Retails: Weather and Event-aware Business Intelligence for the Optimization of Campaigns and Human Resources

Weather and Event-aware BI Sales Strategy Advisor is the solution developed by BIG BANG – in collaboration with EW-Shopp –  to enable maximization of marketing budgets and optimization of the workforce. Big Bang is the biggest Slovenian multi-channel retailer in the segment of Consumer Electronics & Home Appliances with a daily average of 13k visitors with 27% closed sales in 18 brick and mortar stores. 

The goal of this new solution is to optimize Salesforce and Marketing communication planning by predicting the daily number of visitors. A more clear understanding on visitation patterns gives valuable insights for scheduling staff on the floor to match customer demand and improve customer service or at least make it more consistent; improved work organization on the floor and coping with logistics and stock challenges more effectively; maximizing marketing budgets by adapting sales activities on the floor; making more accurate sales predictions.

KEY ASSETS

Weather and Event-aware BI Sales Strategy Advisor enables better, faster, and informed business decisions and clear responses to shifting foot traffic patterns. To achieve this, it uses robust foot traffic predictive analytics to deliver information to

  • the Sales Department – for (weekly) salesforce optimization;
  • the Marketing Department – for marketing activities optimization.

THE APPLICATION

Weather and Event-aware BI Sales Strategy Advisor is built by ingesting fresh and historic data about foot traffic realization and sales parameters from all the Big Bang stores, linking them with information about weather (from Open Weather Map) and enriching them with event data that include external events (e.g., shopping events and calendar events such as paydays and holidays) as well as internal events (e.g., marketing campaigns grouped by channel). The enriched data are used offline to build predictive models to estimate foot traffic and sales based on historic data, weather, and events. The trained models are used online to compute the estimation for the incoming week using fresh weather and event data and feed visualizations delivered to the users through a business intelligence (BI) analytics platform. Weather and event-based analytics are developed with the help of the JSI’s QMiner library. 

The Sales Strategy Advisor web application provides the user with the capability of exploring data and predictions with three main views. 

  • General Overview provides information about the status of all the stores by reporting foot traffic predictions, weather conditions, predictions on sales value and the estimated number of employees needed on the floor for current day and five days in advance.  

  • Store Overview with rolling weekly overview (current day -2, current day +5) that enables drill-down to store level on weather conditions, foot traffic predictions, predictions of all correlated sales parameters, estimated number of employees needed on the floor and calendar of events (together with notifications on store-specific events). It also includes confidence interval data to enable user to more clearly follow the accuracy of predictions and real data to evaluate the differences between predictions and realization. 

  • Detailed Overview that provides aggregated overview of all the parameters at the monthly level with ability of drill-down to micro-level (e.g., by store or date) and dynamic views based on filters (parameters and KPI value by filter settings). 

ACHIEVED IMPACT

The impact of developing Sales Strategy Advisor and introducing it into Big Bang has improved the decision process on many levels. We summarize the main advantages found as a result of an internal evaluation campaign carried out by a team of department managers (sales and marketing), regional and store managers, event planners and analysts.  

  • Moving from ad-hoc spreadsheet-based analytics or rule of thumb prediction and planning processes to end-to-end automated processes with sophisticated predictive analytics enriched with continuous monitoring. In this way we have also gained some competitive edge.
  • Achieving predictive power, with accuracy as high as 80% at the general level and 70% at store level to enable weekly planning. 
  • Adapting sales activities on the floor to make them more beneficial (when store managers receive information to anticipate higher foot traffic, salespeople should act for increasing volume – when foot traffic is expected to be lower salespeople should perform activities aimed at increasing closing rate and/or basket value).
  • Giving information to marketing department on ad-hoc marketing activities, i.e. when to enforce »must have« activities vs. »best performing« activities and when to perform activities with brands/distributors.
  • Adjust schedules with stronger sales force presence when higher traffic is expected (predictions for five days in advance also satisfy the legal obligation of changing schedules three days in advance), which also positively affects in better working climate.
  • Improved work organization on the floor and coping with logistics and stock challenges more effectively i.e. adapting delivery times when lower visitation is anticipated. Scheduling staff to match customer demand furthermore positively influences customer service or at least makes it more consistent.
  • More accurate sales realization forecasting, since based on historical data, a high correlation on the level of store traffic – sales performance is noted.
  • Further advancements include improvements on predictions on the number of employees by making suggestions on how many employees to put on the floor by shifts (morning and afternoon) and not per day only. The first step we are already working on is improving the storage of internal data to enable such predictions. 

INDUSTRY IMPACTED

  • Retailers in the segment of Consumer Electronics & Home Appliances.
  • Brand Manufacturers/Distributors.

RESEARCH & INNOVATION DOMAIN: Data integration, data aggregation, analytical and predictive modelling, user interface and visualization

TOOLS INVOLVED:  QMiner, SPSS, MariaDB, MS SQL, QlikSense

CONTACT PERSON: Patricija Filipič Orel patricija.orel@bigbang.si