Want it Faster, Cheaper, Better? Then You’d Better Sign Up to Prescriptive Analytics

Data is getting bigger and it’s getting cheaper, problems are becoming more complicated, and the speed of decision-making is accelerating. The need for insights was never greater, in these uncertain and ambiguous times, managers are flying blind and need all the advice they can get. What is the slowest, most expensive, and error-prone part of this process? Yes, it’s people like you and me! We are the key reason that insights aren’t faster, cheaper, and better.

Are Humans the Whole Problem?

No! There are lots of things we humans can do that machines can’t. However, when we have a task that an algorithm can do, it does it more reliably, more accurately, faster and it’s less expensive than a human. We need to stop using people as substitutes for machines and use machines to free us, humans, to do things we do best (and which the machines can’t do).

So, we just need to automate?

No! Automation is certainly part of the answer, but on its own, it only makes things faster and cheaper; its impact on quality is typically negative or at best neutral. We need to leverage advances in technology, AI, and, in particular, Machine Learning, to change the way we automate, to ensure we are creating improvements.

And Why Prescriptive Analytics?

The journey for analytics has moved from Descriptive Analytics (showing what is happening), to Predictive Analytics (what happens to Y if I change X1, X2, and X3?), to Prescriptive Analytics (what is the optimal change I should make?) The key breakthrough that Prescriptive Analytics makes is that it can turn large, complex, and sometimes ambiguous data into recommendations. 

Descriptive Analytics was a way that humans could look at the data in better ways. Predictive Analytics answers the questions that humans ask. But Prescriptive Analytics makes recommendations to humans or, where appropriate, implements recommendations.

For example, some learning management systems (LMS) assess a student’s current attainment, map that against what they need to learn and select optimal modules for helping the student master the required end goal. If a human had that job and had access to all the past work by that student, and had access to all the modules available, and unlimited time, they would still be less likely to find the optimal combination. However, humans are unrivaled at finding out what a student’s goal is, finding out what challenges they are dealing with outside of learning, and of motivating and socializing the learner.

Prescriptive Guiding

Bain & Company describes three forms that Prescriptive Analytics frequently takes: Guided Marketing, Guided Selling and Guided Pricing. Guided Marketing lets the algorithms deal with the predictable, allowing customers to explore and travel through purchasing in non-linear ways. Guided Selling helps determine who to approach, what to offer, and when to do it. Guided Pricing comes into its own when deciding what sort of discounts to offer, for which products and services, and at what time.

Prescriptive Analytics in Marketing and Insights

Here are two examples of Prescriptive Analytics in the domain of marketing and insights:

Optimizing Marketing Activities

US-based marketing analytics company, Kvantum has utilized a prescriptive approach to evaluate multi-channel marketing campaigns quicker and with less data than traditional methods. Moving from 6+ months for traditional MMM approaches (and over $100K of spend for an initial model) to their approach that delivers models in 2-4 weeks, with a lower investment ($15K+ depending on the marketing spend). The prescriptive part is that it suggests the optimal spend mix, tracks the results and uses learnings from data to improve the model.

Predicting Customer Behavior

Sydney-based Houston We Have utilizes software that was originally developed for the Australian Defence Intelligence Organisation. Through joint ventures, the company embraces credit risk assessment, assessing investments in start-ups and, through its relationship with Australia-headquartered Potentiate, in predicting customer behavior. Houston We Have married Augmented Intelligence with Subjective Logic, which allows inputs with varying levels of certainty to be combined.

“Creating double the impact at half the cost in half the time”

This is the challenge that Unilever set to their head of insights, Stan Sthanunathan, and it is the challenge he set the research industry. Clearly, we won’t deliver twice as much, in half the time, at half the cost, simply by working harder. The answer is not more people, nor more creativity, nor more visualization. The answer is to use algorithms to make the best researchers better, to make the best researchers more productive, and allow the best researchers to focus on what only they can do. The key to that process is Prescriptive Analytics – analysis that doesn’t just describe, that doesn’t just answer questions, but actually contributes to the solution.

When?

If it had not been for the COVID-19 pandemic, I would have expected 2020 to be a key year for Prescriptive Analytics. Now, as we move on from the worst of the pandemic and into recessionary times, many organizations will need to aim for double the impact, at half the cost in half the time, just to stay afloat. Growth in the use of Prescriptive Analytics in the second half of 2020 will be hampered by scarcity and disrupted work patterns, so I think that we will see a boom in its adoption in 2021 and 2022 – but the smartest movers will be doing it now. It is certainly at the top of my ‘to do’ list right now.