From data to insight: Monte Carlo simulation as a marketing intelligence tool

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
School of Business | Master's thesis
Degree programme
101 + 8
A data-driven marketing approach has gained tremendous popularity in business and implementing business intelligence solutions has become almost mandatory for companies to stay relevant in the competitive business environment. CRM systems are the flagships of business intelligence solutions in marketing. This versatile marketing intelligence solution offers a wide range of possibilities for deriving further marketing insight from customer data for marketers. CRM systems work both as storage for customer and product data and as an analysis tool for this information. One possible method for transforming raw data into actionable insight is to use simulations and especially Monte Carlo simulations to solve marketing-related issues. The applicability and usefulness of Monte Carlo simulations in marketing were researched in this study by building two separate models that dealt with common issues in marketing. Despite being a relatively foreign study method in marketing, it has a wide range of possible usages. This study approached simulation from an operational standpoint, which meant that simulations were modeled in a spreadsheet environment to demonstrate the usability of Monte Carlo simulation for marketing managers who do not possess extensive knowledge in programming or statistics. The first simulation model built was a customer base profitability simulation, which was derived from the extensively researched customer lifetime metric. The second simulation model was built to forecast future sales for one year. A dataset that is commonly used in CLV studies, ‘CDNOW’ was used in the NPV simulation. In the sales forecasting simulation, a unique dataset was obtained from the CRM system of a medium-sized company for the purpose of this study, containing sales data which was then used to build the simulation. Both simulation models were evaluated with commonly used forecast accuracy metrics. The customer base profitability simulation predictions were relatively accurate, indicating that the model in which customers had homogeneous purchases provided more accurate predictions. The second simulation model that dealt with sales forecasting, provided predictions from future sales with high accuracy. The NPV simulation advocated homogeneous modeling of customer purchase behavior. The results from the sales forecasting model advocated using a determined weight that infused learning in sales to the model.
Thesis advisor
Kajalo, Sami
Rosenberg, Laura
CRM, marketing intelligence, Monte Carlo, simulation, analytics
Other note