Dashboard is accessible in the "Specialty Polymers" stream in Qliksense hub.


The aim of the QlikSense "Selective Pricing Optimization Tool"dashboard is to provide users with :

  • Information of the expected impact of the smart AI engine  price recommendations (also called pricing back-end or back-end).
  • Several dimensions of analysis like region, customer segments, market segments, products, sold-to etc.
  • Assistance in understanding some of the treatments done by the smart AI engine to recommend prices.


The primary data sources are :

  • Big Query : Contains all of the data coming from the AI engine hosted on DataIku.
  • SalesForce : Specific data not available in the back-end (End_Use__c object only).


Note : There is a more general “Pricing dashboard” developed by the transparency squad. This one is detached since it is almost exclusively based on data coming directly from the back-end. Also, the users are more restricted and the loading of the data should be done at a specific time. You can find the Pricing dashboard documentation in this Confluence also.


The loading will have to be done manually around once every three months when the DataIku project has been run and BigQuery data has been updated.

In the future, we can imagine an automated trigger when data is refreshed in BigQuery.


Basic explanation of the AI engine methodology and related data :

The objective of the back-end is to give price recommendations on a CPC (customer/product combination) level. 

Every product family of the GBU has its own independent machine learning model trained to do so.

To give recommendations, we are using three major steps (separately applicable to every product family) :

  1. Identify the importance of every price lever. This outputs coefficients also called weights based on the SHAP values of the ML model used.
  2. For every cpc in scope, find its more comparable neighbors using previously calculated weights.
  3. The recommended price of the targeted cpc is then based on the median price of the set of comparable cpc.


To learn more about the back-end methodology, please refer to the dedicated confluence documentation or the slides below :

https://docs.google.com/presentation/d/1FvFOgFyol6TdaU8JCm1IUHxVOhw3IzJKhuSdN5sNMoQ/edit#slide=id.g1161dbc3adf_2_214


Access the dashboard : 

=> Go to the QlikSense hub 

=> Click on the Pricing AI engine dashboard tab

Access is currently restricted to SpP value team.

Contacts : Christopher WILSON, Andrea PAPPADA


What can you find in the Pricing AI engine dashboard ?


UPDATE SS


  1. Recommendation analysis

    Synthesis of the results showing total impact of the recommendations.
     
  2. Impact by H4 and sold-to

    Analysis of the impact allowing to drill-down on product H4 and customers.

  3. Volume adjustment simulations

    Explanation on how we manage differences of volume between CPC when recommending prices.

  4. Volume price dispersions

    Scatterplots of prices against volumes to help justifying price increases and detect outliers.

  5. Details

    Exportable table containing most of the data used in the dashboard.




  • No labels