“Past performance is no guarantee of future results.”
Every investor knows those words. They are the standard caveat in every financial offering, a word to the wise against taking the results of the past—achieved under different circumstances than the present or the unknowable future—too seriously when planning investment strategy. Yet, because historical results are easy to collect, store and access—and because hindsight is always 20/20—people persist in predicting outcomes based on past events.
And that goes for marketing strategy.
Consider the granddaddy of all loyalty paradigms—the frequent-flier programs—in which virtually all of the legacy airlines award status to their “best customers.” The usual formula is that a certain number of miles flown earn silver, gold or platinum rank (or something similar), which entitles the elite customer to such benefits as mileage bonuses, favorable consideration for upgrades and early boarding. Once the customer reaches that threshold, she retains the elite status and privileges for the following year, regardless of her subsequent behavior.
In fact, the customer may have earned that status flying on the cheapest fares. And the minute she reaches the status goal, she may move on to another airline, to earn the same benefits from them. In other words, our golden girl may, in fact, not be a “most valuable” customer to the airline. And she may not be loyal at all.
In this case, the airlines mine for gold using the wrong data.
A colleague buys Clinique products for men from Neiman Marcus. The purchase data identifies him as a Clinique customer—and the programming assumes Clinique customers are women. So he receives all kinds of offers from Neiman Marcus designed for women.
‘The same system would inaccurately predict that I am interested in books about childhood diseases because I bought such a book for my wife, a pediatric nurse.’
The marketers there must wonder why he never buys! They could check his first name on their file, which is not a name of ambiguous gender. More to the point, they could look at the total range of his purchases to better understand who he is and what he buys. Or they could simply look at the specific SKUs or product codes of the Clinique products he buys, and they would see that he buys men’s products.
In this case, the store is focusing on only one aspect of the data, and making incorrect assumptions based on that incomplete understanding.
Even great marketers like Neiman Marcus or Amazon.com make the mistake of using partial data and jumping to conclusions. The collaborative filtering technology pioneered by Amazon and used by many marketers today analyzes customer purchases along with aggregated data from other customers to predict future customer interests and push suggestions out to customers.
The key flaw in this strategy is that the data often does not explain whether the purchase was a gift (especially if the orders ship to the same address). So while it is a correct assumption that, because I bought one American history book by David McCullough, I’m probably interested in similar books, the same system would inaccurately predict that I am interested in books about childhood diseases because I bought such a book for my wife, a pediatric nurse.
In this case, the past purchase data lacks critical elements that could ensure more efficient and accurate marketing efforts.
So what’s the solution?
It’s really no mystery. The fact is that past purchase data is useful only when it is fully understood in context. And in some cases, like airline frequent-flier programs, past behavior might not be predictive at all. The most innovative players in the airline world are moving away from the old “status” structure, creating programs that reward fliers more appropriately by dynamically rewarding behavior in the much more recent past—and incentivizing more of the same in the near future.
It is also important for marketers to make use of customer profile data. Ask customers about their needs and intentions, and reward them for keeping their profiles updated. A blend of past behavior with stated future needs is a much better predictor than the historical record alone.
Finally, make sure that you actually use the available data when you develop strategy and craft executions.
Here’s what I mean. Out of professional interest as a marketer, I signed up for email newsletters from a major direct merchant who sells flowers and gifts. I have never bought anything from the seller. But the minute I signed up for email, it began addressing me as a valued customer, even going so far as to tell me I had earned the opportunity to receive one particular offer because I am such a good customer!
What’s so wrong with calling me a customer when I’m not? First, it belies the promise of relevance and personalization that is so central to online communications. If a marketer thanks me for purchases I’ve never made, what other mistakes will the company make? How can I trust its recommendations? I might even question the firm’s honesty and integrity. And at the very least, such a mistake cheapens the offers it makes and undermines, rather than enhances, the value of our relationship.
Simple segmentation and actually reading the purchase behavior associated with my customer record would correct that mistake.
Its troubled industry notwithstanding, Countrywide does a great job of marketing. I have had several home loans from Countrywide, but I am not a current customer. Most marketers dump their past customers back into the prospect pool and subject them to the regular stream of new-customer acquisition mail. But Countrywide actually remembers and respects my status as a former customer, and regularly sends me print mail and email asking me to reconsider its loan products and become a customer again. These offers are well timed and not so frequent as to bug me but sent often enough so that it gives me the impression that someone at Countrywide actually seems to care about my business and have some knowledge and understanding of my potential needs.
In this situation, I tend to trust Countrywide; I appreciate the fact that the company takes the time to know about me and tailor its messaging to that knowledge. Next time I’m considering a home loan, Countrywide will be high on my list.
Past purchase behavior contains some very important data points and should be considered a vital part of the information mix. But you must beware of placing too much importance on the past. And remember that past performance alone is no guarantee of future behavior.