In Eastern religions, gurus tell their disciples that they must step out over an abyss, and then the bridge will appear. The lesson is that only blind faith will save you. In the real world, you cannot be so cavalier with your actions. After all, if you step out over a deep chasm and fall in, the only one who suffers is you. But when you make business decisions blindly, you pull a lot of people - and a lot of profit - down with you.
Despite the hazards, many organizations continue to make decisions on blind faith. As a result, they make mistakes that cause damage on multiple levels. Call centers that contact customers for acquisition, relationship building and collections are particularly vulnerable to these missteps. Part of the problem is the tools they use to guide their strategies - in particular, relying blindly on the rank-ordering achieved with behavior-scoring applications.
Behavior scores don't unearth the individual insight that can lead to optimum customer contact decisions. Decisioning based on behavior scoring results only in the segmentation of customers broadly - that is, grouping together those with similar attributes and then treating them as a group. Consequently, because we know that no two customers are the same, companies incorrectly interact with customers in ways that waste resources, miss opportunities and increase customer attrition.
To go beyond making decisions based upon broad behavior bands requires methods capable of reaching down to the customer level - revealing each customer's value and propensity to act to various stimuli. Innovative "action analytics" uses advanced modeling techniques and decision methods to make precise customer-level decisions.
Many progressive companies have attempted to crack the code on customer-level decisioning by deploying intelligent predictive analytic applications. And while their objectives are clear - to identify the most effective way to treat each account at a micro level and maximize the value of this data for optimal decision-making at several key customer touchpoints - their results have often not been successful.
The key to success lies in the ability to leverage data for a precise prediction, create an actionable decision framework and create decisions applicable to the business challenges at hand. Once these goals are met, many applications of action analytics can be deployed, including:
- Assessing the risk and fraud propensity of new and existing accounts,
- Identifying accounts at risk of attrition and taking actions to prevent attrition,
- Targeting customers for up-sells and cross-sells,
- Pulling high-value inbound callers out of waiting queues and routing them for optimal service levels,
- Discovering the best collections approach for each delinquent account, and
- Determining the best time of day to call delinquent customers.
Employing intelligent action analytics requires four key steps.
1. Explore multiple databases. What data is needed for account-level decision-making? Reaching down to the customer level versus behavioral-segment level requires assessment of customer data from traditional and nontraditional sources. Typically hundreds of candidates for the final models must be considered from multiple sources.
The next question: What is the most predictive data of any customer subpopulation? Because individual subpopulations behave differently, they have different attributes that are predictive of their behavior. So, answering this question requires an effective and exhaustive exploration of the potential variations in segment performance.
To ensure a consistent and thorough exploration of these data sources and aggregation of predictive factors, advanced techniques are employed to detect subtle variations in subpopulations and to segment effectively and extract as much information as possible for those groups.
2. Maximize data value. A significant amount of data preparation occurs long before any modeling technique is applied. Maximizing the value of a given set of data sources is imperative to building the most effective behavioral predictors. But massive amounts of data can be challenging to process. Innovative processes for automation include:
- Categorical variable performance grouping,
- Numeric variable performance grouping, and
- Segment imputation.
3. Build action models. Action models aim to precisely predict customers' behavioral responses using data developed from experiments designed to capture customer-level behavior or from data that is gleaned from traditional champion and challenger testing. After identifying which variables out of hundreds or even thousands are the most valuable for predicting customer-level action, action analytic models are built with the idea of quantifying an account's propensity to act to specific actions that are being analyzed.
A variety of modeling techniques are used in action analytics, and these scorecard-building methods are designed to take multiple data inputs in a hierarchical manner. Specifically, and when merited, they can allow the creation of multistage, sequential models that incorporate different families of predictive characteristics. This capability is particularly helpful when designing a solution to fit a certain operational window. For example, in transactional fraud scoring, a model segment summarizing the information value of account and personal data (available at day's end) eliminates the need to carry more than that stage's score into the authorization system.
The score can then be combined with recent authorization activity to create a final prediction. Also, additional stages in the model can be fired post-authorization in order to incorporate data for operational decisions that would otherwise be inaccessible due to latency concerns. Effectively, a multistage scoring approach allows for maximum model strength while reducing the computational intensity in the inline authorization process and ensuring expedited but precise decision support.
4. Undertake optimal decision-making. Action models are the basis for developing and deploying effective decision-making. In a simple framework, they can be used to determine the best actions for the account or customer. This can be accomplished within the standard decision-making structure, such as a rules-based process. Alternatively, they can be used in an optimization structure to balance action assignments against real-world operational and business constraints.
In either scenario, rules or an optimization framework action, analytics can answer important questions at several critical customer touchpoints, including acquisition, relationship building and collections, as described in the following examples.
Acquisition. Action analytics can predict which customers will take advantage of your relationship-building offers and which will just take advantage of you. For example, a credit card company could offer nonprime customers credit cards with low credit limits until they identify that the new account is a good high-value customer candidate.
Relationship building. Action analytics can predict which high-value customer will respond positively to up-sells and which will not. This will both improve companies' relationships with high-value customers and reduce the risks of making offers to the wrong customers.
Collections. Action analytics will predict a new level of insight, such as which high-value customers should not receive calls to pay a delinquent account, and which low-value customers should be routed directly to collections agencies.
In whichever way companies choose to employ action analytics, they will gain the same results: a clear-eyed, dependable, results-focused prediction of which action to take with each customer for maximum impact at all times.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access