Predictive Marketing Blog

How to Find Your Best Prospects Using Analytics – Part 3

In part 1 and 2 of this series, we examined how marketers sometimes employ incorrect or sub-optimal methods to define their target market for new customer marketing campaigns.

In part 1 of this series we examined a direct marketing campaign with 17,000 sales and identified the demographic target of the new customers based on a demographic market research study of new customers; we found that this targeting approach well underperformed a random set of marketing prospects who received the same offer.

Part 2 further analyzed the campaign and used a penetration index approach to identify demographics where new customers differ from the market. While the penetration index is superior to the market research target, it does not guarantee that the optimal set of demographics are discovered, nor that the most prospects with the highest likelihood of becoming new customers are discovered for targeting advertising.

Part 3 in our series will address the automated statistical approach to identify of your marketing universe of likely new buyers and how the marketer can take advantage of big data to find the best set of targeting criteria.

Identifying the Marketing Universe

Using a 200,000 random sample as the base to compare new buyers to, we can determine how many marketing prospects can be identified by each previously described method (market research and penetration index). The market research target approach finds 43% of the market as meeting the targeting criteria (employed, at their residence for 3-5 years, and under 65 years old), but as identified in part 1, these are very poor buyers who actually underperformed the random names. Using the penetration index approach in part #2, we find a much better prospect list for new sales but the approach only identified 2% of the market as being “best.” While good prospects are found in the penetration index approach, you can’t be assured that you maximize the number of these good marketing leads with this approach.

Figure #6: Examining the random database and using the three demographic variables under analysis, we can compute the size of the market for those identified by market research and penetration index approaches – market research finds the largest universe while penetration index has the best sales rate.

Fig #6 Sales rate and market size

So what method allows you to achieve both goals: finding high penetration characteristics and largest market universe?

The Statistical Approach maximizes the number of best new marketing prospects

The statistical approach utilizes a multivariate regression analysis. The multivariate statistical model is a set of rules, using tables and equations as necessary to estimate (“score”) the degree of penetration. It discovers the characteristics and values that separate your new customers being analyzed vs. your market in general. This statistical model can be applied to a new prospect database for selecting marketing leads with the highest statistical scores of the regression equation; in other words, selecting those marketing prospects with the highest probability of becoming your next new customers. The statistical approach can find best prospects in your current market (defined by zip codes where you now sell) as well as find new markets that have high concentrations of high probability prospects but are outside your current market.

Each record in the prospect database receives a score from the model. While a single record has this probability, it is the grouping of similarly high scoring records that identifies the target segment of good prospects. If a random 10% of your market is penetrated by at 1% and the model finds that the top 10% is penetrated at 6%, then the top 10% is 6 times more concentrated in likely new customers than random. By targeting advertising to the top 10% of prospects, your advertising dollars are reaching 6 times more prospects than random. So you are advertising more smartly.

Figure #7: Using the 17,000 sale records and 200,000 record random campaign, we created a database of non-customers in the same geographic areas as the new buyers. Both datasets were appended with 500+ demographic characteristics. A statistical model was developed that maximizes the likelihood of finding the best and highest volume of new prospects in the data. The model result is applied on a separate random database:

Fig #7: Predictive Marketing findings

In figure 7 the statistical models finds 5% of the random market (with an sales index of 184); using the penetration index approach we found 2% (with an index of 191); the model approaches finds 2.5 times more names.

Conclusion

Marketers should employ the statistical technique to achieve an optimum solution for targeting new customers. The statistical approach maximizes sales, minimizes advertising expense, and allows for managing to an acceptable ROI.

Marketers should not rely on market research targeting to analyze new customers in a vacuum. Nor can marketers rely on the manual penetration index approach as it will not guarantee the maximum number of best prospects and will be too unwieldy to manage effectively.

Marketers should employ statistical methods such as regression modeling as their methodology for comparing hundreds of demographic characteristics of new customers vs. the market’s demographic characteristics.

By using a multivariate, statistical modeling approach and using new customer’s demographic data vs. a random market dataset to create a modeling database, marketers will be assured that the right next prospect’s (next likely new buyer) characteristics are uncovered for exact targeting.

And your marketing dollars will then be used to optimally and most efficiently reach the maximum number of best prospects who most look-like your new customers.