Selling Price Prediction of Mobile/Lifestyle Business Using Machine Learning

Redington Gulf is amongst the leading broad line distributors across Middle East and Africa, distributing over 20 global IT brands. Founded in 1997 in UAE, the company today offers a wide array of IT products to resellers in the region. Product categories under the Redington Gulf include Computing, Lifestyle (IT accessories and IoT), Mobility Solutions, Printers and Peripherals and Software Products. Redington Gulf deals with reputed brands like Samsung, Apple, Huawei, etc. in the mobile phones industry and offers various end solutions to retailers like Promoter Management, VAN Sales, Indirect Sales, and Marketing. Redington provides end-to-end supply chain solutions for all categories of Information Technology, Telecom, Digital Lifestyle products.

Our team of Infrastructure specialists enabled Serverless transactions in the entire solution stack in order to minimize cost and enhance functionality.

Redington is distributing more than 500 SKU’s of mobile phones across multiple regions like UAE, Qatar, Saudi Arabia etc. Currently team is predicting the prices of each unit based on their experience and instinct, as there is no single source of truth to empower data driven decision making. It was difficult for them to predict the prices every day because of multiple factors needs to be considered while predicting the price, also the volume of SKU’s is enormous, which led to the pricing being an extremely tedious and arbitrary exercise, with no tracking, and or forecasting in place.

Redington needed a Forecasting solution which shall predict the prices of all the SKU’s based on historical data, market intelligence and other features across all the countries and customers. Redington also intended on a What-if model to predict the selling price of each SKU based on the quantity.

Citrus undertook multiple sessions with Redington domain experts to understand the use case and the availability of the data. After the thorough understanding of the use case and data, Citrus understood that Redington wants a single model which will give the price predictions for all their skus by providing the quantity they want to sell. Since the problem is related to multiple time series model and what-if scenario is included, Citrus proposed Amazon DeepAR plus algorithm to create the price prediction Model which will fulfill all the customer requirements.

In order to build the model, created a Data Pipeline to get the Sales and Inventory historical as well as live data from customer SAP Hana database to Amazon S3.

The team undertook exploratory data analysis and visualization of the dataset in order to understand its properties and prepared for further pre-processing. They selected specific products according to quantity and quality of historical data therein removed duplicates and did outlier analysis in order to identify anomalies and the business associations with regards to the same.

The team built the models based on different versions of the pre-processed data. Did experiments with different algorithms and services like AWS DeepAR plus, Aws Forecast etc. and respective hyper-parameter combinations. The best models were selected with their hyper-parameters and did further research in optimizing their performance. The same was therein deployed after testing and created an endpoint for the further usage.

Our team of Infrastructure specialists enabled Serverless transactions in the entire solution stack in order to minimize cost and enhance functionality. They created a Lambda function which will give predictions for any SKU. The Predictions include price upper bound, lower bound and actual predicted value. The Endpoint is exposed for input requests through Lambda function. A Rest API was created using API Gateway and attached the prediction lambda to API for the inference request.

For Implementing end to end Price Prediction solution for Redington it took 4 Months starting from March 2020. Divided the project into three phases and in each phase delivered the above requirements. By end of phase 1 we selected the SKU which has good quality and quantity of data and identified the key features that are responsible for building the model. In second phase prepared the data as required by the model and tried with different forecasting models finally selected DeepAR plus algorithm which worked best. By the end of phase 3 created the model and endpoint which provides real time predictions and automated end to end model training process.

• The solution improved the accurate prediction of the selling Price thereby reduce the loses.
• Forecasting time reduced because of less manual intervention
• Cost to business and manual intervention for prediction also reduced
• Identified the factors that influence the price prediction.
• Price fluctuation analysis.
• Model end to end automation without any manual intervention.