Retail, some basic lessons I learned

Based on a SAP platform, some basic lessons I learned about retail.

The stores:

  • The POS are terminals that retrieve data from a server, typically located on the same store.
  • The server of the store receives/sends data to a SAP PI.
  • The frequency of these communications depends on your capacity to manage big volumes of data.
  • The technology on the store can be independent to SAP. SAP PI will send the appropriate data format.


  • The Origin of all data comes from ERP, specially from SAP SD (sales & distribution).
  • SAP Retail will gather the detailed data that goes and come to the POS servers.
  • SAP Retail have all detailed data. The quality and the granularity of the data will determine the ability to posterior data mining actions.
  • Typical problems of SAP Retail are: communications, and reporting (have the ability to gather the required data in terms of volume or detail).
  • The management of promotions and affiliations is other major topic on the processes.
  • To simplify the data complexity, SAP Retail sends the data to SAP SD taking into account that each store is a single customer.
  • Affiliated people can be on a CRM or on SAP SD.
  • SAP PI sends/receives data and transforms it into the right formats. This enable the integration with different POS vendors.

Data mining:

  • For high levels of detailed data: retrieve it from SAP Retail. SAP HANA is not enough for handling big volumes of data. Hadoop is the preferable solution for these big volumes.
  • For detailed data: retrieve it from SAP SD.
  • For generic data: retrieve it from SAP FICO.
  • A serious BW retrieves data from SAP Retail.

Real time & transactions:

  • Frequency of transactions between POS and SAP can be complex when you have so many terminals.
  • The need to increase this frequency can stress the architecture and force changes at that level.
  • The demand of data miners and the need to act on prices at the end of the day depending on the consumption trends can be something complex if the architecture is not well implemented.
  • Majority of transactions are done during the night, but it could be different.

Leave a Comment