Course5 Rekoncile

Today’s retail business thrives on a brand's multiple channels to market and sells its products. Among the various channels, the Marketplaces command the lion's share of the sales for the brand, along with building their brand equity. Almost all large retail brands sell a significant merchandise volume through platforms like Amazon, Lazada, Taobao, JD, Rakuten, eBay, and many more. As part of the process, these marketplaces share the sales data dump along with product attributes to the brand on a daily/weekly basis.

On the Brand side, a team of data analysts manually compiles and reconciles this data with the Brand product catalog to identify the actual Product/SKU sold on the Marketplace websites. This is a critical exercise to generate a unified sales report for the executives hungry to look at the regional, national, and global sales by Department, Category, Product, and SKU. If the data is reconciled correctly, it becomes possible to present clear and actionable facts to the executive team.

To solve this problem, at Course5, we have built an accelerator called REKONCILE, which automatically matches the product attributes between the Brand and Marketplace product catalog. This accelerator has been built on Azure Stack and leverages multiple out-of-the-box services from Azure, i.e., COSMOS, Azure Cognitive, Azure Functions, Azure Synapse, and Azure DevOps. Behind the scenes, we have built a massive data dictionary in partnership with the brand and also made use of the OpenAI GPT service to enable the category/merchandising managers to interact with the reconciled data conversationally and request deep-dive insights.

Here are some of the critical aspects of this accelerator:

• The Category/Merchandising managers can upload the downstream marketplace catalog (Amazon, Lazada, Jet, eBay, Taobao, JD, etc.) and get that reconciled based on the standard product metadata of the Brand selling its products through these alternative e-commerce channels.
• The Category/Merchandising managers can bulk upload transactional data from multiple marketplaces at once or do a single file upload via UI.
• Every uploaded file goes through multiple checks before we use it for reconciliation:
• Permissions and Privacy Compliance
• Security and vulnerability checks
• Format and Schema Understanding
• Data Quality Assessment
• Checksum/Digital Signature Verification
• File Type and Format Validation
• The solution can also match up the uploaded product category data with the US Harmonized Codes, the lifeline for all trading activities between nations worldwide. It leverages GPT to parse through various HS Codes, Chapters, and Sub-Sections.
• The solution allows the Category/Merchandising managers to share their feedback through a UI to fine-tune the model and improve its accuracy.
• There are multiple filter options in the interface for the Category/Merchandising managers to slice and dice the reconciled data.
• The solution also enables the Category/Merchandise Managers to present a unified transactional story to the regional, national, and international executives.
• The final reconciled report can be downloaded in a CSV/XLS format.
• The solution can generate a PPT response and curate a summary with quantitative insights for the executives by leveraging the GPT behind the scenes.
• While generating the summaries and content, the solution always considers relevancy, coherence, neutrality, and accuracy.
• The solution can generate the transactional report in multiple currencies by respecting the current exchange rate.
• Category/Merchandising managers can audit what % of their marketplace catalog got reconciled with an exact match, broad match, and not match. The broad-matched ones usually have one-to-many relationships, and human intervention is required to resolve them. Those who don’t get matched go through an orphan bucket and demand human intervention.
• Behind the scenes, we maintain a knowledge graph that beautifully encapsulates the relationship among different products, their attributes, parent categories, and brands.
• The Audit log is maintained for every upload and reconciled activity for the administrators. In the audit log, we capture both failures and successes.

To learn more about Rekoncile, we propose a 2-week consulting gig to discuss your problem statement, assess the fitment of the Rekoncile accelerator for your business problem, and the way forward. During this stage, we will also apprise you on leveraging Azure Data, Synapse, and Cognitive AI stack, to bring value for your business. After the culmination of this two-week assessment, we will explore the possibilities of doing a Proof-of-Concept for six weeks, leveraging your enterprise data within your secured and scalable Azure data infrastructure.