Thumb1

Analytics has a value problem.

This is not new. In fact, we have been talking about this for decades. That notorious failure rate to meet expectations from analytics has been stuck at 60-70% for as long as I can remember.

When we really think about it, this is abysmal. Imagine if you didn’t meet your job expectations 60-70% of the time—you wouldn’t have a job! The number of analytics produced but do not see the light of the day is astounding. Stories about models idly sitting on shelves have been around for decades. If they are not used, they do not produce value. Somehow, analytics has been able to survive such poor performance. Perhaps because analytics has been novel and maybe even sexy enough to overcome?

This is bad enough in normal times. I have already heard on multiple webinars that analytics would get cut in the current COVID-19 crisis if we do not start showing its value soon. Various sources have noted that analytics professionals have not been spared from furloughs, headcount reductions, hiring freezes, and rescinded offers. Possibly, the worst is yet to come.

The businesses will not, and should not, keep the analytics professionals employed if they do not see clear value from what they do, crisis or otherwise. However important, analytics becomes a pure expense if it does not produce value; at the end of the day, expenses get cut in tough times. We have a value problem already without the crisis, and the crisis only makes it worse. The time is ticking on the novelty and the sex.

While I don’t mean to be an alarmist, it is the reality we face today. So, how do we ensure we get value from analytics?

The “business value” of analytics

One question I hear too often: how do you show the value of the analytics to the business?

This is usually an indication that the analytic is built before the analyst(s) understand what is valuable to the business. Therefore, my question in response is “what is the value you are trying to bring to the business?”

Problems in Software Design

It may surprise you how often analytics professionals are unable to articulate a response to this question. As a result, they get caught up in things like model performance metrics, methods and techniques, and a host of other buzzy stuff they think the business wants to hear. None of these really address the value to the business. In a cost-cutting environment, one is not going to survive expense reduction by explaining a better r-square. Even in normal times, what does that really mean to the business, tangibly? This fundamental miss on the value creates an expectation in the analysts’ minds that their sole explanatory responsibility is technical. The r-square is not the end, not even a means to the end. It is a means to the means to the end. Analytics has a much bigger purpose, but we lose sight of it very quickly.

To be fair, this is not entirely the fault of the analytics professionals. Take the age-old conflict: the analysts want the business to involve them from the beginning, while the business says the analysts do not understand its needs. What analysts do not realize is that the beginning is too late. On the other hand, that the analysts do not clearly understand business values means that something about the organizational infrastructure—the people and the process but not the technology—is inadequate to support value generation from analytics. Consequently, the objective of analysis often becomes misaligned from the value that matters to the business.

“Begin with the end in mind”—Stephen R. Covey

If analytics is in fact important to the business, we really need to zero in on the value. As a first step, we need to change the way we view our respective place in the world. The business needs to own the articulation of the business value, and the analytics professionals need to own the support of those business values. Pointing fingers at each other accomplishes nothing.

We can learn a lot from process improvement approaches that focus on value. Personally, I am a fan of the Lean approach as a philosophy. We need to keep in mind that the value of a “product”—and an analytic is a product—is determined by the end consumer, the business, not by the producer, the analytics professional. The approach to value, therefore, must be end-focused.

In my previous professional life, I taught courses on music theory to music majors. The first few semesters covers a lot of basic classical harmony, with rules about which harmony can follow which and how the individual notes in the harmony can move from one chord to another. A common exercise is to provide a harmonic progression to a given melody.

Despite the word “progression,” a key trick in successfully completing these exercises is to start from the end and work backwards. This is particularly useful when one gets stuck in the middle, as it ensures the harmonies connect well from the beginning to the final cadence. In fact, going strictly forward from the beginning often results in a problem that is impossible to solve without starting all over. Or at least sounding funny.

Too many analytics efforts today are stuck in the middle at best. At worst they are impossible to remedy without starting all over. Not to overplay the oft-mentioned parallel between math and music, but this is exactly the same.

Abstract Futuristic infographic with Visual data complexity , represent Big
data concept, node base programming

If you can’t define it, you certainly can’t measure it, let alone manage it

Start from the end. And I mean, really the end. For each analytical opportunity, what is the business decision to be made? To inform the decision is consumption, so focus on the specific decision and not just on a business question. They are not always the same; in fact, they usually are not.

There are two common traps. The first is that the business question is usually too broad to connect us to the decision to be made. As a result, focusing on the business question can give everyone a false goal. This is fine in the exploratory stages but is detrimental in the finalized analytic.

The second trap: we all have heard some variants of “if you can’t measure it, you can’t manage it.” However, we are often so focused on measuring that we neglect clarifying what the “it” is. We rarely spend enough effort on it.

Analytical professionals are problem solvers by nature. This is both a blessing and a curse—too often, they jump right into solutioning. The business needs to help articulate the exact decision that it expects the analytic to support, with little room for interpretation. A good check: imagine you pull a stranger (with some relevant knowledge, to be fair) off the street. State the business decision you are trying to make. Are you confident that there will be no questions?

This is not about data or analytics

It is important to do this without any reference to data or analytics. Believe it or not, this is very hard. In the sessions I facilitate for my clients, I make words like “data,” “model,” “analytics,” “metric,” “score,” and so on, illegal. It is so easy to gravitate to “we can look at this data” or “we can use AI for that” before clearly articulating what we want them for.

Data, AI, machine learning, etc., are all sexier than “decision making”! References to data and analytics like this cheat us from understanding the real value; this is a quintessential case of solving the problem before we know what the problem is. Once we have the “it” clear, then we can address the question of how to measure or evidence it. If your “it” is sufficiently well defined, this is often trivial.

Start from the end to move forward

From this business decision, work backwards into how analytics can support it and then what analytic can produce that support. If you map forward from analytics, you only hope to connect to the decision. In contrast, by working backwards from the decision, you guarantee the lineage from the analytic. That is, you guarantee that the generation of the insights is concretely traceable to the consumption of the insights.

More often than we would like, the analytic originally intended is not the analytic the business needs. The originally intended analytic, had it gone forward (pun intended!), would have a value misalignment problem mentioned earlier.

If you cannot back into analytics, then either it is not a question for analytics to solve, or you do not know what you really need. Or worse, yet often the case, you have a much bigger issue than simply the capability to execute analytics. Analytics professionals cannot solve any of these.

Contrary to popular belief, the capability to execute analytics is relatively straightforward to solve for. Organizational success in analytics needs much more than that. However, the organizational stuff is beyond the expertise of the typical analytics professional. This is an entirely separate series of topics.

Does your analytics initiative have a value problem? There is about a 60-70% chance that it does.

Previous articleThe political debate that has not been
Next articleVirtual reality reaches out to save relationships
Michiko Wolcott
Michiko is currently the managing partner and principal consultant of Msight Analytics, a management consulting firm specializing in data and analytics. Backed by 20 years of experience in analytical project execution and delivery, she has helped organizations of all sizes in the development of enterprise capability and effectiveness in data and analytics. She has led multi-national analytical consulting practices, with clients and colleagues from across the globe in financial services, media and communications, retail, healthcare, life sciences, public sector, as well as humanitarian response and disaster management, and has spoken at many industry conferences and forums. Her prior responsibilities include serving as the Lead Data Scientist at North Highland and leading the international analytics practice at Equifax as the Vice President of International Analytics. Michiko holds a Master of Science degree in Statistics from Florida State University among other degrees from Florida State University and the Peabody Conservatory of the Johns Hopkins University.

LEAVE A REPLY

Please enter your comment!
Please enter your name here