To attract footfall in-store, a leading retailer of electronics invests in
  • Newspaper, radio and online advertising
  • Social media management
  • Web site and mobile app development
  • Refurbishment of their stores

but it is having trouble tying these investments back to revenue which is limiting their ability to understand their true ROI. They now want to be able to measure the effect on footfall in their various stores.


The geography the client was operating in, online channels were more of research channels rather than sales channels. So, our approach was to analysis these channels to understand how promotion spends on different channels impacted the traffic on these channels and then correlate the increase in traffic on a product level to predict the footfall in the store. Historical sales and footfall data were used to develop base forecasting models. An ensemble of these base models was developed to arrive at the final prediction.


  • Facebook data scraped to gain a measure of the level of engagement on a daily basis
  • Web traffic, app traffic, and Facebook engagement was incorporated with a lag in predicting footfall
  • Media spend was translated to expected reach
  • Grid search used to optimize each of the forecasting models
  • An ensemble of all base forecast models was developed


ARIMA, Holt-Winters, Neural Networks, Linear Regression, and Ensemble Modeling


Models predicted the result of each investment on the footfall at the store and allowing the business to plan in-store promotions, staffing requirements, and project revenues.