The client is a leading data storage solutions provider with products falling in both verticals: hardware and software. They deliver modern data storage experience by reducing the complexity and expenses behind it. Having a customer centric approach and ability to execute and our completeness of vision made them Magic Quadrant Leader.
Our expertise and solutions aligned clearly with our client’s vision and goals. Being a leading data storage solution provider, the client knew how fragile data handling could be. It is undeniable that Google is a global leader when it comes to handling and processing data, Pluto7 being a premier Google Cloud partner and being decorated with Partner of The Year for Data & Analytics award build the unwavering trust for us. The client wanted an implementation expert who can accurately predict the demand of products over a period of six months, which is an area of expertise for Pluto7.
The real journey of the customer started with our Demand ML solution on GCP Marketplace . After initial exploration and trial of the solution, it was clear to the client that Demand ML built by Pluto7 is the key to their business success.
Being in the data storage solutions business the client had a huge spectrum of customers differentiated on multiple factors like data size, team size etc. This diversity within their customer segments made them build a wide range of solutions to fulfill the market demands. While this was a paradise for their customers, the client’s biggest challenge emerged as accurately predicting the demand patterns for all the products.
The client had need to forecast demand patterns for next 15 months, but their data aggregation cycle was monthly and only 17 months of historic data was available to train the machine learning model. The implementation team at Pluto7 took their evergreen customer centric approach and consulted the client to generate forecasts for no more than 6 months, since with available data volume a forecast for a longer horizon would introduce low accuracy and multiple errors.
Another challenge came up with inconsistencies across available data for all the products. Within a given historic range there were multiple missing points, which were filled by extracting specific data from their ERP system, which would be stored into the GCS bucket. Since the volume of data was low, the team moved forward with statistical models rather than deep neural network models.
In the first few weeks a minimum viable model (MVM) was built. Once the client got familiar with the model and how the demand forecasting works, we automated the entire process by building a pipeline that would deliver end to end solutions from data processing to training the ML model and generating reports on dashboards.
The tailored demand forecasting model, Demand ML was successfully deployed to generate accurate demand predictions for a shorter horizon period of 6 months. Pluto7 team built an automated pipeline to take care of data inconsistencies across the range of products, improving team productivity by automating processes from data processing to training ML models and generating reports on the dashboard.