Quantitative models to understand causality, levers, and influence in a complex world
Individuals and organizations interact through many channels, on multiple screens and devices, across a myriad of touchpoints and over time. Given the ubiquity of data, there is a new opportunity for firms to more fully understand the effect and value of their marketing actions. Big data, experimentation, and new models enable us to gain new insight into the causal levers and influences in this complex, extended world. Significant research is needed to develop better models that enable causal inference.
Topics include, but are not limited to, the following:
- Improving multi-touch attribution, marketing mix, and ROI models — across all media, digital and non-digital
- Understanding “omni-screen” and “omni-channel” drivers of customer decision making and behavior
- How can we efficiently and effectively detect signal versus noise in big data, and eliminate extraneous data?
- Identifying the critical paths to purchase in B2B environments using causal models
- Understanding and measuring the impact of creative — incorporating creative in causal models
- Given the enormous amount of data firms now have, are shorter time periods sufficient for causal inference?
- Identifying what we can do to drive behavioral change versus identifying a change wave that we are simply riding (i.e., not caused by our marketing actions)
“#1 priority is attribution!”
“We need improvements to ROI modelling that more accurately identify and quantify impact from various digital efforts.”
“We need to understand the effects of synergies across touchpoints — paid, owned, and earned. Current advanced analytics is providing learning and ROI on contribution to sales for each touchpoint, but is not taking synergies into consideration.”
“How can you measure creative quality in real time?”
“We need to develop ‘deep learning‘ (machine learning) methods; applications of deep learning methodology will be required to solve the most important marketing problems. When do traditional MMM or attribution models work better, and where does deep learning work best?”
“There is a lot of unrealized potential in the field experiment approach, where different levels of a marketing activity (or different activities) are employed (along with a control condition) and both immediate and long-term effects are tracked.”
“There are key tradeoffs between models that are focused on big important problems with many variables — sometimes hundreds — and the models we often see in the literature that are based on a handful of variables. These tradeoffs require a focus on important issues, an acceptance that we’ll never get pure causality, and a willingness to be approximately right on less essential issues.”