Whatever Happened to Precise DM Measurement
In ancient times — say 10 years ago — catalogers prided themselves on having a precisely measurable medium. They were the scientists of the marketing world. Most catalogers took the majority of their orders by phone and spent a great deal of effort capturing source codes and order IDs from every call.
As computer technology and database expertise became cheaper and more widely available, we were not only able to measure precisely which customers responded to our mailings, but also what they bought and from which editions of our catalogs. We measured the performance of every square inch of every edition and smugly thought we could answer every question about our audiences and their shopping behaviors.
The Direct Marketing Association defined the direct trade of that day as measurable in costs and results, improvable through testing and analysis, and expandable with confidence.
For better or worse, the advent of the Internet, the expansion of retail and, more importantly, consumers’ desire to shop when they want through the channels they want, made precise measurement a major challenge for catalogers as well as other direct marketers.
Addressing Today’s Challenge
The typical cataloger now receives about 50 percent of its direct orders online. Analyzing individual list performance poses a greater challenge as customers, especially prospects, receiving catalogs have no incentive to provide the marketer with a source code unless it’s imbedded in a promotional coupon code. Product sets and merchandise depth also vary. So if a catalog serves as an effective driver of customers to your Web site, those customers may end up buying items not even featured in the promotion they received.
Ten years ago, we knew all the media the customer had seen from us (namely, a stream of catalogs), and they willingly provided our telephone operators with information about the precise vehicle from which they were ordering. Now we’re lucky to be able to precisely measure half our sales.
On the Internet side, at first we figured that we’d know exactly what banners or keywords drove customers to our Web sites and be able to precisely determine the amount to spend on advertising — and where we ought spend it. Catalogers have been delighted to find new customers placing orders on their Web sites, driven from Google paid search terms and other Web media.
Too few catalogers, though, are looking critically at their Web ad spends. Those who carefully track their Web customers have come to the unsurprising conclusion that many of these “new customers” who came through paid search and other online advertising avenues were already receiving catalogs from them.
Epsilon’s Abacus division has done some groundbreaking research within its catalog co-op database on the consumers who arrive at the typical cataloger’s Web site through Web referrals. “We typically see as much as 85 percent of these buyers were recently mailed a catalog,” observes Epsilon Senior Vice President of Product Strategy Casey Carey, “and on average, 45 percent to 55 percent are existing customers.”
So, catalogers vigorously debate whether they need to mail as many (or any) catalogs to online buyers as to others. Pure e-commerce companies, such as Amazon.com, Zappos.com, eBay and others, develop offline mailing programs and then strive to determine the payoff from those efforts.
Retail Makes It Cloudier
Multichannel marketers with robust retail chains also struggle to determine how much of their Web and catalog ad spending can be properly allocated to store sales. Even companies like AT&T try to optimize hundreds of millions in media spend over outlets that range from the lowly bill insert to sponsorship of the Olympics. Their goal, just like more traditional catalogers, is to drive orders to their Web sites and call centers.
A number of approaches — some homegrown, some available through solution providers — have been developed to allow multichannel marketers to properly attribute ad spending to multichannel results. Fundamentally, there are two basic approaches:
1. matchback methodologies
2. experimental test design.
Matchback technologies examine historical data. Test design anticipates future results. Here, let’s focus on matchback approaches. Multichannel matchback methodologies are usually integrated with catalogers’ processing of databases. To engage such tools, marketers must capture data for the following metrics:
• each customer relating to mailings received;
• e-mails delivered to customers; and
• visits to the Web site originating in paid and organic search, among others.
Incoming orders or retail transactions are then matched back to mailing files, e-mail records, telemarketing logs and others. Each order from a household is matched to all the media you know you’ve targeted to that household over a period of time, thereby identifying the costs of those media.
Then, typically, these methodologies use a series of business rules to allocate the sales from the order to the media that might have generated it. Here’s a very simple example of a business rule: If the household was mailed four times in the past four months, assign all of the sales from the order to the most recent mailing.
Practices and approaches to multichannel performance measurement are evolving quickly, and almost every serious cataloger is experimenting with several analytical approaches. The business is still lacking a universally accepted set of best practices, however.
Allocate Orders Realistically
“The key is to define rules that realistically allocate orders to each channel,” says Bob Gaito, founder/CEO of data services provider I-Centrix. He advises mailers who have invested heavily in search engine optimization, banner ads and affiliate programs to be careful not to overallocate the traditional direct mail channel.
If an order is placed over the Web and it can be determined that it came via paid search, for example, is it really appropriate to allocate the order to a catalog the consumer received 90 days ago? Perhaps, but that’s where the difficulty lies, Gaito points out. Work through these scenarios in detail before implementing a definitive matchback program.
Define two or more different rule sets, and ask your analysts to run each set on your historical data, Gaito advises. This provides a chance to look at the net impact of each rule. “Although this approach may be time-consuming,” he says, “if you’re using the matchback results to drive future channel expenditures, it’s time well spent.”
Intelligent Rules Rule
If you’re diligent in collecting and storing contact data, emerging multichannel matchback technologies can do a great job of matching sales to all of the media that could’ve caused the sale. Without intelligent business rules and a long-term view of the data, they’re less effective at telling catalogers precisely which media or interactions of media caused the sale.
The trick is to develop sets of business rules and test them against your business realities over time — and continue to experiment. As you do this, your complicated multichannel business will become more measurable, improvable and expandable.
Mark Swedlund is a partner and senior vice president of Haggin Marketing, a Mill Valley, Calif.-based direct marketing services agency. He’s an author and frequent speaker on direct response topics. You can reach him at (415) 289-1110 or via e-mail at email@example.com.