Predictive Modelling for Marketers: Analyzing Long-run Trends in Timeseries Data

This is how to use Holistic Marketing Modelling to incorporate a trend into marketing attribution modelling.

The Challenge: Integrating a trend into marketing attribution modelling

Marketing KPIs such as sales or revenue shouldn’t be stationary but instead, evolve and contain long-run positive/negative trends. Even more, these trends are not constant. They have their own development over time, changing and shifting at crucial points in the historical data. Our goal is to embed this in our core marketing attribution modelling.

Specifically, to isolate the long-run trend, find its changepoints (trend shifts), and forecast this into the future. It is important to distinguish the long-run trend from seasonal effects, short-run marketing effects, and other factors. Such factors of influence may include the competitor spend or even the weather. Isolating the trend is a complicated process.

Furthermore, the trend might also affect the factors of influence in return.  For example, the media effect might fluctuate over time in a positively or negatively trending market. The challenge is to perform our trend analysis while considering all the actual factors of influence. And, then, predict it all into the future.

The Solution: Our Holistic Marketing Model with trend-shift estimation

This case study demonstrates our approach to simulated data that contains a long-run trend and trend shifts.

Objective Platform developed an attribution model that blends methodologies like Multi-Touch Attribution (MTA) and Marketing Mix Modelling (MMM). Our Holistic Market ng Model includes experiments, past business knowledge and a trend-shift forecast algorithm. Data-driven marketing becomes even more sophisticated.

Read the full story?

Download case study