Customers and prospects present unique profit potential. Economically originated efficiencies are being forced upon marketing budgets everywhere. This "doing more with less" requires higher rates of return on campaigns than what was previously acceptable.
Campaign management is the process of planning, managing and assessing outbound campaigns. Although consumer marketing examples will be used, this article focuses on the use of data warehouse data in campaign planning, regardless of industry. The data warehouse provides the empirical data that is needed to determine the ideal size of the campaign market, develop the target market for the campaign and assess the campaign's effectiveness in near-real time. As you will see, the key to campaign planning is up-front segmentation of the customer/prospect data into actionable segments. That's the science part, anyway. Marketing is, and always will be, a crucial art form, regardless of how much we think we can automate the process.
The near-term benefits of targeted campaigns would come from the simultaneous reduction in campaign-related expenses and increase in purchase volume by customers. For the marketing department, this would likely mean decreases in outbound marketing budgets enabled by more precise targeting than is currently practiced. It is not unusual for targeted campaigns to simultaneously increase effectiveness by boosting spend from the targeted customers and decrease campaign expense. This swing could be several million dollars in a positive direction each year.
Even as event-based triggers create one-to-one marketing opportunities, outbound targeted efforts will continue to be a primary means of attracting new customers in today's saturated marketplaces where increasing revenues, market share and new customers come from our competitors' customers. The customers that represent our growth opportunities do not present themselves and their information at our storefronts, kiosks, Web pages or in our mailboxes without provocation.
Is your data warehouse ready to support campaign management? Most data used in campaign management is inadequate. The first step is to investigate the relevancy of the current data and develop customer/prospect attributes and segmentation possibilities with the existing data.
Once the existing data is organized, the second step is to determine the desirability/necessity and extent of augmenting customer/prospect attributes with internally available data, commercially available demographics and other attributes, such as clusters. These attributes should consist of sales and geographic data as well as customer states which predict what customers are likely to do next.
Finally, the third step is to formulate marketing plans to make best use of the campaign management process or software within the operational marketing framework. Many software products enable companies to not only deliver coordinated messages and value propositions via multiple channels, but also enable channel parity the wide distribution of consistent, timely, integrated, usable customer data. Consider its use as you transition to year- round continuous campaigning on a continuous cycle.
Preparing the Customer Data
Complete Customer Identity
It is possible that customers are submitting only minimal, if any, identification. Because of this, retailers, in particular, have been restricted to marketing to anonymous customers. This results in a corporate marketing database that is very difficult to use for developing customer relationships. Determine an appropriate set of methods to identify every customer. Customer-identifying operational systems and procedures will help this effort.
"Garbage in, garbage out" applies to customer data. While it is possible to impute missing data, front-line employees should be incented to accurately identify customers and their attributes as a way to help this effort. Quality metrics need to be added to the existing quantity metrics for assessing the performance of call center and order entry personnel.
Once customers are identified, taking the identification to the third-party marketplace progresses the identification. Among the data sourced, feedback national change of address (NCOA) standard address data into the data warehouse on a periodic or continual basis. Better address quality increases delivery rates and complete identification rates for unidentified customers.
Perform data quality checks including existence counts, minimums, maximums and counts by value for each customer attribute internal and external as they are being loaded into the customer data warehouse. These reports often find that customer data is incomplete and inconsistent.
Perform pattern analysis on the customer data warehouse to identify significant customer purchase pattern variation in the marketing life cycle. Once identified, these core purchase patterns can be used as a basis for segmenting customers for marketing campaigns that match their particular purchase patterns and life cycles.
Once customer data is augmented and profiled, the key to any targeted marketing effort is identification of customers by rank. Once identified, customers are assigned attributes and segmented according to their personal preferences, purchasing behavior, state and characteristics, and economic value to the enterprise. Economic value typically includes last quarter, last year-to-date, lifetime-to-date and projected lifetime values.
Most key attributes have financial linkage that maps directly to return on investment (ROI) of the company. Where possible, analyze the customer data warehouse purchase history by customer for the following econometric attributes at a minimum:
- Lifetime spend and percentile rank to date. This is a high-priority item.
- Last year-to-date spend and percentile rank.
- Last year spend and percentile rank. This is a high-priority item.
- Last quarter spend and percentile rank.
- Annual spend pattern by market season and percentile rank.
- Annual spend by vendor and percentile rank.
- All of these attributes for profit.
- All of these attributes for projected future spend and profit.
- Frequency of purchase patterns across product categories.
- Using commercial demographics (RL Polk, MediaMark or equivalent), match the customers to characteristic demographics at the census block and block group levels. A census block is approximately 15 homes. A block group is approximately 100 to 150 homes.
Consider the improvement some basic econometrics give us in this example. Given the economic basis for selecting customers as a primary criterion, the changes in customer spend and profit per transaction from a baseline point provide a nearly instant indicator of success.
Using basic demographics, we would identify a person as: female, 25 to 32, Midwest U.S., suburban. In typical marketing, this customer is promoted to buy a pair of Levi's jeans. In a typical evaluation, the customer is put on a targeted list. When she buys a pair of Levi's jeans, a hit is counted. The percentage of hits indicates the effectiveness of the campaign.
From this perspective, on a mass mailing, a hit rate of 2 percent would mean that two out of every 100 targeted consumers purchased a pair of Levi's jeans. With appropriate co-pay from the vendor, this may even be considered an economic success.
Shifting to an economic basis for targeting, the market segmentation cell could be: female, 25 to 32, Midwest U.S., suburban, prefers Calvin Klein's, average per-pair spend $125. A brand-switch campaign to Levi's causing the customer to buy a $35 pair of Levi's would be considered a success when counted in this method.
When counted against the economic segmentation, the $35 drags this customer's metric down from the previous per- pair spend of $125, regardless of what pants were purchased. It is likely that an enterprise-based marketing campaign would have sent a different, higher- priced campaign to this customer.
New Customer Profiling
Once the customers that have ample data are profiled, the remaining customers need to be identified and their comparable econometric data modeled from customers from similar block group demographics. Have these "cloned" customers "inherit" the economic profile of known customers in their block groups or with customers sharing similar demographics until the new or unknown customer has his/her own profile.
This practice, known as profile inheritance, can effectively and quickly create a fairly robust basis for economic analysis of marketing efforts. Different economic performance by the cloned customer than the original profile gives a good financial and qualitative indication. Different percentage responses by the cloned customer than the original customer mean that expectations for the profile may need more detailed modeling based on the different results.
Standard statistical software can be used to determine the significance of multiple economic-based segmentations and combinations. The software can be used to generate predictive models for the economic performance of customers sharing particular attributes. This would permit the gradual increase in accuracy of attribute selection for campaigns over time.
Customer State Modeling
Individual customers are initially assigned a current state based on the history of their interactions with your company and other customer-related information (see Figure 1). These states provide a way to mathematically summarize the current condition of each customer relationship for purposes of predictive modeling. They allow us to predict what customers are likely to do next. Customer states should be updated automatically from transaction-level data entering the data warehouse such as product purchases, contact updates, inquiries, complaints, back orders, returns, demographic changes and responses from customer surveys.
Figure 1: After analysis of the client, their customers fall into 6 states which form the basis for further state modeling.
Once a customer's state is known, sales, marketing and business development questions can be supported. These questions include:
- What product or product bundles should I offer each customer next?
- When and how often should I contact a customer?
- In what order should I offer my services to each customer?
Results can be deployed in real time to condition the most recent purchase. They allow responses from marketing programs to dynamically change with each customer interaction.
The idea of the state-based modeling is to find the unique combination of specific factors that best identifies the customers most likely to expand their relationships or churn over time.
Recent modeling with a cable company showed customers who subscribe to two products are almost twice as likely to drop a certain premium channel than customers with three products (6 percent versus 3 percent probability). However, separation for basic and digital services ranges only from approximately 1.8 percent to 2 percent. We want good separation between different products for customers in the same group. Customers who have subscribed to six products have the same chance of dropping basic, expanded basic, digital and premium channels. Build states that capture the important variables and past customer behavior for consistently predicting customer behavior over time.
From any state, the likelihood to drop any product, add any product, transition to any other state or churn should be determined and used as a consideration in building outbound lists for campaigns. The accuracy depends on how well you identify cause/effect relationships in available customer data.
Finally, use the customer identification, profiling and states in small-scale campaigns. Evaluate campaign performance from precampaign, during-campaign and post-campaign perspectives. All these periods should be tracked in the data warehouse following these guidelines:
- Use economic criteria primarily for the evaluations of success and failure even if noneconomic attributes are used in building the campaign.
- Measure the uptake and falloff effect across the various evaluation periods.
- Use standard statistical software where possible to assess the significance of results, keeping in mind that positive promotion numbers are not always positive to the overall bottom line. In this situation, the term significance means significantly different from chance or random outcomes.
- Systematically and continuously monitor known customer performance against chosen state criteria. Reevaluate state criteria periodically.
Small-scale marketing can be gradually expanded as comfort grows and the customer attributes are expanded, the data quality improves and the customer state management and profile inheritance improve.
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