Retail category management solution

RETAIL CATEGORY MANAGEMENT SOLUTION

Proper management of PLU turnover, placement and promotion

Situation

According to latest researches, lack of a stock organization and management causes around 8% of revenue loss, so weeding out that problem impacts on money flow in egregiously positive way. The largest Central & Eastern Europe retailer asked Unicsoft faced a need of a tool that would be a capstone of category management: having over 50000+ of PLUs and getting a dozens of new names every couple of days, Client had only a manual tool for category definition. So far, the basic intention of a tool was a creation of “tag clouds” supposed to tell one sub-categories from the other ones based primarily on a descriptive information, and, consequently, for proper management of PLU turnover, placement and promotion.

 

Solution

First step was defined as data preparation and enrichment : given more than 50k PLU names alongside with variety of features (main category name, manual tags, supplier notes, etc.), a main thing was to set up an proper DWH solution for storage and handling with further data pre-processing activities. Starting with SQL-type DB combined with Python libraries for pre-processing, a huge bunches of data was cleaned up and enriched. Next step was a tokenization of data: main goal was to create a tags cloud  for working out subcategories : I.e, given the “sweets” category through applying these methods based on keywords, a list of sub-categories came up (diabetic sweets, soya sweets, etc.). Subsequently, that approach was applied to entire portfolio of PLUs providing opportunity not only to categorize existing goods but to process incoming ones through fetching their description and its analysis. For the scalability purposes the second generation of solution was moved to Apache Solr keeping previous successful NLTK-based NLP approach for text analytics.

 

Result

Unicsoft involved highly-skilled Data Science and DWH experts for development and further tool support; break-even point for this particular solution was reached after 9 months of use through saving a significant amount of money by the cost-cutting on the staff that previously did such activities manually. The Client also was provided the support team for this particular solution for keeping all of analytical models up-to-date alongside with the highest level of accuracy.

The Client noticed high quality of Unicsoft delivery process as well as overall solution quality and expressed a willingness for further collaboration.

TECHNOLOGY & TOOLS: Apache SolrNLPNLTKPythonArtificial intelligenceMachine Learning

Platform: Web