By Cheap Calling Cards

ith the downturn in the economy and the maturation of markets, the importance of customer support has risen dramatically. The adoption and integration of customer relationship management (CRM) systems throughout the 90's has provided a means to collect the valuable data on an individual customer. However, it has not turned the corner to enable companies to exploit that data to generate additional revenue. Each customer contact -- a support call, payment received data, or marketing offer, for example -- is logged into customer portfolio records to create a wealth of information for call service representatives to review and act upon. Although the support side of CRM systems has improved vastly, the use of CRM technology as a sales and revenue generator has not met expectations. Despite their investment in tools to collect immense amounts of customer data, companies have struggled to leverage this investment to extract the true value of their existing customers.

Optimization is a promising new technology that is helping companies maximize information contained in CRM systems to better understand their customers. This is increasing the depth of customer prospecting and generating new cross-sell and upsell opportunities, as well as strengthening customer retention -- all increasing the value of their customer portfolio.

All About Optimization
According to Tower Group, optimization is a mathematical methodology that allocates finite resources across multiple competitive, conflicting, and overlapping initiatives, each with unique constraints, to achieve an overall objective1. Optimization theory includes a variety of advanced techniques, including calculus of variations, control theory, convex optimization theory, decision theory, game theory, linear programming, Markov chains, network analysis, and queuing systems.

From a practical perspective, optimization is needed when business goals -- such as maximizing profits or revenue -- have multiple variables and constraints and therefore become difficult or impossible to achieve using conventional mathematical techniques. For example, a large telecommunications company might use optimization to determine which offer to make to which customer in order to retain and maximize customer portfolio value. This type of data becomes invaluable to the marketing departments as companies connect their customer touch-points to ensure that the most effective offer is being made with each customer interaction.

Companies with very large customer portfolios need to solve a range of other constrained marketing problems, including pricing, prospecting, and retention management. During the pricing process, companies need to factor in price points that deliver the best combination of acceptance rates, long-term profitability, and risk. When prospecting, companies seek to balance the attractiveness of introductory offers against the need for long-term profitability to achieve customer acquisition goals.

In a recent study, Knowledge Capital Group found that for optimized marketing programs, "The more penetrating the analysis, the higher the opportunity for developing programs that will strengthen customer profitability and expand customer relationships (through cross-sell/upsell). In many cases, investments in such capabilities are paying for themselves within a few campaigns…it is not unusual to generate returns of 250% or more."2

Optimization Vs. Conventional Processes
The idea of optimization is to take actions that generate returns with the most optimal (e.g., profitable) results.

Figure 1 below illustrates how optimization differs from a conventional "sort by score" approach to maximizing profits in the context of a marketing cross-selling campaign.

Conventional Approach
("Sort by Score") Optimization Approach
For each customer, use predictive models or other techniques to assign a model score for each available product based on the expected long-term revenue if that product is offered, and then:
Select highest priority product.

Rank customers by score for that product.

Select customers by score, top scoring customers first, to satisfy product requirements.

Remove selected customers.

Repeat.
Simultaneously evaluate all product and customer combinations to determine the most profitable matches within the given constraints (such as minimum number of customers that must be selected for a specific product).

Figure 1. Optimization Vs. Conventional Approach To Cross-Selling

Note that due to the sequential nature of the conventional approach, some customers are matched with less profitable products simply because those products are evaluated first. More profitable product matches are not considered once a successful selection is made.

In a small-scale illustration of the process, assume there are two products to be offered (P1 and P2) and that there are a total of five accounts that can be cross sold (A1, A2, A3, A4 and A5). P1 has a higher priority than P2 for account selection.

Assume also that there are predictive models in place for each product that will permit calculation of potential profit (by combining a likelihood of response prediction with the revenue value for each product if purchased).

This information could be summarized as in Table 1 below. Assume that based on the specifications provided, we must assign two accounts for P1 and two accounts for P2.

Account P1 P2
A1 $0.30 $0.90
A2 $0.70 $0.80
A3 $0.60 $0.10
A4 $0.40 $0.50
A5 $0.90 $0.10

Table 1. Estimated Product Profit

Simultaneously evaluate all product and customer combinations to determine the most profitable matches within the given constraints (such as a minimum number of customers that must be selected for a specific product).

Based on the conventional approach, which allows products with higher priority to select the most profitable accounts first, the following product assignment would occur:

Account P1 P2
A1 $0.30 $0.90
A2 $0.70 $0.80
A3 $0.60 $0.10
A4 $0.40 $0.50
A5 $0.90 $0.10

Table 2. Conventional Approach

A5 and A2 would be selected for P1 for an estimated $1.60 in profit
($0.90 + $0.70).
A1 and A4 would be selected for P2 for an estimated $1.40 in profit
($0.90 + $0.50).
The total profit for this method is $3.00 ($1.60 + $1.40).
In contrast, the optimization approach would view all combinations simultaneously and assign products in the following manner:

Account P1 P2
A1 $0.30 $0.90
A2 $0.70 $0.80
A3 $0.60 $0.10
A4 $0.40 $0.50
A5 $0.90 $0.10

Table 3. Optimized Approach

A3 and A5 would be selected for P1 for an estimated $1.50 in profit
($0.60 + $0.90).
A1 and A2 would be selected for P2 for an estimated $1.70 in profit
($0.90 + $0.80).
The total profit for this method is $3.20 ($1.50 + $1.70).
This example illustrates in simple form how optimization achieves higher profits than the conventional approach to offer selection.

When determining the optimal assignment, optimization evaluates all customers eligible for a product and all imposed constraints, such as budgets, capacity, contractual obligations, policies, etc. In some cases, the product with the second (or third, or fourth, etc.) highest value may be selected as optimal based on other customers eligible for the product.

Constrained optimization of the marketing problem is difficult, with exponentially more possibilities, as shown below in Figure 2.

Customers Actions Possibilities Solution Time
4 2 6 600 picoseconds
8 4 2,520 252 nanoseconds
12 6 7,484,400 748 microseconds
16 8 81,729,648,000 8 seconds
20 10 2,375,880,867,360,000 66 hours
24 12 151,470,000,000,000,000,000 480 years

Figure 2: Exponential Solution Time Growth

An Example from Retail Banking
The results that can be achieved by applying optimization have already been proven in the retail banking space. An example is a leading financial services company with a diverse set of banking products and a growing credit card business that needed to better manage and respond to its 80,000 retention-related calls per month.

In this case, after reviewing six months of data, the results exceeded expectations, and included a 33 percent increase in the rate of accepted offers, a 5.6 percent increase in saved accounts, and a reduction in voluntary attrition of 13 percent. In addition, accounts handled by the optimization solution showed increasing levels of performance compared to existing methods, including increased balances over initial levels, and increased annualized profits, finance charges, and receivables.

Conclusion
Optimization is an important new technique for extracting additional profits from existing business processes, particularly those that involve large numbers of customers, products, and interactions.

Companies in financial services, telecommunications, and other industries with large customer portfolios are now using optimization to quickly and dramatically improve revenues and profits by augmenting, and sometimes even replacing if necessary, the selection processes in existing customer treatment programs.

It can be expected that optimization will follow an adoption curve similar to other operations enhancement technologies. As optimization techniques become more widely adopted, the use of optimization will shift from a competitive advantage to a competitive requirement for leading companies.

1Tower Group, "CRM IT Spending: What Are Retail Financial Services Institutions in North America Spending on Customer Knowledge Technologies?", November 2001.

2Knowledge Capital Group, "KCG MarketView: Marketing Optimization", 2000.

Craig Macdonald is the group vice president of the Opportunity Solutions Group at HNC Software, a leader in optimization software solutions. In this role, Macdonald is responsible for the vision, development and delivery of the Opportunity Solutions Group offerings that are focused on helping customers drive incremental revenue by optimizing marketing decisions, enabling customer-centric programs, and increasing customer loyalty.