So I thought I would do just a true how-to on progressive profiling: no cute song titles, no random quips or quotes.
In order to provide the level of detail I’ve been asked about, this and my next two posts will be dedicated to the topic – so buckle up and enjoy the ride.
The idea behind progressive profiling is allowing email marketers to learn more about customers over time – through direct and indirect interactions – with the ultimate goal of leveraging those learned elements to increase relevance and drive engagement. In theory, this sounds like a desirable way to approach the customer, but it requires some planning to do it right. So where do you begin?
advertisement
advertisement
Areas of Need
The first thing you need to do is identify the data points you want to leverage in
your email program, but don’t currently know about your customers. For example, would it be helpful for you to know the preferred shopping categories of your customers (men’s,
women’s, or kid’s clothing, etc.) or would you like to know how they view themselves (expert, leisure, novice, or what’s- an-airport traveler?).
Defining what you need to know, versus what you want to know, is going to require some self-editing. There’s a litany of data that brands want to know about customers, but when using progressive profiling, it is important to ask questions or position inferences in such a way that you can begin applying that learning immediately. If there is no intent to use the information right away, hold off on asking it.
Direct and Indirect Profiling
There are two ways to gather additional information from subscribers. The first, a direct approach, is to ask a specific question of your
subscriber and solicit a single response. Typically, a single question is posed with multiple answers to choose from, a la survey style. The difference here is that the method for collecting the
response is the recording of a click action – and the result of that click action needs to drive the recipient to a logical destination. So in the above example, if you asked the question,
“What type of clothing do you shop for? Men’s, women’s, kids’?” the act of clicking on the “Men’s” button would redirect the clicker to a page focused
on men’s clothing – and the click action would be recorded against the profile of the subscriber.
The indirect method applies the same concept of recording click activity as
a profile marker, but the click is not in direct response to a question. Success is best realized for brands that have topically driven navigation or featured/unique content elements that
support your content strategy. Imagine a resort that offers various activities on its property, and you include separate “spa” and “golf” offers in one email. Some will
identify with one or the other, still others with both activities, and a fourth group won’t identify with any of the presented activities. The other big consideration here is that one behavior
or action alone does not define a profile attribute. In leveraging indirect profiling, you must build that case out over a period of time.
Map the Path
Once you have determined
what you want to know, and how you want to get to the information, you have to map the questions and the path. And this is exactly where we will start in my next post.
I think it's better to profile which types of product shoppers look for, when searching or browsing, rather than interrupt them with survey-style questions. But does anyone know of published research that compare the ROI of these two approaches?
Pete - I agree that behavioral information such as browse and search behavior are also valuable for understanding the customer - but those are very "moment in time" data elements that may or may not reflect the long term and on-going desires of the subscriber. Finding the right combination of behavioral data and self-reported data elements typically generate the greatest ROI - so long as you are actually taking action against the data.
I have not seen any published data on the topic of ROI as it relates to the source of demographic and psychographic data, but if I do, I will post it up.
Thanks for the comment!
KT