Recently a mail order cosmetics and hair products company launched a test direct mail campaign in which they sent coupons to a percentage of their in-house mailing list. The conversion rate of the campaign was about 10%. The marketer then decided to roll out the campaign to an un-mailed portion of their list, but wanted to focus on the segment that had the highest propensity to convert. After some analysis they found that 45% of the conversions from the test campaign were from blondes, 35% brunettes and 20% were other.
Based on this information, they rolled out a new campaign only to blondes, expecting better than a 10% response rate. But the conversion rate of the rollout dropped to 8.3%. After some serious head-scratching they decided to analyze the people who received the coupons but did NOT convert, and found that were 55% blonde, 20% brunette and 25% other. So they launched a third campaign directed only at brunettes and produced an 18% conversion rate.
Converters & Non-Converters in The Attribution Formula
When any marketing attribution management solution is developed, a decision must be made right upfront on which available data is going to feed the attribution process. Obviously the mathematical science that’s employed to calculate the amount of credit to be attributed to every channel, campaign and tactic used by marketers can only utilize the data that’s made available. So if certain data within an organization is excluded and deemed unimportant -- such as the “non-converter” data in the example above, before that data has been mathematically proven to be unimportant -- then that solution is inherently suspect.
And frankly, doesn’t the application of even a simple, common-sense barometer find it obvious that only when the traits associated with the non-converting population are compared to those associated with the converting population can marketers identify the difference between the two (and by implication the traits that have an impact on conversions)? But despite this apparent no-brainer, attribution solutions exist in the marketplace that exclude all data (traits) from the non-converting population.
Only Your Hairdresser Knows for Sure
Now, to put this in a true cross-channel attribution context, instead of hair color, think of brunettes as channel “A” and blondes as channel “B” – or, since attribution is actually a multidimensional exercise, think of brunettes as a given set of channel, publisher, creative, size, price or date traits and blondes as a differing set of traits. In a very real sense, a marketer only has the complete picture of her marketing performance if she looks at the population that has been exposed to her marketing efforts as a whole—and can prove or disprove the importance of ALL the data (traits) that has been included. Only then can the most informed, accurate conclusions be drawn from the attribution solution, and most effective optimization strategies be enacted as a result.