FREE BI Review Site Registration!
Sign-up today and access BI Review on the Web!

Your FREE registration entitles you to:
FREE email newsletters

FREE access to all BI Review content

FREE access to web seminars, resource portals, our white paper library and more!

   

Analytics and Business Performance

The Marketer's dilemma is to identify the unique customer-product-channel combination that maximizes profitability, meets and satisfies business-driven goals, yet operates within typical business constraints. This is a complex task that involves millions of prospects/customers with different demographic profiles and needs and a variety of products to offer. In life insurance, for example, there is term life, whole life, universal life and variable universal life insurance. There are numerous communication channels: email, direct mail, Internet, phone, and face to face. There are simply too many factors and combinations to allow an optimum targeting scheme. This is where predictive modeling comes to the rescue, by allowing us to compute the probability of an individual purchasing a product through a certain channel, and helping us forecast the value of each purchase.

The Power of Analytics

With a nationwide network of more than 12,300 financial advisors and registered representatives delivering financial solutions to more than 2.8 million individual, business and institutional clients, Ameriprise Financial needed to develop highly sophisticated solutions to meet the always-evolving needs of our business.1 We found the answer in analytics, which is a primary focus of Ameriprise Information Management (AIM). AIM's mission is to drive change across the company by applying unique, proven skills and capabilities to our most important strategic decisions. AIM delivers and integrates highly sophisticated analytic capabilities for predictive modeling, targeted marketing lists and campaigns into the company's decision-making processes to drive change and build shareholder value. AIM supports all company marketing programs including direct mail, email, statement inserts, statement messaging, client research lists and telemarketing. We also deliver statistical econometric analytics such as profiling, modeling and campaign results. We employ highly skilled economists, statisticians and econometricians who are responsible for defining, creating and delivering leading-edge analytics. All activities are supported by a 30 terabyte data warehouse that receives product, client and advisor data from most company systems, and is enriched with purchased demographic household data.

At Ameriprise Financial, analytics includes predictive modeling (or econometrics) as well as customer, prospect and advisor profiling. Simply, a predictive model is an equation that identifies main characteristics affecting the likelihood of certain behaviors or attributes of a client or a prospect. There are various statistical techniques that can be used for predictive analytics; ordinary least squares (OLS), logistic models and decision trees are the most commonly used. Typically, the following steps are needed in any predictive modeling process (see Figure 1):


Figure 1: Predictive Modeling

  • Model Design: Every modeling project begins with Model Design. At this stage, a business objective is transformed in to an analytical objective. The target variable and the methodology to predict that variable are identified.
  • Data Pull: Companies usually have transactional and demographic data on prospects and their clients. Extracting the right data for the analysis is crucial. Two sets of data are pulled at this stage: one for model development and the other for model validation.
  • Model Development: This step starts with preprocessing data, and determining the most significant variables that predict our target variable.
  • Model Validation: This is the step where we evaluate the performance of the model we built. We estimate by using the validation data set to make sure it performs well.
  • Implementation: Once a model passes the validation test, it is ready to be used in marketing campaigns. The modeling logic is automated and put into production to create campaign lists.
  • Model Tracking: This is an ongoing process for all the models used in marketing campaigns. The performance of every model in production is tracked regularly to ensure its robustness over time.
  • Remodeling: Predictive accuracy of a model will change over time due to changes in economy, introduction of new products or different channels, etc. When a decrease in model performance is observed, a new model should be built.

The Evolution of Analytic Solutions

Building an analytical practice is an evolution that grows from past experiences and matures over time. Typically, there are four distinct stages: Judgmental Prescreening; Segmentation; Scoring and Ranking; and Optimization. At Ameriprise Financial, we have marketing teams operating at all of these stages (see Figure 2).


Figure 2: Evolution of Analytic Solutions

  • Judgmental Prescreening: In this initial stage, there is no analytical support available, so simple selection or exclusion criteria are based on the marketer's judgment. Response rates for these types of campaigns tend to be very low.
  • Segmentation: This stage involves data analysis and analytical techniques to identify customer segments to target. Marketers using data driven client segmentation can more easily target client groups with appropriate offers and improve campaign effectiveness. Segmentation is the simplest way of analyzing the data and its results are very easy to communicate to non-analytical groups.
  • Scoring and Ranking: In the segmentation phase, only the groups of customers that are likely to accept an offer can be identified. Some individuals within those segments may still be unlikely to accept the offer. Predictive models are constructed to analyze individual customer behavior. The analytics used for predictive modeling are far more sophisticated than segmentation analytics. Typically the predictive analytics involve two levels: probability and profitability predictions
    • Probability Predictions: Probability models help predict the likelihood of each individual to behave a certain way. A variety of these models are built to serve different marketing goals like acquisition, client deepening and retention. The outputs are used to rank order clients based on their probability scores. Marketers then decide the size of their campaign based on their budget constraints and pick the clients with the highest possible probability scores. Campaigns based on probability models result in major marketing cost reductions since the number of individuals to be solicited reduces significantly with this targeted approach.
    • Profitability Predictions: Probability models will improve conversion rates but do not ensure profit maximization, so profitability models also need to be applied. This way, marketing decisions are based on estimates of the probability of accepting a marketing offer and also the change in net revenue due to the offer.
  • Optimization: This is the most sophisticated stage of the analytical evolution. At this stage, there are typically probability and profitability models for each product and channel. The next step is to build optimization algorithms that incorporate all channels and offer combinations for profit maximization while taking into account a variety of business constraints. This helps marketers plan and prioritize their campaigns to come up with an ideal marketing offer mix for a given situation.

The Linkage to Business Intelligence

Although our analytical capabilities are complex and refined, until a few years ago reporting output relied primarily on ad hoc or monthly reports delivered via Excel and email. Outside of AIM there was no centralized, company-wide reporting solution (a single source of the truth) for decision-makers to track and manage their business-critical metrics. A decision was made to create a BI environment fed by AIM's comprehensive data warehouse and delivered through Actuate's Enterprise Reporting Application. A fully scalable data model and SAS ETL provide the consistency, sustainability and data quality we need to run BI.

Ventana Research defines Business Intelligence in terms of insight, empowerment and effectiveness. Insight enables individuals through discovery to understand unknown aspects of the business and gain visibility. Empowerment enables individuals to take action and/or make decisions based on information. Effectiveness enables the organization to gain an understanding of those activities needed to improve performance.

Given AIM's mission to drive change across the company, we did not want to settle for traditional business intelligence reporting of historical and current activity. Our BI environment must be able to reflect our analytics. We needed to deliver reports that could look to the future (see Figure 3).


Figure 3: Decision Intelligence

Today, our "Decision Intelligence" application enables decision-makers to have the facts they need at their fingertips. It is an enterprise-wide strategic reporting solution for decision makers to track, manage and evaluate metrics that are critical to their business. Decision Intelligence applies Ventana's principles by focusing on strategic decision-making with an eye on the future. By integrating predictive capabilities we can deliver truly actionable information and promote forward-thinking decisions.

Opportunities are the key to moving to this next level. Predictive reporting moves us from a reactive client/market environment (learning) to a proactive organization (improving) by enabling decision-makers to take a more holistic approach to maximizing product/client potential. Incorporating a single econometric variable into a traditional report transforms that report. Rather than viewing a client list that simply displays individual holdings, the user is able to rank clients by a predictive variable, or cross-match them using lifecycle value measurements, creating opportunities.

Business Transformation from Analytics

The need to reengineer inefficient processes has grown in importance as companies identify improvements that have a positive impact on the bottom line. Business transformation (BT) is more than cost cutting, important as that is. There is a far greater opportunity to add real value to the company's bottom line by focusing on revenue, structure and strategy. At Ameriprise Financial, annual BT targets are developed to drive efforts that transform the P&L and performance of the company. These targets close the gap between where our performance stands and where it needs to be. Using advanced techniques such as financial and customer analytics, corporate model integration, business intelligence and strategic reporting, AIM's transformation focus is a critical component to business leaders who understand that it is not enough to run a business. You have to drive change, and that's what BT is all about.

Many of the predictive models we deliver are transformational. We are currently assessing multiple projects in the BT pipeline that will deliver real, quantifiable value to Ameriprise. For example, we can cut acquisition costs in several business lines by embedding predictive logic into the client acquisition process. Similarly, we are working on better ways to rank order and target clients that could result in new revenue streams from deepening client relationships. As businesses recognize the benefits of transformation, AIM has become a critical component to many business leaders. We are accountable for shared goals, and indeed, their business plans are heavily structured around the opportunities that BT offers.

Advanced capabilities such as analytics and Business Intelligence are assets to any company. The challenge will always be in integrating them. The temptation is to allow these disciplines to mature independently because they can both be successful in their own right. If you can find the key to delivering predictive reporting you unlock enormous opportunities to the users and the organization.

Reference:

1. Internal management reports and public disclosures as of 6/30/06.


Ozlem Kinav, Ph.D., is Director of Life Insurance and Annuity Analytics in Ameriprise Information Management.

Philip Marquis, CPA, is Director of Business Development in Ameriprise Information Management.

For more information on related topics, visit the following channels:



Industry Vendors