I am often asked about the kinds of Analytics I perform as a consultant to address the questions my clients pose. The “real question” behind this is: What kind of Analytics do I get to engage in? The focus of this article is on what kinds of things I do as a Predictive Analytics Consultant. Other types of Analytics are not discussed here.
Typically a client will realize that there is more that they can do to engage their customers, keep their customers, or reach a broader share of the market. They often do not know what specific analytic questions to pose. As a consultant, I do several things, but one of the most important is to help the client pose their questions in a way that can be addressed by Analytics, if appropriate. After that, it is a matter of translating the client’s question into an Analytics question that can be addressed with predictive models, for example.
Asking the Right Question
In “Making Analysis Relevant”, Mr. Vince Roske reiterated a recipe provided by John D. (Dave) Robinson, MG USA, Ret, when he was the Director of Joint Staffs J-8. The recipe is provided below:
- What’s the question?
- What’s the “real” question?
- What do the final slides look like?
- What do I already know?
- How do I get the remaining information that I need?
So, this idea of what is the real question is not new and it is not unique to businesses. A client might express a problem, for example as: How do we get more people to use our innovative car buying service? Well, that is not a very specific question, for it potentially has lots of answers, all of which cannot be arrived at through Analytics.
As analysts, we have to draw that real question out through dialogue with our customers. Doing this is often as much an art as it is a science. If we do not do it, we stand a good chance delivering a well-built solution that happens to be the wrong one. I have done this at least twice in the last three years, knowing that getting the question right was paramount. In spite of my failures, I have found some keys to help with formulating the business case.
- Tie it to a metric. Even if currency is not at stake, you have to have a metric to focus on. Otherwise, you can fail to see the forest for the trees. As your project rolls along you will forget what you are building to without a target. It can be Return on Investment, Share of Wallet, Key Performance Parameter, or something else, but you need a metric.
- Do not mention the solution. Do not state your analytic solution method in the business case. The customer may want a model and a model may be the right analytic solution for answering the question, but most customers do not understand what models do and don’t do. The business case should focus on the question/problem, not the solution. If you allow it to, the customer will start building your model for you, without the expertise to do so. I have seen this happen.
- Use timing. Guide the customer to include timing in their business case, i.e., achieve metric A with X months, detect behavior B at least Y months in advance, and so on. With a timing “device” you can help the customer realize the temporal nature of an analytic solution. On one hand, the solutions do not work forever, but many believe otherwise. On the other hand, solutions apply only to events and behaviors in a specified time period. Behavior farther out from the phenomenon or event are quite different from behavior closer to it.
- Tie it to the business unit. That might sound like a no-brainer, but I have seen business units responsible for web activity or call center activity want to tie the business case to a product. If you are trying to improve call center operations, you are trying to improve call center operations, period. Now, you are probably doing so with product sales as a downstream goal, but you cannot put the cart before the horse. Fix your call center problem and then ask a new question.
After going through this process, the analyst may have helped refined the client’s original question and arrive at this business case, for example: Identify customers who would engage in a car buying service 90 days before they reach a buying decision, in order to increase our share of wallet in auto loans.
I am sure there is a lot more that can be said about this subject, but this is what I have learned over time and I am just one analyst. The fact remains that if we do not get the “real question”, then we may not provide the “right solution.”
Questions we can answer with Predictive Analytics
Once we have an appropriate business case for developing an Analytics solution, we have to translate that to an Analytics question. This may seem simple, and it often is, but it must be performed before we delve into model building. For example, using the business case above, the Analytics question then might be: Which customers would have the propensity to engage in a car buying service 90 days before reaching a buying decision. That one is pretty straight forward—others may not be. Other general types of Analytics questions might include:
- Who has a high propensity to purchase?
- Who has a high propensity to engage in a web service?
- Who has a high propensity to shed products?
These kinds of questions place us in the arena of predictive Analytics. I have been often asked how marketers can use predictions to develop more profitable relations with their customers.
Propensity models are what most people think of when they hear “Predictive Analytics”. Propensity models make predictions about a customer’s future behavior. However, keep in mind that even propensity models are abstractions and do not necessarily predict absolute true behavior (see, “What is Predictive Modeling?”). I’ll go through six examples of propensity models to explain the concept.
Predicted customer lifetime value
CLV (Customer Lifetime Value) is a prediction of all the value a business will derive from their entire relationship with a customer. It is based on the Pareto Principle that states, for many events, roughly 80% of the effects come from 20% of the causes. When applying this to e-commerce, it means that 80% of your revenue can be attributed to 20% of your customers. While the exact percentages may not be 80/20, it is still the case that some customers are worth a whole lot more than others, and identifying your “best” customers can be extremely valuable to your business. We can construct algorithms to predict how much a customer will spend with you long before customers themselves realizes this.
From the moment, a customer makes their first purchase you may know a lot more than just their initial transaction record: you may have email and web engagement data, for example, as well as demographic and geographic information. By comparing a customer to many others who came before them, you can predict with a high degree of accuracy their future lifetime value. This information is extremely valuable as it allows you to make value-based marketing decisions. For example, it makes sense to invest more in those acquisition channels and campaigns that produce customers with the highest predicted lifetime value.
Predicted share of wallet
The predicted share of wallet refers to the amount of the customer’s total spending that a business captures in the products and services that it offers. Increasing the share of a customer’s wallet a company receives is often a cheaper way of boosting revenue than increasing market share. For example, if a customer spends $250 with you on auto maintenance, is this 10% or 90% of their auto maintenance spending for a given year? Knowing this allows you to see where future revenue potential is within your existing customer base and to design campaigns to capture this revenue.
Propensity to engage
A propensity to engage model predicts the likelihood that a person will engage in some activity, like unethical behavior or post purchases. For example, a propensity to engage model can predict how likely it is that a customer will click on your e-mail links. Armed with this information you can decide not to send an email to a certain “low likelihood to click” segment. A propensity to engage model can often be used for deciding whether or not certain services should be offered to customers when they log into your website.
Propensity to unsubscribe
A propensity to unsubscribe model tells you which customers not to touch: if high-value customers are at risk of unsubscribing due to e-mail marketing, you need to find other ways to reaching out to them that are not by e-mail. For example, you can predict how likely it is that a customer will unsubscribe from your e-mail list at any given point in time. Armed with this information you can optimize e-mail frequency. For “high likelihood to unsubscribe” segments, you should decrease send frequency; whereas for “low likelihood to unsubscribe” segments, you can increase e-mail send frequency. You could also decide to use different channels (like direct mail or LinkedIn) to reach out to “high likelihood to unsubscribe” customers.
Propensity to buy
The propensity to buy model tells you which customers are ready to make their purchase, so you can find who to target. Moreover, knowing who is ready and who is not helps you provide the right aggression in your offer. Those that are likely to buy won’t need high discounts (you can stop cannibalizing your margin), while customers who are not likely to buy may need a more aggressive offer, thereby bringing you incremental revenue.
For example, a “propensity to buy a new vehicle” model built with only data the automotive manufacturer has in their database can be used to predict percent of sales. By incorporating demographic and lifestyle data from third parties, the accuracy of that model can be improved. That is, if the first model predicts 50% sales in the top five deciles (there are ten deciles), then the latter could improve the result to 70% in the top five deciles. That is, 30% of your customers will be responsible for 70% of your sales.
Propensity to churn
Companies often rely on customer service agents to “save” customers who call to say they are taking their business elsewhere. But by this time, it is often too late to save the relationship. The propensity to churn model tells you which active customers are at risk, so you know which high value, at-risk customers to put on your watch list and reach out to them. Armed with this information, you may be able to save those customers with preemptive marketing programs designed to retain them.
Sometimes propensity models can be combined to make campaign decisions. For example, you may want to do an aggressive customer win-back campaign for customers who have both a high likelihood to unsubscribe and a high predicted lifetime value.
A Word about Selling the Solution
Most analysts do not realize that they are also salespersons. If the client does not use the analytic solution you developed to address their business case, not only have you wasted time and money, you may not be called upon for further consulting. This is one of the reasons we must tie the business case to a specific metric. When we present our Analytics solution, we focus on that metric. For example, rather than providing a detailed analysis of the model effects, the strength of variables, and the log-odds, we should concentrate on demonstrating model performance using ROI. This shows the customer how the model helps achieve their business goals. They will want to know what variables are in the model, but for all practical purposes (other than legalities), it really doesn’t matter.
Predictive analytics models are great, but they are ultimately useless unless you can actually tie them to your client’s business case. This business case must be carefully developed through methodical dialog with your client. Also, it must be tied to a business unit, tied to a metric, related with timing and should not be associated with an Analytics solution. The properly stated business case will drive the appropriate Analytics.
In Predictive Analytics, the solution may very well be some type of propensity modeling. These will help identify the client’s “All-star” customers for purchasing, engagement and share of wallet. It will also help detect their high-risk customers for unsubscribing and churning.
Finally, the Analytics consultant knows they have achieved success when the client implements their proposed solution.
Jeffrey Strickland, Ph.D., is the Author of Predictive Analytics Using R and a Senior Analytics Scientist with Clarity Solution Group. He has performed predictive modeling, simulation and analysis for the Department of Defense, NASA, the Missile Defense Agency, and the Financial and Insurance Industries for over 20 years. Jeff is a Certified Modeling and Simulation professional (CMSP) and an Associate Systems Engineering Professional (ASEP). He has published nearly 200 blogs on LinkedIn, is also a frequently invited guest speaker and the author of 20 books including:
- Data Analytics using Open-Source Tools
- Operations Research using Open-Source Tools
- Discrete Event simulation using ExtendSim
- Crime Analysis and Mapping
- Missile Flight Simulation
- Mathematical Modeling of Warfare and Combat Phenomenon
- Predictive Modeling and Analytics
- Using Math to Defeat the Enemy
- Verification and Validation for Modeling and Simulation
- Simulation Conceptual Modeling
- System Engineering Process and Practices