Implementing more accurate sales forecasting to drive more efficient operations
On average, the new predictive model engineered by Data Minded is 33% more often within the ±10% error range compared to the existing legacy approach. In addition, the end-to-end data processing and analytics is now fully automated. The Data Minded specialists succeeded in completing the entire project within a period of two months.
The stream of consistently more accurate sales predictions enables Newpharma to streamline its operations accordingly. Considering the large scale of Newpharma's operations, this involves expenditure savings that are huge. The scalable and structured approach of Data Minded intrinsically leads to a more efficient and targeted use of human resources.
Why Data Minded?
Through referral, Newpharma heard about Data Minded as being the right partner for data science and engineering. Right from the start of the project, the data engineers of Data Minded felt a connection with Newpharma in terms of company vision and culture.
Newpharma is market leader in Belgium and France in online pharmacy and plays a prominent role on the European market. The company reached out to Data Minded to help kick off its analytics transformation. The project started off with the case of predicting future sales.
In the past, Newpharma steered the planning of its logistics operations workforce using basic sales forecasting models enriched with market and process experience. As the number of its customers and products increased dramatically, Newpharma decided to look for a more data-driven and scalable approach to forecast sales and organize its operations.
In a first two-week period, we have explored a wealth of data (historical sales, product catalogue, marketing campaigns, weather, etc.) as well as different predictive modelling techniques. Subsequently, we were able to outperform the legacy model using the acquired information and insights.
After this positive result, Data Minded has designed and implemented a data platform based on AWS Cloud, Airflow, Spark and Python to train and run an LSTM model on a daily basis and publish order predictions. Kubernetes is used as an orchestration solution for automating and scaling infrastructure and applications.