Customer-driven campaign big data analysis of behalf of an international online retailer

Project executed on behalf of an international online retailer

Project Objective

Online retailers possess a lot of information about their clients’ preferences and habits that, if properly exploited, may be used to enhance their overall performance. However, finding the right takeaways involves the analysis of several millions of registers, rendering this task significantly complex. One of the leading international online retailers commissioned Axon with the objective of providing analytic results and conclusions on the behaviour of customers in relation to the various campaigns launched by our client and building a statistical model that predicted customer behaviour and recommended marketing actions based on a pre-defined set of objectives (e.g. maximise profitability, increase sales, attract new customers).

Project Description

Firstly, we performed an analysis of all the orders placed in our client’s in-store and online channels over a two years period. This analysis allowed us to provide valuable insights to our client in terms of:

  • Demand seasonality
  • Stores’ performance
  • Relevant customer segments according, for instance, to their time of purchase habits:
  • Impact of different marketing campaigns in attracting different customer groups.
  • Effect of external factors (e.g. sport events, weather, special events) on sales
  • Assessment of the performance of all marketing campaigns, including: Their ROI (Return On Investment), their impact on the ATP (Average Ticket Price), the efficiency of the different channels (e.g. leaflets, TV, radio, SMS, e-mail, social networks), their impact on sales, and the comparison between marketing spending and contribution to margin:
Figure 1: Marketing spend vs contribution margin of campaign groups

Following our initial backwards assessment, we built a predictive model, fit with machine-learning algorithms, that allowed our client to design its marketing campaigns according to the specific objectives pursued at any given time. The model was continuously fed with new and live data, thus improving the accuracy of its estimates (e.g. impact on sales, acquisition of new customers, ROI) over time.

Figure 2: Overall data management and analyses model implemented for our client

Key Takeaways

The predictive model we implemented for our client became the “right-hand” of its marketing department over the coming months. In the 12 months following its deployment, our customer’s sales increased by over 20% and the number of unsuccessful marketing campaigns was slashed by over 80%.