Ecommerce Personalization and why you shouldn’t push too hard

It’s easy to get enthusiastic about personalization given the immense UX and marketing power afforded to us by the march of sophisticated automation, complex analytics, and user accounts that span entire digital ecosystems. After all, online marketers love experimenting with new methods — so much goes into selling online that inflexible sellers don’t last very long.

But while ecommerce personalization starts out well — like a warm, welcoming smile — it’s very possible to overdo it. If you get it cataclysmically wrong, it can more closely approximate a stranger walking through your front door, offering you an unnaturally-wide grin, and asking you about the specific details of your last bowel movement. We all know that tech can be creepy!

Even so, when you’re looking to work personalization elements into your designs and systems, you might choose to believe that it’s better to risk overuse than underuse. You can scale back if necessary, right? Maybe, but I recommend always erring on the side of underuse. Here’s why.

The basics of ecommerce personalization

In its simplest incarnation, ecommerce personalization consists of configuring a UX to provide dynamic elements determined by the history and settings of the customer using it. Whether in a marketing email or on a store page, information is drawn from the account associated with the user (either attached to the email address or gleaned from the current store login).

The simplest example: you visit an ecommerce store, provide your login details, and proceed to a version of the homepage that greets you using your first name. What does that accomplish? Well, not all that much, but it does manage two things: avoiding clunky and sterile constructions like “Hi customer” or “Hello visitor”, and subtly demonstrating that you are known to the business. You’re not some anonymous visitor to them.

Back when you’d actually visit a brick-and-mortar store, being on a first-name basis was a sign that you had established a good rapport with the staff, usually suggesting that they’d be better equipped to understand your needs as a customer. Online, it doesn’t hold that meaning, but it does allude to it — and in doing so, it carries over a sliver of the value.

Expanding the value for the customer

Because the objective of UX design work is to make a system work better for a customer, that kind of shallow personalization doesn’t go too far. The real meat of ecommerce personalization lies in catered offers, services, and product recommendations — things that outright make a site or email meaningfully and practically better. Think of that in-store employee knowing your needs as a customer: using algorithms and analytics data, online stores don’t need actually know someone to accurately determine what they value and how they spend their money.

A promotional email pushed out to previous customers doesn’t need to have a fixed set of products listed. It can dynamically base the array on the order history and expressed preferences of the email recipient, so the same email template sent to five different people can have five different product sets.

From the perspective of the customer, this is highly useful. If you know that there’s a higher likelihood that you’ll see something you’re interested in if you open a marketing email, the experience will be more valuable to you, making you less likely to ignore it as you would most emails (we all have inboxes packed with spam, after all).

Retailers can also go down the route of personalizing products as a method of building customer loyalty — it’s nice to receive a free gift as a reward for being a long-time customer, but it’s even nicer to receive something personalized. It’s easy enough for a seller to test out a print on demand store and apply low-cost custom tweaks to common items (a T-shirt with the customer’s name on it, a notepad in their selected background color, etc.). The customer gets something unique to them, and is made to feel that their custom is truly appreciated.

How you can push personalization too far

The problems with personalization manifest when you do one of the following two things:

Feign familiarity

We all know that personalization of the “including personal details” variety is artificial — there’s no familiar person on the other end responsible for carefully expanding site content with your information. Now, “artificial” doesn’t mean “bad”, and we’ll gladly tolerate some degree of artificiality because tasteful use doesn’t generate any discomfort, but it can go too far.

Perhaps the best comparison in the technology world is the uncanny valley in robotics and 3d modelling. Give us a robot that looks humanoid and we’ll be fine with it, but try too hard to make a robot look human and our minds will view it as an attempt at deceit. We don’t mind contrived language when we use chatbots because we know what they are to begin with — but if we open up “live chat” systems that appear to use bots with no disclosure, we’ll feel misled.

Think about getting an email from a store containing the following sort of construction: “Hey there [your name]! So I saw the other day that you bought [recent purchase] — awesome choice! You clearly have great taste, and that made me think that maybe you’d also be interested in [similar item]. I know you’re a busy person, so I wouldn’t waste your time with anything mediocre. Why not take a look? Anyway, keep being awesome!“.

Why is that so objectionable? Well, it tries far too hard to pretend that there’s someone on the other end who knows you and understands your tastes, and attempts to conceal the sales-focussed intent using the sort of tone you’d use when talking to a friend. That kind of familiarity needs to be earned over time by actual people.

Appear invasive

We know how retail works, and we understand that value in the online world is reciprocal — given primarily so it can be returned. Most of the time, we’re absolutely fine with that. We willingly give up anonymity in order to have more convenient purchasing experiences, and don’t feel exploited when our freely-relinquished data is used to market to us — but businesses that fly past our expectations with complex processing can damage that exchange.

How does that happen? It’s a little like sleeping on a friend’s sofa (go with me on this). If you’re a respectful houseguest and you don’t get in the way, they’ll probably be happy for you to stay a while. They may even forget to ask you to leave. But if you start eating all their oatmeal after a couple of months, they’ll feel that you’ve broken the implicit terms of the deal, and show you the door. They can no longer trust you to be responsible with your belongings.

For a real-world example of this kind of behavior, I always go back to the classic case of Target’s customer tracking technology analyzing a teenager’s shopping activity, discerning that she was pregnant, and sending coupons for baby items to her house — the problem being that she lived with her parents and they didn’t know about the pregnancy. You can understand the father and daughter alike being unhappy with that situation!

You may be able to infer a great many things from a customer’s data, of course. Perhaps you can conclude that it’s very likely, based on their previous buys, that they suffer from weight problems. But if you start adding weight loss products to their recommendations when they’ve never expressed interest in such products, they’ll be deeply unhappy, and rightfully so. You might rightfully note “Well, that’s a sensitive topic”, but any topic can be sensitive to someone.

I’m firmly of the belief that the customer must always feel in control of the personalization, and it should never stray too far from the parameters established by their purchasing history. I don’t care if your algorithms can glean that I’m an ambitious entrepreneur — unless I’ve specifically told you to take that leap (more on this below), I do not want you to send me business book recommendations unprompted.

The double-edged sword of automation

When we use complex personalization, we inevitably rely heavily on automation to achieve it, because there isn’t enough time in the day for anyone to manually cater their content to countless niches while taking all the variables into account. And that’s a good thing, by and large, but it’s fundamentally a double-edged sword — and you’ll understand why if you’ve ever delegated responsibility to an employee only to see them get it all horribly wrong. You don’t have to do the work yourself, which is good, but you lose all control, which is bad.

Because modern machine learning is so remarkably fast, we can make the mistake of thinking it’s actually smart in the way that a person can be, but it isn’t. In truth, much of the digital world runs on precarious stacks of what Andrew Smith recently called franken-algorithms: myriad chunks of code that are expected to interact seamlessly and coherently, but operate too mysteriously at a deep level to be very trustworthy.

As marketers, it’s important to remember that automation tools are just that: tools. They’re not wholesale replacements for customer service teams, and you must keep your use of them within reason and not be so quick to cede responsibility — that’s how the robots eventually take over.

The importance of user direction

So, we’ve seen that ecommerce personalization scales from the basic use of a first name to the extrapolation of boundless likelihoods from digital analytics. You can say “Hello Joe”, or you can use a supercomputer to calculate the entire range of items someone is most likely to buy and configure marketing to push them at the optimal times. And no extent of personalization is inherently damaging — it’s all about context.

What this ultimately comes down to is user awareness and determination. The terms of your personalization can be whatever you want them to be as long as you get sign-off from the user. In ideal circumstances, you could give each new customer a fully-fledged set of personalization options, allowing them to shape their experience for themselves.

Imagine becoming a customer of store and being presented with the opportunity to do everything from customizing the style of the store (changing the background, choosing the colors, adjusting the font sizing, etc.) to identifying the specific things they want done with their data.

Do they want to be recommended products similar to those they’ve bought? Choose a variety of interests and get suggestions in those areas? Sync up their social media accounts and let the world of big processing figure out their optimal cart contents?

Conclusion

In essence, pushing too hard is just pushing harder than the user is comfortable with, and it will rapidly drive people away. If you want to massively scale your ecommerce personalization, make certain that your customers are up for it first — and when you don’t know exactly what your customers will tolerate, err on the side of caution!