Clarifying the distinction between Data Science and Decision Sciences can be tricky because of the overlap between the technical skills required for each role. There is not a ton of writing on this topic, yet, but what ranks high in web searches can sometimes seem like the splitting of hairs. This is because the technical foundation of both roles is very similar. After all, Decision Scientists are doing data science at the core!  What is different, however, is how these skillsets are applied and, moreover, the end goal to which they are applied.

Below I begin to propose a new working definition of the Decision Science function within Marketing, specifically, that can be the starting point of a wholly revised framework for Decision Sciences in our field. One where the driving force behind all efforts is a relentless pursuit of the understanding and optimization of incrementality and lift:

Incrementality: measures the actual impact that a marketing activity has on a result or key performance indicator (KPI). It refers to the difference in any given metric between two or more groups, treatments, or segments, such as incremental orders, revenue, customer acquisitions, satisfaction scores, ad placements, etc. 

Lift: refers to the additional likelihood that customers will take a desired action after receiving a campaign or promotion compared to those that didn’t receive it. It is typically expressed as a percentage, such as “8% Lift.”

After all, improving return on marketing spend (ROMS) is a direct result of understanding the (estimated) impact of two or more choices based on the value each should generate (or not) over some baseline or expectation (Note: since first writing this post, it has been brought to my attention that author Daniel Vaughan refers to incrementality as “the holy grail of data science.”)¹

Moreover, measuring incrementality and lift provides marketers the information they need to demonstrate the real impact of their campaigns and programs. These companion measures quantify the improvement or benefit that marketing has generated above baseline demand, such as the increase in sales, website conversions, leads, or other key performance indicators (KPIs) that would not have occurred without marketing efforts.  Baseline demand — also sometimes called native demand — refers to sales that would have happed anyways, without marketing influence.

Why are these measures, in particular, so important to Decision Sciences in Marketing?  Because, at the end of the day, the ultimate inflection point for nearly every major Marketing decision should be centred on discussions about value creation.  And value creation is best measured through the incremental impact — and subsequent lift — that occurs as a direct result of marketing actions and spend decisions. By anchoring the outputs of Decision Science programs in these core measurements, we provide a concrete basis for enhancing the effectiveness and efficiency of marketing decision-making, ensuring that every decision is backed by robust, data-driven insights that drive the continuous improvement of value creation, de-coupled from baseline demand (as best as possible).

Failing to differentiate incremental value creation from baseline demand can overstate Marketing value.  It can also divert attention away for the real opportunities for optimization and growth.  For example, often Marketers will use predictive models to target customers with the highest probability of converting. This is very tempting because we want our campaign and program to be successful (i.e. bigger numbers are always better).  However, if many of these customers will convert anyways (“Sure Things”) without the additional influence of Marketing, then are resources not better allocated to the lower probability customers where the additional influence of a well timed and targeted campaign could increase their likelihood to convert (“Persuadables”)?²  Indeed, there is a predictive method called Uplift Modelling that is designed for this purpose and, in many cases, may be more appropriate than a straight-up propensity model. All things being equal, a Decision Scientist will embrace the methods that gets us closer to the truth about incremental performance so that decision-makers can make investment decisions with the best risk/reward trade-off.

Almost all data science methods can be applied to amplify focus on incrementality and lift. Within Marketing, this includes Marketing Mix Modelling (MMM), Multi-Touch Attribution, Predictive Modelling, Probabilistic Programming, Causal Analysis, Linear/Non-Linear Optimization, Simulations and even Cohort Analysis, to name a few. These are all important methods for understanding incrementality and, moreover, quantifying the risk associated with different decisions.  It may not always be possible to perfectly decouple baseline demand from true incremental gains. However, it’s vital for Decision Scientists to leverage these methods accordingly.

1. Vaughan, D. (2023). Data science : the hard parts : techniques for excelling at data science (First edition.). O’Reilly Media. ↩︎
2. Uplift Modelling: Making Predictive Models Actionable ↩︎