Lists : More Customers, Please!
Using marketing data to find lists and media that match your existing customersFebruary 2014 By Mary Vasek
The idea behind targeted marketing is to make your efforts more effective—to deliver more return for every dollar spent—by targeting people more likely to become your customers.
For most marketers, the best place to start is with your current customers, who like your products and services and buy them repeatedly. So when looking for new customers, it only makes sense to target prospects who share key characteristics with your current customers.
Major brands have long relied on data and analytics to pinpoint those characteristics. Recent improvements have made these tools simpler, faster and less expensive—making data-driven targeting an option for virtually any advertiser.
But I Know My Customers?
Most marketers know who their best customers are—or at least they think they do. But acting on assumed knowledge, rather than applying analytics, is a mistake for two reasons:
1. You may have an outdated picture of your customers. Your core customers change over time, reflecting changes in the population, in consumer behavior, in lifestyle preferences or in purchasing behavior. Even subtle changes in your offerings or pricing can shift the appeal of your products or services.
For example, a recent client, a mail-order garden supplier, envisioned its core customers as older, rural residents with big yards. Analyzing the data showed its customer base had shifted: While the majority were still rural customers, a significant number were now urban or close-in suburban residents, more likely to have patio gardens or window boxes. That change not only affected the lists the client bought, but caused them to reconsider everything from creative approach to the plants and products offered.
2. Even if you're generally right about your customers, you may be missing some important details. For example, you may know your core customers are women ages 25 to 55 with incomes above $50,000. But a detailed analysis might show that you perform significantly better with a subset of that group, such as women ages 35 to 55 with incomes above $75,000. Knowing that, you could focus more profitably on the smaller, more narrowly defined group.
The simplest way to "clone" customers using data, profiling is the first step to reveal targetable insights instead of relying on intuition.
To profile your customers, a data provider appends demographic, behavioral and other consumer data to your customer records, which are then compared to the provider's consumer database. The results show how your customers differ from the overall population in each specific demographic or other attribute.
The profile chart, for example, shows the company has a higher customer penetration among consumers with household incomes between $75,000 and $399,000. Marketing to households with those incomes will produce a higher response rate and yield more new customers than marketing to other households.
Income is only one attribute. A typical customer profile includes data for hundreds of attributes, including demographics such as age, gender, home ownership, family size; interests and hobbies; buying behaviors; and other attributes. Each attribute in the profile is separately indexed against the overall population.
Because each characteristic is covered separately, choosing lists or media based on a profile leaves room for judgment. Marketers must evaluate which characteristics are most important in distinguishing their customers from the general population—and two experienced marketers can reach different conclusions.
The next step up on the data targeting ladder is look-alike, or clone, modeling. The process starts much the same way as profiling. But modeling uses statistical tools and techniques to analyze the results across all the attributes in the profile—both individually and in association with each other.
Non-predictive attributes are eliminated, and the remaining, predictive attributes are mathematically weighted. The final product divides prospects into deciles—with the first decile most like your customers, and the 10th decile least like your customers.
A look-alike model essentially uses data and analytics to make selections for you. You simply specify how many deciles to include in your marketing program.
Modeling can combine attributes in ways that are not addressed by profiling. For example, a model may show women ages 25 to 35 with incomes above $50,000 are strong prospects, but when the age shifts to the range of 35 to 44, the income level shifts to above $75,000.
Moreover, because modeling uses rigorous statistical analysis to examine customer attributes, it can uncover an unexpected predictive attribute that would otherwise be overlooked, simply because it is so unexpected.
Finally, look-alike modeling lets you examine the combination of attributes in your top-performing deciles to gain insights into your targets. This helps you hone product selection, messaging and creative to further increase response rates.
The next step in using data is a response model. This approach builds on the clone model by incorporating results of past direct marketing efforts—comparing the attributes of responding prospects against the universe that received the offer.
The results capture both the characteristics of your customers and those of the prospects most likely to respond to a specific marketing channel, without including customers you may have acquired by other channels and marketing approaches. As a result, you can expect an even higher response rate than you would achieve with a look-alike model, and should reach your marketing goal with fewer contacts.
Which Is Best?
The answer depends on your goals in using data, and your choices may be limited by the customer data you have.
Response modeling requires contact histories from past direct marketing efforts—data on both the consumers who were contacted and those who responded. If that data is not available, response modeling isn't an option.
Profiling is a strong first step, and because it delivers discrete data for multiple attributes, it provides easily understood insights into your customers and a strong foundation for many marketing activities.
Profiling can be accomplished with 3,000 or 4,000 customer records, and rarely requires more than 50,000. It's a good choice for smaller marketers, and for larger marketers who want to segment their customer bases—profiling only customers who spend more than a threshold amount or customers who have stayed for a certain amount of time.
Compared to profiling, look-alike modeling takes more guesswork out of list and media selection, and, by combining multiple variables, generally produces more predictive results and higher response rates.
However, look-alike modeling also requires more records, at least 20,000, precisely because it involves multiple variables. In practice, we prefer to start with even more records, because incomplete data and data irregularities are discarded during the modeling process.
Look-alike modeling also takes longer than profiling, and is more expensive.
The good news is many data providers have automated processes for profiling and look-alike modeling that both reduce costs and accelerate the timeline, even for as little as $500. At that price point, every marketer should consider using data to improve acquisition marketing. It will make your campaigns more efficient, improve response rates and ROI, and help you build a stronger customer base for the future.
Mary Vasek is director, data products, for Richardson, Tex.-based marketing and customer engagement services provider KBM Group. Reach her at email@example.com.