Information Management: Strategies for Gaining a Competitive Advantage with Data (2014)

Chapter 3. You're in the Business of Analytics

When it comes to seeing the business landscape to make process change or to strive for maximum profitability, derived from customer- and product-specific catering, you need forward-facing data. You need to know a future that you can intervene in and change. You need to know how to turn the likely future in a more profitable direction.

Keywords

analytics; predictive analytics; data mining; customer lifetime value; churn management; next best offer

I am now going to present a caveat to the premise of Chapter 1—that you are in the business of information. While undoubtedly that is true, it is a form of information that is prominent enough to replace information in the mantra, and that form is analytics.

Basic information operates the business and it is available in abundance to accumulate, publish, and be available from the myriad of data stores I will discuss in this book. Basic information provides rearview-mirror reporting and some nearsighted ability to look forward and get ahead the next few feet.

When it comes to seeing the business landscape to make process change or to strive for maximum profitability derived from customer- and product-specific catering, you need forward-facing data. You need to know a future that you can intervene in and change. You need to know the future that will not actually happen because you’re intervening and turning it in a more profitable direction.

The lack of precise forecasting, caused by constant change, may leave the analytics process in doubt. Yet, you must build up trust in the analytic process through trust in the quality data, the right models, and the application of those models in the business.

What Distinguishes Analytics?

Many approach analytics as a set of value propositions to the company. However, from a data use perspective, the definition of analytic data relates to how it is formed. It is formed from more complex uses of information than reporting. Analytic data is formed from summarized data providing information that is used in an analytic process and yielding insightful information to be used in decision making.

Addressing the propensity of a customer to make a purchase, for example, requires an in-depth look at the spending profile—perhaps by time slice, geography, and other dimensions. It requires a look at those with similar demographics and how they responded. It requires a look at ad effectiveness. And it may require a recursive look at all of these and more. Analytics should also be tied to business action. A business should have actions to take as a result of analytics—for example, customer-touch or customer-reach programs.

There are numerous categories that fit this perspective of analytics. Customer profiling, even for B2B customers, is an essential starting point for analytics.

Companies need to understand their “whales” (most valued customers) and how much they are worth comparatively. Companies need a sense of the stages or states a customer goes through with them and the impact on revenue when a customer changes stages. Customer profiling sets up companies for greatly improved targeted marketing and deeper customer analytics.

This form of analytics starts by segmenting the customer base according to personal preferences, usage behavior, customer stage, characteristics, and economic value to the enterprise. Economic value typically includes last quarter, last year-to-date, lifetime-to-date, and projected lifetime values.

Profit is the best metric in the long run to use in the calculations. However, spend (shown in the bullets below) will work, too. More simple calculations that are simply “uses,” like purchases, of the company’s product will provide far less reliable results.

The key metrics to use should have financial linkage that maps directly to the return on investment (ROI) of the company. Where possible, analyze customer history for the following econometric attributes at a minimum:

• Lifetime spend and percentile rank to date (This is a high-priority item.)

• Last year spend and percentile rank (This is a high-priority item.)

• Last year-to-date spend and percentile rank

• Last quarter spend and percentile rank

• Annual spend pattern by market season and percentile rank

• Frequency of purchase patterns across product categories

• Using commercial demographics (Polk, Mediamark or equivalent), match the customers to characteristic demographics at the block group¹ levels

• If applicable, social rank within the customer community

• If applicable, social group(s) within the customer community

These calculations provide the basis for customer lifetime value and assorted customer ranking. The next step is to determine the attributes for projected future spend. This is done by assigning customers a lifetime spend. Lifetime spend is based on (a) n-year performance linear regression or (b) n-year performance of their assigned quartile,² if less than n years of history is available.

Customer Lifetime Value: The Prima Facie Analytic

CLV=Present Value (future profits (revenues minus expenses) from customer in n years)

There are three major components to the formula: revenues, length of the relationship (n), and expenses.

Revenues. Future revenues are largely based on recent past revenues. With a few years of data and more sophistication, regression of past revenue forward serves to determine future revenues.

Length of the Relationship. Retention modeling can be used to understand leading indicators for customer drop off. Calculating CLV for different estimated customer lifetimes shows the value of keeping the customer for longer periods; this shows the potential CLV. Most organizations that do this valuable exercise are amazed at the potential CLV of their customers and how it grows over time.

The goal becomes keeping the customers with highest CLV as long as possible and reverse engineering the attributes of those high CLV customers for use with marketing behavior, thereby increasing overall CLV. Retention modeling usually accompanies CLV modeling.

Expenses. The major difficulty in computing CLVs is not in computing customer income. It’s on the expense side of the ledger. It can be difficult to determine how to allocate company expenses to a particular customer, but it’s immensely worthwhile.

Choose key characteristics of each customer quartile, determine unique characteristics of each quartile (age, geography, initial usage), match new customers to their quartile and assign average projected spend of that quartile to new customers.

Defining the relevant and various levels of retention and value is an extension of customer profiling. These are customer profiling variables like the ones above except they are addressing the need for more immediate preventative action as opposed to predicting the volume of future profit.

Also, regardless of churn³ potential, the determination of the point at which customers tend to cross a customer stage in a negative direction is essential to analytics.

Customer profiling and customer stage modeling should combine to determine the who and when of customer interaction. Actions are dependent on the company but could be a personal note, free minutes, free ad-free time, and/or free community points.

In addition, in markets where customers are likely to utilize multiple providers for the services a company provides, the company should know the aspirant level of each customer by determining the 90th percentile of usage for the customers who share key characteristics of the customer (age band, geography, demographics, initial usage). This “gap” is an additional analytic attribute and should be utilized in customer actions.

This is simply a start on analytics, and I’ve focused only on the customer dimension, but hopefully it is evident that many factors make true analytics:

• Analytics are formed from summaries of information

• Inclusion of complete, and often large, customer bases

• Continual recalculation of the metrics

• Continual reevaluation of the calculation methods

• Continual reevaluation of the resulting business actions, including automated actions

• Adding big data to the mix extends the list of attributes and usability of analytics by a good margin

Big Data and Analytics

With the ability to explore previously unrealized correlations between certain metrics and/or attributes, big data—and the combination of big data and relational data—greatly increases the effectiveness of analytics. While big data enhances analytics with additional, albeit very granular, data points, it also opens up the possibilities for analytics, taking them into the realm of minute fine-tuning.

If we’re in the business of analytics and analytics are required, it stands to reason that eventually it’s analytics comprising all data, especially the mammoth big data, that will create the leading businesses of tomorrow.

Consider the field of telematics (i.e., automobile systems that combine global positioning satellite (GPS) tracking and other wireless communications for various purposes: automatic roadside assistance, remote diagnostics, etc.), which has serious traction in the auto insurance industry, most famously by Progressive Insurance. If a consumer opts in by placing a sensor device in their car, which feeds its data to the insurance company, they can save money on their insurance. These devices capture fine movements of the car and the car’s location, both of which decrease the odds to very minimal that the insurance package would be less than profitable.

Many using big data analytics to personalize products for customers—such as Netflix, which can recommend movies that model a selection pattern and Amazon, which offers customized recommendations for purchases based on buying habits. Another example is an electric company that offers personalized energy management alerts and recommendations based on smart meters, enabling customers to be in the appropriate plan for them.

Health insurance companies routinely analyze customer health records, correlating granular statistics about patient conditions to outcomes. Green energy systems can increase output with minute adjustments to energy conversion devices like wind turbines. Social media is mined for customer preferences and best times, locations and wording for posting to social networking services. And, of course, for better or worse, governments monitor citizen activity by tracking communications and movements. Most video, phone, and internet activity, with many pixels, sound waves, and fine cursor movement being recorded in subseconds, is big data.

Predictive Analytics

Analytics is a business strategy that must be supported with high quality, cross-platform-border data, as just discussed. The data is formed in order to make predictions about the business. We use “Predictive Analytics” to refer to the class of analytics focused on creating a better future for the company, from grand process change to individual customer interactions.

If done well, predictive analytics help companies avoid business situations analogous to being struck by a bus. Business situations, however, are usually less dramatic and much more nuanced than avoiding a moving vehicle. And, unlike the bus, a company will often not even know there was a situation worth avoiding.

Therein lies the fate of many analytics—in order to prove its worth, you need to build trust in your predictions. Occasionally, I have let predictions of minor doom pass with a client in order to build trust (“We had 142 churners in Maine this month, just like the model predicted. Now can we apply the model nationwide and prevent the churn?”).

Predictive analytics are key to the prevention of loss by fraud, churn and other unwanted outcomes—the equivalents of being hit by the bus.

The Analytics Approach

1. Control data systemically that is detailed, accessible, wide-ranging, and well-performing

2. Focus on a business problem

3. Choose a modeling technique

4. Build models to translate the data into probable business actions, with associated probability of action

5. Avoid the undesired future with effective business action

6. Evaluate effectiveness

7. Refine the model

8. Repeat

Predictions should be communicated as a probability distribution. This turns model output such as “likely to be fraudulent” into “75% likely to be fraudulent.” This can better correspond to a range of actions appropriate to the event (fraud) and the percentage (75%).

In terms of “avoiding the undesired future with effective business action,” these are the next steps that turn all of the data analysis and preparation and the building of the model into business.

Building Predictive Analytic Models

Predictive analytics are applied in the process of determining business events that are likely to occur and be actionable. The probability threshold of “likely” differs from event to event.

Companies that do predictive analytics without attaching a probability to events are seriously impeding the profit potential of predictive analytics.

Company profiles also come into play. Actions take cycles and a fast-moving company driving a strong top line is going to care about something different than an established multinational company with a large customer base and low margins. The former may only prepare for relatively low probability events that come with a particularly bad outcome or try to only enact change that leads to high increased profitability. The latter companies are more risk averse and will seize every opportunity to move the needle even slightly.

FIGURE 3.1 Decision model with churn management.

A predictive modeler might produce a result that indicates a customer is likely to churn, yet the model might not indicate how likely it is to happen or whether the company should care about this.

There is a set of company reactions to any likely unwanted business event or business opportunity. These range from highly invasive actions, like terminating the credit card, to simply changing a metric about the customer that may, one day, lead to a more invasive action if compounded. And, of course, there’s “do nothing.”

Example 1 Customer Lifetime Value

Customer lifetime value is a means to an end. It supports operations as a data point to justify taking other actions, such as whether to market to a person/company, how to support the customer, whether to approve financing, whether to challenge a transaction as fraudulent, etc.

Example 2 Churn Management

When a customer appears likely to churn, companies are increasingly turning to customer lifetime value and other predictive analytics to temper the instinct to rush to salvage the relationship. The operative term is “churn management,” not “churn prevention.”

Regardless, proactive intervention to salvage the relationship, if so desired, is a multidimensional decision.

Example 3 Clinical Treatment

Caregiving organizations want to provide the best care at the lowest cost. To reach this balance, multiple procedures for the patient are considered based on probabilities of efficacy. This efficacy is formed from transaction patterns and, increasingly, big data read from monitoring.

Example 4 Fraud Detection

Predictive analytics is used to determine the potential fraudulent nature of a transaction. Here again, we find that analyzing a transaction without bringing to bear a customer profile built on summarized and recent transactions can lead to false assumptions and actions. Increasingly, a customer profile is required input for any model performing fraud detection.

But on a larger and broader scale, another trend is bringing more data into the predictions, and that includes web-scale data and other big-data environments. For example, customer usage burdens on support can contribute to expenses in the CLV calculation.

Example 5 Next Best Offer

Descriptive modeling classifies customers into segments that are utilized in a large variety of marketing-related activities. These segments should be formed dynamically in conjunction with campaigns and should correlate to the various activities of the campaign. Rather than marketing to everyone determined “likely to purchase,” a “probability to purchase” should be produced and used with other factors that make the effort worthwhile to the company in the long run. A factor like the customer’s income might increase the company’s interest in encouraging the customer through a smartphone or tablet app alert.

A related use of predictive analytics is in decision modeling, which might focus on the next customer interaction and whether it should be proactive and driven by the company (like extending an offer) or reactive (like responding to a financing application).

Proving the need for multiple dimensions in predictive analytics is like proving I should not have stepped in front of that bus. It’s sometimes hard to demonstrate what you have prevented.

Analytics and Information Architecture

You are going to learn quite a bit in this book about the various types of data stores that are legitimate for corporate data. As you travel through the data stores, you may wonder where to place your analytic data and where to actually do the analytic processing. Analytic data is increasingly interesting everywhere processing is done. It will be important to make the analytic data accessible to every data store if the data is not actually in the data store.

In the chapter on master data management, I will make a case for Master Data Management (MDM) to be a primary distribution point, perhaps in addition to being the system that originates some of the analytic data.

As far as analytic processing goes, the goals of all the processing that goes on throughout the enterprise is only enhanced with analytic data. Many enterprises already acknowledge that they “compete on analytics.” Many more will join them.

Analytics are used to assess current markets and new markets to enter. They are used in determining how to treat customers and prospects, in very detailed ways. They are used for bundling and pricing products and services and marketing products and services. And clearly they are used to protect a company’s downside, like fraud, theft, and claims.

You may have a workable supply chain that gets a product to the store (or whatever passes for a store in your business), have low prices, and tactically everything may seem to “work.” That is not enough today. Today, the supply chain must be very efficient, prices should be set based on a firm grounding in analysis, and customers must be known at an intimate level.

The use of analytics comprises the major area of competitive focus for organizations in the foreseeable future.

Analytics Requires Analysts

A host of articles and books have promoted the idea that nearly full company automation is possible, supported by analytics, which are also automated. This automation rivals the intellectual properties of the business analyst who currently translates these analytics into the achievement of company goals. These systems will purportedly exhibit behaviors that could be called intelligent behavior.

The question is not whether or not information systems supported by analytics reason. Of course, they will not. Nor is the question whether or not systems will be able to create the illusory effect of reasoning. They will also clearly get better at it. Information management displays apparently intelligent behavior when it automatically alters in-process promotions to be rerouted to prospect profiles that are responding to the initial mailing. When information management uses analytics to reroute procedures to best-of-breed providers, it displays intelligence. Additionally, when it uses analytics to automatically change pricing in response to demand, it displays intelligence.

However, good engineering cannot yet take the place of the skills and experience of the business analyst. The essence of human reason is the aptitude to resolutely manipulate the meaning of the inputs encountered to create perceptibly favorable situations and arrive at a basic cognitive orientation. The development of this ability within the experienced business analyst makes him or her more adaptable to the business environment. This is what business analysts do—they reason. Analytics don’t.

Determination of the best fit of data for broad organizational needs is another multidimensional reasoning function many business analysts provide to an organization. Business and data requirements are seldom able to be completely coded. Requirements misfire frequently.

At a minimum, and where many programs are today, business intelligence simply provides access to corporate data more efficiently and occasionally does some automated cleaning of that data. While an analyst’s role in manually accumulating disparate corporate data can be diminished with good information management, the higher value-added role of reasoning cannot be. There are, clearly, non-analytical, operational functions being served up to automation, such as industrial manufacturing.

Computers are better at fast calculations than analysts, but that’s not reasoning. There is no scalability from syntactical computation to the input manipulation, abstraction, and perception—the functions that comprise reasoning—based on continued innovation in information management and advances in computational power alone.

The chances of successful analytics efforts significantly correlate with having people with the right characteristics for success on the project. This success goes far beyond technical skills. Analytics work best when they foster organizational communication.

Putting together the data, the processes, and the people around analytics has created the most successful businesses in the world. Welcome to the business of analytics.

Action Plan

• Survey your use of information—are your users using only base information or is it pre-summarized and enhanced to draw real, actionable meaning out of it prior to user interface

• Ensure analytics are a part of corporate strategy

• Enlist the support of business representation in determining the analytic calculations and predictive models

• Determine the analytic data that will be useful to your business strategy

• Understand the calculations on base information necessary to bring the data to full utility

• Understand where the base information resides today and if it is in a leverageable place

www.mcknightcg.com/bookch3