Lessons I’ve Learned About Personalization (Or, What Catholic Nuns Always Knew)

Brands have more customer-supplied data than ever before, and consumers now expect to see personalized product recommendations based on their interactions with eCommerce sites. This is awesome for both brands and purchasers, but, like so many amazing tech innovations, practical implementation is often easier said than done.

After spending nearly a year with a direct role in implementing personalized product recommendations in behavior-triggered emails for an apparel eCommerce site, I have a few takeaways to share, which I hope can help your brand leap the hurdles and quickly arrive at the fun part.
On long car rides as a child, I loved when my mom would pass time relaying anecdotes from her youth. (I promise this will relate to personalization.) She had five sisters and one brother; they were Irish; they were mischievous; they would tease each other to tears. One of my favorite stories showed my grade-school skinny (before skinny was cool) mom at Catholic elementary school, trembling in fear at her desk as Sister Mary Margaret would slowly pace by, wood ruler tap-tapping in her palm, inspecting pupils’ uniformity and cleanliness, as individuals and as a group. Un-tucked white collared shirt? Thwack! Dirty finger nails? Smack! Too-short plaid skirt? Whap! One pupil out of line? The entire class punished.

The nuns had their canny eyes out for uniformity and cleanliness, and they knew the importance of the group as it relates to the individual. Awareness of these same elements, as they apply to data, can lead to success in implementing personalization in eCommerce digital marketing.

  1. Uniformity:
    You should have a uniform method of attribution for existing and upcoming products in your digital inventory system(s). It should be practiced without exceptions. This will likely involve teams outside of digital marketing. This is more important the more narrow your strategy of recommending products similar to the product the person purchased or browsed or abandoned in their cart. Recommendation engines rely on being tapped into your digital inventory system(s) and being able to look for certain defined parameters that you set. As an example, the standard attribution for socks could be the tags: socks, accessories, (specific gender), (specific category). If a new sock is added to the inventory and is not tagged completely, (let’s say it’s missing gender and category), then the recommendations for someone who buys that sock may not be in the same gender or category as the purchased sock, which may not be relevant for the consumer, or according to the business marketing strategies.
  2. Cleanliness:
    A personalization engine is only as good as its consumer data. The data you collect from consumers should be clean and without inconsistent wrinkles. (See how I’m so subtly trying to tie this to the Catholic uniform illustration! Genius.) This is important to keep in mind when trying to execute business strategies based on certain demographic profiles. It’s difficult to achieve that personalization if the audience file does not have complete and consistent types of date.
  3. Group Affects the Individual:

    I think it’s important to remember personalization engines are not curated dynamic content engines. The products they recommend are based on affinity data, for example: which of your brand’s products is purchased the most by a certain demo; which of your brand’s products is purchased the most with another specific product purchased by a certain demo, etc. This can be frustrating as a marketer because, basically, what the engine says is the best product to recommend for a person, according to its data, is not necessarily the product the brand wants to be showcasing at a given time. Let’s say the latest style of those socks I mentioned earlier just launched and your brand has multichannel marketing devoted to them. The socks can be forced to come up in recommendations by restraining the recommendation engine through business rules to only show latest socks, but this usually does not amount to as much revenue from recommendations as when the engine is allowed to truly recommend products based on the customer’s profile data. If the new socks are purchased by a majority of a certain demo, then they will bubble to the top of the recommendable products pool, which also means people like these products, so it’s more natural/authentic representation of what’s hot.

Additional reading:

I came across a really interesting end of 2013 guide to omnichannel personalization, created by the CMO Group with analysis of best practices in the industry as well as actionable next steps. Read it here.

About eROI

eROI crafts compelling digital experiences across email, web, and social channels. Our work has been consistently successful in driving revenue and exceeding goals for our partners.