Steps to Improve E-Mail Targeting
Step 1: Basic Targeting -- Interests, Demographics and Source
Interests of the subscriber. The simplest type of targeting, and the most commonly used, is to the explicit interests of subscribers and the experiences of campaign managers for targeting. To send out 20,000 e-mails offering a three-day vacation package in Orlando, FL, a campaign manager might select subscribers who have indicated an interest in travel, who are from a list obtained from an affiliate who publishes a travel newsletter and who have been subscribers for one month or less.
In a recent mailing of more than 70,000 e-mails for an international financial information services company, one incentive for subscribing to its service was entry into a contest to win a free trip. Two different interests were targeted -- people interested in travel and people interested in business, investing and finance. Response rates between the two targets were dramatic, with those interested in business, investing and finance responding at three times the rate of those interested in travel.
Basic demographics.The next simplest type of targeting is based upon demographics. Prospects may have disclosed their sex, age, ZIP code, homeowner status, etc. Targeting based upon this information is done by doing a select from the list to restrict the prospect list to those fitting the targeted demographics, such as males, 20 to 35 years old, living in a metropolitan area.
Using interests as a selection criterion can be effective. When combined with correct demographics, the effectiveness of an e-mail campaign is further increased.
We worked with a supplier of greeting cards to do a series of e-mail campaigns. Adding a single additional criterion -- household income greater than $35,000 -- to the standard selects increased the response rate by as much as 60 percent.
When we worked with another advertiser, adding income and age resulted in a nearly 243 percent response rate increase over using interests alone.
The source of the subscriber. Not all of the prospects in the prospect list may have been acquired in the same way. Some may have been acquired directly and some shared from a marketing partner. These two groups generally would respond differently to the same campaign.
A large consumer packaged goods foods company mailed hundreds of thousands of e-mail offers across seven campaigns using four suppliers of eligible names. All the suppliers recruited names using opt-in messages, but different sources were used to obtain the names. The results varied dramatically across campaigns and across different sources.
Even when the same demographic selects and message were used for each campaign, responses varied widely. Responses for Campaign V ranged from 0.9 percent to 5.4 percent, a difference of six times. Equivalent performance by source is the exception rather than the rule for this manufacturer. The average difference was closer to 100 percent -- a doubling of the response rate. This means that using the correct source can lower the marketing costs 50 percent. In this case, the selection of source was a critical factor to the marketer's bottom line.
Step 2: Add Segments
Basic segmentation.A segment is simply a sublist of prospects that behave similarly. Basic segmentation can take two forms. In its simplest form, it represents subcategories of demographics (for example, segmenting subscribers into high-, medium- and low-income ranges). What makes the segments important is that they are treated differently, with different offers targeted to them. Choosing different offers for different segments can dramatically raise the response rate of each segment and the overall response rate.
More complicated segments are formed by using combinations of demographics, such as income level and purchase history.
It is common to separate subscribers into high, medium and low spenders or into high-, medium- and low-income ranges. Offers can then be tailored to best address the needs of each segment.
The recency of the prospect. When a person first signs up for a service, he generally has a higher level of interest than he will days, weeks or months later. All other things being equal (interests, source of name, demographics), newbies are more likely to respond to e-mail solicitations than those who have been in the database longer. Our experience is that newbies are three to six times more responsive than the rest of the database.
Step 3: Add Response Models
A response model provides a response score for each prospect in the list. The score represents how likely a prospect is to respond to an offer.
Response models are simple to understand. Given a list of 100,000 prospects, a typical offer might produce 1,000 responses. Without a response model, for every 10,000 prospects mailed, about 100 additional responses are generated. A response model identifies which prospects are likely to respond. Using a response model to mail only 10,000 prospects in a list might produce the same number of responses as mailing 40,000 without using a model.
Response models are statistical functions that relate one variable -- the response -- to one or more other variables, such as the age of the prospect, the number of times the prospect has responded to similar offers, the total amount of purchases by the prospect, etc.
Response models are used for several reasons, including:
• To reduce the size of the mailing. Response models produce the same number of responses with fewer e-mails. This is important for several reasons and is often misunderstood. Though the per-unit cost of e-mail -- compared with conventional direct mail -- is low, e-mailing every one on a list for all offers is usually not a good idea, if only because this quickly burns out a list.
• To delay burnout. With a response model, a prospect is more likely to receive e-mail that is of interest and less likely to burn out after receiving mail that has little or no interest.
• To monetize an entire list. A common practice is burn and churn. A prospect is acquired, mailed very frequently since the response tends to be high, burns out and is dropped. With a response model, prospects likely to respond to specific campaigns can be identified, allowing a list to be monetized much more effectively than would be possible without using a response model.
• To order offers. The order of offers is often critical. There is a tendency in e-mail to fill an e-mail with many offers. On the other hand, the lower an offer appears in the e-mail, the less likely a response. A response model can place the best offers in the best order to maximize overall response.
Simple Things to Watch Out for
• Good targeting cannot improve a bad offer or a bad treatment. The response to a campaign depends in large part upon three things. First, the offer. Prospects may not want a CD at the price you are asking. Second, the treatment. Fancy flash may irritate some prospects while others may be moved by free shipping. Third, the list. The prospects in a purchased list may be too burned out by previous e-mails even to look at the subject line of a new e-mail from the same marketer. No matter how good the targeting is, if the offer is bad or if the treatment is bad, targeting can be of only limited value. It is sometimes easy to forget this -- until the results of the campaign come back.
• Mailing an entire list is not targeting. Say you have 100,000 prospects and use targeting to send a campaign to the 10,000 most likely to respond. Good targeting can improve the response rate. However, if you mail to all 100,000, no targeting is required and you should expect the response rate to be lower.
• Do not forget the importance of the test-measure-refine loop. There are always surprises. Creating different tests, measuring the outcomes carefully and refining as necessary is the best way to design a campaign.
• Some lists are just plain bad. Some lists can look great, but after a few mailings it is clear there is no targeting you can do to lift the response rate off the floor. Move on and remove the names from the master list.