“I know half of my advertising is wasted, I just don’t know which half”. – John Wanamaker

What has been a fuzzy picture at its best historically, has become a blurry, murky mess over the last few years. As Catalog Choice knocks buyers off our mailing lists, consumers tighten their belts, and postal rates increase, the importance of clearing our vision as to which marketing efforts are working and how well never has been more pressing.

The obstacles to vision come from two sources:

1. Sales can take place through any channel, regardless of what type of advertisement “drove” the order; and
2. Consumers often are exposed to several different ads before making a purchase which makes accurate order allocation a challenge.

We can crack the first nut of cross-channel behavior by using a sensible combination of tracking and surveys. The goal here is to get a good sense of how the “untracked” folks got to your checkout line.

Taking the most complex case of the retailer who mails catalogs and has a website, there are three prominent order channels: the phone, the web and the stores.

The traditional technique of using dedicated 800 numbers to represent marketing channels has a few new wrinkles. In addition to having a different number on your website than your other marketing vehicles, consider adding tracking parameters to your online programs and a dynamic gif or text on the bottom of each page that will allow phone reps to ask web callers to “Please scroll to the bottom of the page and read off the number on the bottom of the screen.”

For the website, consider a post-checkout survey asking customers “Why did you buy from us today?” and offer a discount on the next order for those who participate. With a touch of programming this can be done randomly to minimize hassle, and can be done only for customers coming through as untracked by other programs.

For store traffic consider putting a POS coupon in local print ads and online simultaneously, but with different codes. See what fraction of coupon users in the retail store use the web version versus the online version.

Use catalog match-back data to show who got a book before they ordered.

In all of the above cases, we advocate coupling the statistical analysis with some phone surveys of each group. Speaking with a handful of customers periodically is good for your soul and good business, too. Don’t outsource this to your call center personnel, have marketing people do these calls and dig deeper than simply: Q: “How did you find us?” A: “The Web”. The key is in the follow up threads: “How did you find our site? Did you look for us by name? If so, how did you hear about us?” “What made us appealing?” “Would you buy from us again? How?” “Do you read our emails?” “Do you read the catalog?”

Again, you’re not looking at this as a stats project as much as getting the pulse from real people as to how they’re interacting with your company. This doesn’t replace statistical analysis, but it will often add valuable insights on interpreting the data you see.

The channel spillover is half the battle. Figuring out how to allocate credit between different programs is the other half.

As we see it, none of the tools on the market do this well, because the rules aren’t adequately flexible. Crediting the first ad, the last ad, or allocating proportionally are all pretty bad hacks at real understanding. Consider the following Scenarios for Acme:

Scenario: 1

1. Acme Catalog received
2. Search on “Acme Catalog”, click on PPC ad
3. Checkout interrupted as customer searches for “Acme Coupons”, click on affiliate link to get coupon code
4. Purchase

Scenario: 2

1. Acme Catalog received
2. Acme Email received
3. Checkout interrupted as customer searches for “Acme Coupons”, click on affiliate link to get coupon code
4. Purchase

Scenario: 3

1. Customer searches for “Widgets” clicks on Acme PPC link, doesn’t buy but signs up for email
2. Acme Email received
3. Purchase

Scenario: 4

1. Acme Catalog received
2. Search on “Acme” click on Organic link
3. Search on “Widgets” click on PPC link
4. Purchase

Scenario: 5

1. Acme Catalog received
2. Search on “Widgets” click on Organic link
3. Search on “Acme” click on PPC link
4. Purchase

First, think about how your current credit allocation system would stamp these five orders. Then, think about how you as a marketer would allocate credit for them in a perfect world. Indeed, you might want more information, like: how much time passed between the various steps. The answer would effect which channel should get the credit.

In Scenario 2, if instead of searching for an “acme coupon” and coming through a “coupon affiliate”, the customer searched for “widget deals”, clicked on a link for widgetdeals.com – an affiliate – and then clicked on the Acme link on that page, your answer might be different still. Not all affiliates are created equal.

I’m willing to bet that not one retailer would say that her/his allocation system comes up with the same results for each scenario that s/he would assign. Given the stakes in this game and the amount of money involved in mailing books, paying affiliate commissions, SEO optimization firms, designing and managing emails, and buying search ads, research is warranted.

The challenge then, is to take a comprehensive look at what fraction of orders stamped affiliate by your system were folks who also: a) received a catalog; b) received an email; c) clicked on a non-brand PPC ad; d) clicked on a non-brand organic link.

Play this same game with the other channels to look at the differences. If possible, bucket the “coupon affiliates” separately from others to see if the patterns look different. Put some time-frames on these classifications, like “received a catalog within 2 weeks”.

From our perspective a channel should never be credited with driving an order if the preceding search included your brand name in any way shape or form. The real driver was whatever prompted them to search for you by name.

When customers touch multiple messages the only true way to get a sense of whether one or both messages were necessary is to do hold-out tests. These are hard to implement cleanly, but remain the gold standard for determining whether the catalog and/or email are pulling their weight with a given segment.

As retailers study numbers that suggest major changes in where they spend their marketing dollars we think it’s very important to poke those numbers thoroughly to make sure they paint an accurate picture.