How the Retail Personalization Wars Will Be Won

One of the hottest areas in product over the past couple years is “Personalization.” When my title changed to the “Product Manager for Personalization” at my last job, my LinkedIn inbox was flooded with recruiters looking to build out their algorithms and website’s personalization capabilities. However, when I would talk to each recruiter regarding their goals for the role, I always got a different answer. This most commonly manifested as on-site recommendations, email recommendations, and preference-management tools, but sometimes it was about login credential management or even data mining. It seemed like no one really agreed on what personalization really was at its core, even though they all seemed to be on the periphery of it. I think that Wikipedia defines it well (as it often does):

“Personalization, also known as customization, consists of tailoring a service or a product to accommodate specific individuals, as opposed to general groups.”

However, I offer a simpler version: 

Personalization is any way a website makes a user’s experience unique and more useful.

Over the past 5 years we’ve seen a sharp spike in the use of the word “curation” and e-tail companies that have built their entire model around it— often subscription services such as Trunk Club, Peaches & Petals, and Loot Crate. Subscription services, however, aren’t the only ones making big moves in personalization. We see major players like Amazon and Google thriving in the personalization game. Right now you could go to any page on Amazon and you’ll likely see between 6 and 15 exposed recommendation features at a given time. So the question begs to be asked:

If companies with vast resources like Google & Amazon are so serious about personalization, how long is it until they can do it better than the startups (or until they just buy them out)?

It’s a fair question, and with the vast amounts of data that these (and even mid-size) companies are able to analyze and infer from, it’s hard to argue against the idea that it’s an inevitability the big guys will win. In my opinion, though, there is a real “thermal exhaust port” in these Deathstars:


To solve a complex problem the usual response from big tech companies is to throw engineers at it — lots of engineers; and this can be effective…to an extent. See, to offset many of the problems that arise from rapid scaling such as decelerated innovation, increased bureaucracy, and a growing lack of focus these large companies allow teams to organically spin up around specific problems, target markets, or areas of expertise. Usually, though, this comes along with a, “to the victor go the spoils,” mentality — pitting teams against each other to create a superior product (which certainly can yield impressive results). An under-addressed symptom of this, however, is tracking infrastructure — since it changes so often, recording it feels like an exercise in futility. Because of this, it becomes a much bigger deal to make changes to underlying systems, such as data warehouses, analytics platforms, and for retail businesses, their catalogs.

Product catalog & taxonomy at first glance seems like a pretty stable area of the business — I mean, how often are we going to change how we measure pants? Rarely. However, the context around your catalog changes all the time. For instance, think about how often terminology in the fashion industry changes. Or think about the seemingly endless points of view regarding something as simple as a vest: Is it a jacket? Is it an accessory? Is it a top? Is it suiting? Is it casual? The answer could be yes to any or all of these at a given time. Because of this, big businesses are forced to create layers and layers of filtering, mapping and attribution software on top of archaic catalogs and taxonomy structures — after all, it’s much safer to do that than risk breaking the whole website and potentially losing millions of dollars for a change that users won’t even directly see. Even if you wanted to take that leap, it’s difficult to pitch such a risky move to a numbers-driven exec for something as hard to measure as catalog structure. Tracking impact on conversion would be a nightmare, and even if you could find a good way to do it, so many other factors are at play on these massive sites that attribution numbers would almost certainly come with an asterisk.

This is a problem that most of the little guys don’t face. When a retail company is still considered a “startup” or even just a small business, the focus is on building the best possible experience. They have cleaner catalogs, simpler data pipelines, and less internal dependencies to worry about. With smaller catalogs to manage and more focused customer bases providing data, reasons for iteration are often more obvious. Even beyond that, these smaller, more niche sites have a key advantage: they speak the same language as nearly all of their customers. Nastygal speaks Millennial Twenty-Something Fashionista; and they can focus on speaking it fluently because they don’t need to learn other languages. Massive retailers like Nordstrom or Macys have to learn to speak the languages of mothers, grandfathers, high-school girls, ten-year-old boys, athletes, and many more. This usually results in a streamlined yet less informative catalog versus a focused, more insightful one.

So how can big retailers or even your company compete with these personalization ninjas? Spoiler Alert: There is no silver bullet. This is mainly because your ability to address these challenges will be dependent upon your organizational structure and how much freedom is given to employees that don’t have “VP” or “Director” in their title. However, if you happen to work in a company that affords you the slack needed to really address this issue, I have a five suggestions:

  1. Look at Personalization from a user perspective, not from a team perspective. Too often, our need for labels within companies limit our thinking. For example, if you think of Personalization as just recommendations, you will see it as the consumer of the catalog rather than the co-creator. Thus, you probably won’t give them authority to make upstream changes. When you remove this mental barrier you may see opportunities to combine teams that would never have worked together before.
  2. Open up your catalog to your users. In today’s (hash)tag-based social media landscape, we see that users are willing to help provide context to images, products, places, and experiences. Opening up your catalog to users by allowing them to tag & describe products in their own terms not only makes your catalog more robust, but it also helps you speak languages that you haven’t yet mastered. This helps you have a more useful on-site search experience, you’ll rank better in web search, and you may even discover unseen trends in your own customer base.
  3. Spend more time on building out your customer profiles. There seems to be a perception in e-retail that customers are very secretive and resist any attempts to self-describe online. While this may be true for older generations, today’s millennial customers see the value in providing useful info for a better shopping experience (for a great example of this, look at the profiles in Sephora’s online community). Even if some customers don’t want to share their info, that shouldn’t prohibit other users from having a better experience. Imagine how much better you could recommend clothing if a customer uploaded their measurements into their profile and noted their fit preferences. Suddenly, exposing unknown (and perhaps more lucrative) brands in search results is more effective, because users can have confidence that their garment will be flattering on their body. The applications are endless; you just need to allow for it. Besides, now that users can connect social media accounts to fill in many of the blanks on profiles, it’s easier than ever for customers to help themselves get a more personalized experience online.
  4. Add human behavior experts to your data science teams. When creating recommendation algorithms for your products, having team members that are able to understand the context at the micro (behavioral psychologists) and macro (sociologists) levels could make the difference between useful recommendations and wild guesses.
  5. Leverage your online marketing team’s data to inform your knowledge on-site. As a marketer, there’s a treasure trove of behavior data you can ethically access which could lead to the ideal on-site experience for visitors. Many display ad networks and social platforms with advertising allow for cookies, pixels, etc. Often this data is used in a vacuum, kept separate from website/customer data and compared at a high level as a performance baseline. This application is reactive, and not proactive. Imagine the amount of improvement you could make to the user experience if you could connect the dots between even half of your customers and their off-site behavior. Powerful stuff.
    1. *Disclaimer: Data leakage & breaches are to be taken seriously. Don’t get cavalier with linking data if you can’t guarantee that you’ll be the only audience.

If you have basic personalization services on your website and can execute well on even three of those suggestions, you should be in good shape. Got any other suggestions? I’d love to hear them in the comments below!