using a control parameter, to gradually spread from the center of the distribution to their optimal locations.
The BaNG algorithm is a non-trivial extension of the K-means clustering approach. It is usually faster than the K-means, and is guaranteed to converge.
BaNG vs Breakpoint
The BaNG algorithm generates statistically optimal clusters based on the number of clusters specified by the user. Breakpoint generates clusters based on user input breakpoints, and the number of clusters generated depends on the breakpoints.
In order to generate PoC clusters that vary by Dept, users need to specify a group by option of Dept. Breakpoint clusters PoCs based on average PoC sales within each Dept, while BaNG can consider an additional dimension for generating the clusters. For example, BaNG can cluster based on weekly sales of each PoCs within the Dept. Here, weekly sales are the coordinates over which the clustering is performed.
Advanced Clustering
If Advanced PoC Clustering is being used, the clusters are interfaced and imported into the Assortment Planning solution (potentially from the Oracle Retail Advanced Clustering solution). There is no need for the Basic Clustering workbook configured for this task. The merchant/trader is still able to review the cluster versions in the Look Maintenance task and assign a cluster version to a look.
Basic Clustering has the following steps:
To create the Clustering workbook:
Select the New Workbook icon in the Clustering task.
The workbook wizard opens.
In the Select Look Group page, select the Look Group for which the clusters are to be created or maintained. Click Next.
In the Select Source for Perf Clustering page, select a cluster source for assessment. Click Next.
In the Select Time Periods for Source page, select the weeks for which the clusters are to be created or maintained. Click Next.
In the Select PoC Channel page, select the channels in which the clusters are to be created or maintained. Click Finish.
The workbook is created.
There is one custom menu available for this step.
Seed WP with CP
In this custom menu, you can copy the Current Plan (CP) version of the clustering weights and attribute measure to the Working Plan (WP) version of the same measures. You would do this, when you have been testing some other options and decide to revert back to the last approved version of the measures.
To run the custom menu option, access the Custom Menu Option and select Seed with WP with CP.
| Menu Item | Seed WP with CP |
| Views |
|
| Trigger Measures | NA |
| Trigger Condition | NA |
| Additional Detail | NA |
The following table lists the updated measures:
| Updated Measure | Source | Notes |
|---|---|---|
| WP Weight Sales R | CP Weight Sales R | In Set Perf Clustering Criteria, PoC Analysis, Cluster Results |
| WP Weight Sales U | CP Weight Sales U | In Set Perf Clustering Criteria, PoC Analysis, Cluster Results |
| WP Weight Sales AUR | CP Weight Sales AUR | In Set Perf Clustering Criteria, PoC Analysis, Cluster Results |
| WP Weight Gross Margin R | CP Weight Gross Margin R | In Set Perf Clustering Criteria, PoC Analysis, Cluster Results |
| WP Weight Gross Margin % | CP Weight Gross Margin % | In Set Perf Clustering Criteria, PoC Analysis, Cluster Results |
This step has the following views:
This step has the following custom menu options:
Seed WP with CP
For information on this custom menu option, see Seed WP with CP for this task.
This custom menu option is available in the Review/Set Weights and Select Cluster Attributes views.
The Set and Review Weights view enables the Buyer Analyst to perform the following, as applicable to performance clustering (not space clustering):
View the strategy/intent results defined in the Setup task and the resultant weights for each of the performance measures.
Override the weights assigned in order to change the performance weights used to calculate the Combined Performance Index to Average.
The weight of metrics might vary for a given class for different buying periods to give the planner flexibility in determining the clusters for each buying period.
This view defaults the strategy/intent defined for the Brick and Mortar channel for the selected buying period and the related weights. The planner is able to edit the weights for any and all of the metrics provided; however, strategy/intent may not be edited on this view.
|
Note: The total sum of the weights assigned to all the metrics should be 100%; if the weights do not sum to be 100%, they are re-normalized upon the next commit and refresh. |
The planner can override the weights on this view in order to influence the PoC clustering for this assortment. Making overrides only applies to this assortment and does not impact the default weights assigned to this strategy or the class.
This view enables the Buyer Analyst to select additional PoC attributes that can be further broken down by performance group. A maximum of two attributes can be chosen in addition to performance. These attributes are selected in the WP PoC Attribute measures.
In this step, you create the clusters.
This step has the following views:
This step has the following custom menus:
Seed WP with CP
For information on this custom menu option, see Seed WP with CP for this task.
This custom menu option is available in the Define Perf Clusters, Define Space Clusters, and PoC Perf Analysis views.
In this view, the Buyer Analyst can choose:
The algorithm used for performance grading. The default algorithm is Breakpoint, but you can select a check box to switch to using the BaNG algorithm.
The number of performance-based cluster parents (A, B, C, D, and E) to be created. The maximum is five.
The label for each Cluster Parent (A through Z).
The uppermost performance breakpoint for Cluster Parents A through E.
In this view, if using space rather than performance-based clusters, the Buyer Analyst can choose:
The algorithm used for space grading. The default algorithm is Breakpoint, but you can select a check box to switch to using the BaNG algorithm.
The number of space-based cluster parents (A, B, C, D, and E) to be created. The maximum is five.
The label for each Cluster Parent (A through Z).
The uppermost performance breakpoint for Cluster Parents A through E.
Note: When clustering for Direct Channel, the cluster parents' name will be labeled as AE, BE, CE, DE, and EE. The images for Direct cluster parents will still show as A, B, C, D, and E.
This view enables the Buyer Analyst to perform performance analysis and cluster assignment.
Performance Analysis
In this view, you can see the following:
The actual performance of each PoC in the source data period as a list of metrics.
The Index to Average for each performance metric, Combined Index-to-Average, and Space-Index-to-Average.
The performance group and space cluster in which each PoC would fall, based on the algorithm and additional attributes selected on previous views.
Cluster Assignment
This view enables the Buyer Analyst to view performance classification for each of the algorithms per PoC.
The Buyer Analyst may manually assign a specific performance cluster or space cluster for any individual PoC. For example, a PoC that fell in the B-Grade may be assigned to A-Grade if the Buyer Analyst desires.
This step enables the Buyer Analyst to analyze the clusters created in the previous steps.
This step has one view.
This view enables the Buyer Analyst to analyze the clusters created in previous views. It displays the WP performance measures based on the source used for clustering: actual history, forecast, or location plan. This view also serves to indicate how the clustering then impacts the Assortment Plan.
The Cluster Results view is read-only, although metrics may be rolled up at the cluster level to ensure that the clustering is aligned with the higher level plans. Note that dimension splitting can be used on this view to analyze the clusters based on any of the PoC attributes.
This step enables the Buyer Analyst to approve the clusters created in the previous steps.
This step has one view.
This step has one menu option.
After the Planner generates the cluster parents and clusters, the cluster version can be approved to the CP version by executing the custom menu option:
This marks all the PoC clusters as approved by populating the CP versions of measures with WP values.
The CP version is exported to external systems.
To run the custom menu option, access the Custom Menu Option and select Seed with CP.
| Menu Item | Approve Clusters |
| View | Approve Clusters |
| Trigger Measures | NA |
| Trigger Condition | NA |
| Additional Detail | When the Approve PoC Clusters custom menu option is selected, the workbook is committed at that time. |
The following table lists the updated measures:
| Updated Measure | Source | Notes |
|---|---|---|
| CP Weight Sales R | WP Weight Sales R | In Approve Clusters |
| CP Weight Sales U | WP Weight Sales U | In Approve Clusters |
| CP Weight Sales AUR | WP Weight Sales AUR | In Approve Clusters |
| CP Weight Gross Margin R | WP Weight Gross Margin R | In Approve Clusters |
| CP Weight Gross Margin % | WP Weight Gross Margin % | In Approve Clusters |
| CP PoC Attribute 1 | WP PoC Attribute 1 | In Approve Clusters |
| CP PoC Attribute 2 | WP PoC Attribute 2 | In Approve Clusters |
| CP Use Bang! Algorithm | WP Use Bang! Algorithm | In Approve Clusters |
| CP Nbr of Perf Groups | WP Nbr of Perf Groups | In Approve Clusters |
| CP Perf Group Label | WP Perf Group Label | In Approve Clusters |
| CP Perf Breakpoint Upper Bndry | WP Perf Breakpoint Upper Bndry | In Approve Clusters |
| CP Perf Breakpoint Lower Bndry | WP Perf Breakpoint Lower Bndry | In Approve Clusters |
| CP PoC Count (Perf) | WP PoC Count (Perf) | In Approve Clusters |
| CP Nbr of Space Groups | WP Nbr of Space Groups | In Approve Clusters |
| CP Space Group Label | WP Space Group Label | In Approve Clusters |
| CP Space Breakpoint Upper Bndry | WP Space Breakpoint Upper Bndry | In Approve Clusters |
| CP Space Breakpoint Lower Bndry | WP Space Breakpoint Lower Bndry | In Approve Clusters |
| CP PoC Count (Space) | WP PoC Count (Space) | In Approve Clusters |
| CP Sales R | WP Sales R | In Approve Clusters |
| CP Sales U | WP Sales U | In Approve Clusters |
| CP Sales AUR | WP Sales AUR | In Approve Clusters |
| CP Gross Margin R | WP Gross Margin R | In Approve Clusters |
| CP Gross Margin % | WP Gross Margin % | In Approve Clusters |
| CP Combined Index to Avg | WP Combined Index to Avg | In Approve Clusters |
| CP Perf Group (Breakpoint) | WP Perf Group (Breakpoint) | In Approve Clusters |
| CP Perf Group (Bang!) | WP Perf Group (Bang!) | In Approve Clusters |
| CP Adj Perf Group | WP Adj Perf Group | In Approve Clusters |
| CP PoC Space U | WP PoC Space U | In Approve Clusters |
| CP Space Index to Avg | WP Space Index to Avg | In Approve Clusters |
| CP Space Group (Breakpoint) | WP Space Group (Breakpoint) | In Approve Clusters |
| CP Space Group (Bang!) | WP Space Group (Bang!) | In Approve Clusters |
| CP Adj Space Group | WP Adj Space Group | In Approve Clusters |
| CP Cluster Label | WP Cluster Label | In Approve Clusters |
| CP Gross Margin Index to Avg % | WP Gross Margin Index to Avg % | In Approve Clusters |
| CP Gross Margin Index to Avg R | WP Gross Margin Index to Avg R | In Approve Clusters |
| CP Sales Index to Avg AUR | WP Sales Index to Avg AUR | In Approve Clusters |
| CP Sales Index to Avg R | WP Sales Index to Avg R | In Approve Clusters |
| CP Sales Index to Avg U | WP Sales Index to Avg U | In Approve Clusters |
| CP Seed from Sister PoC(s) | WP Seed from Sister PoC(s) | In Approve Clusters |
| CP Sister PoC 1 | WP Sister PoC 1 | In Approve Clusters |
| CP Sister PoC 1 Weight % | WP Sister PoC 1 Weight % | In Approve Clusters |
| CP Sister PoC 2 | WP Sister PoC 2 | In Approve Clusters |
| CP Sister PoC 2 Weight % | WP Sister PoC 2 Weight % | In Approve Clusters |
| CP Sister PoC 3 | WP Sister PoC 3 | In Approve Clusters |
| CP Sister PoC 3 Weight % | WP Sister PoC 3 Weight % | In Approve Clusters |
| CP Volume Adjustment % | WP Volume Adjustment % | In Approve Clusters |
| CP PoC Count | WP PoC Count | In Approve Clusters |
| CP Sales R | WP Sales R | In Approve Clusters |
| CP Sales U | WP Sales U | In Approve Clusters |
| CP Sales AUR | WP Sales AUR | In Approve Clusters |
| CP Gross Margin R | WP Gross Margin R | In Approve Clusters |
| CP Gross Margin % | WP Gross Margin % | In Approve Clusters |
| CP APS Sales R | WP APS Sales R | In Approve Clusters |
| CP APS Sales U | WP APS Sales U | In Approve Clusters |
| CP APS Gross Margin R | WP APS Gross Margin R | In Approve Clusters |
| CP APS Gross Margin % | WP APS Gross Margin % | In Approve Clusters |
