Blog

Unleash the power of simulation in marketing strategy development

  • Facebook
  • Twitter
  • Linkedin
  • Email
Explore

Jingfen Zhu

Chief Scientist, Analytics, Chief Science Officer

December 19, 2017 - Strategy development has evolved through several phases. What started as a reliance on expert judgement and primary research evolved to a data-assisted phase — and today confronts the challenge of having so much data. Excitement about the explosion of data is matched by confusion over its utilization. We often hear: How much data goes to drive strategy development? How can we integrate various forms of data? Are all of these equally useful? How can we uphold the quality of our evaluation when we bring larger quantities of data together?

In order to make data-driven decisions, people usually resort to tracking reports, or more complex methods such as modeling. The most popular approach for the latter is marketing mix modeling. This concept, developed in the 1950s, has been widely adopted for decades and remains the standard method in advanced analytics to inform strategy development. And yet, despite vast improvements in speed and flexibility in the past decade, this method still suffers from fundamental limitations:

  1. Data ingestion: We can only ingest structured information for modeling. Unstructured data — word of mouth, perception, and so forth — is difficult to quantify and integrate.
  2. Parameterization: Construction and selection of parameters have a big impact on the performance and robustness of a given model. Improper parameterization often leads to issues like multicollinearity and misrepresentation. For instance, when a campaign allows customers to respond online or by mail, how to parameterize the distinct contribution of online vs. mail?
  3. Distribution assumption: While some emerging algorithms are less strict with their assumptions around data distribution, many others have stringent requirements. The more dimensions our data represents, the more likely that it will violate the underlying assumptions. As a result, the analytical results become questionable.

Thanks to interdisciplinary approaches to data science, simulation is fast emerging as a new promising alternative. We have seen extensive application of it in operational fields such as network traffic and chemical processes, but rarely as much in sales and marketing. Some of its unique properties come as a solution to the problems inherent in the modeling approach:

  1. Using simulation, we can integrate disparate data sources, which are easily integrated regardless of their form. Structured data like sales and unstructured data like tweets and consumer preferences can all be integrated in simulation.
  2. Simulation aims to reconstruct the full dynamics of the market, and thus generates output with a higher degree of validity and reliability. Simulation can incorporate customer preference changes over time and model the influence of social media on purchase behaviors. However, these present huge challenges to a modeling approach.
  3. While a marketing mix model handles one target at a time, simulation estimates multiple targets simultaneously. Instead of building two models — one for revenue and one for customer growth — using simulation these KPIs can be synced in one system.
  4. Since simulation emulates interaction and interdependence of the true market, there is no “destiny" or outcome. Changes can flow in one or multiple directions. Effective digital campaigns lead to strong sales, which in turn reinforce the brand image and market perception.
  5. We can estimate channel synergy during scenario iterations. Simulation's ability to fully replicate market dynamics makes estimating channel synergy seamless, intuitive, and accurate. Moreover, simulation does away with the constraining assumptions that underlie most marketing mix models.

In addition to the above advantages, simulation is also a scalable option. Traditional marketing mix modeling focuses on marketing activities without considering other parts of the value chain such as supply chain management, partially due to the computational burden. But advances in technology and infrastructure have made enterprise-level touchpoint-attribution possible. We can optimize resource allocation across multiple product portfolios and multiple segments simultaneously.

Born outside of the marketing arena, simulation is an increasingly compelling approach to deriving insights and guiding strategy development. The beauty of simulation is that it reconstructs full market dynamics to generate ROI estimates and trade-off scenarios within a unified process. It encompasses all the mathematical rigor of modeling — and goes beyond it. Analytics armed with simulation should take the practice of strategy development to the next level.