Web personalization is a marketing and optimization technique that is often misunderstood.
On one side, you have those who look at it as a silver bullet, something you can simply implement everywhere and use to start raking in more money.
And on the other side?
The other side views web personalization as impossible — something you can’t even approach or try because of the perceived complexity.
As with most things, the truth lies somewhere in the middle. Personalization is hard. Getting it right is even harder. But there’s nothing magical about it.
The reality is most businesses can take advantage of web personalization techniques to provide a better user experience and to acquire more users.
Defining Web Personalization
First off, what is web personalization?
Web personalization is the targeted delivery of an experience due to some specific set of criteria.
It may sound like a simple definition, but if we can frame things correctly from the get-go, everything else will be easier.
Of course, it’s easier to imagine with an example.
To illustrate the concept, consider the homepage for a blogging tool called Wordable. The product lets you upload Google Docs drafts to WordPress in a couple clicks.
The site’s owner wants to target return visitors with a separate message from new visitors. As they can detect return visitors (gather the appropriate data) and deliver the content in real time (provide an experience delivery), this is possible. Look at how Wordable can target returning guests with a “welcome back” message:
Now imagine they also want to break things down between current customers and prospective customers. They could, if they wanted to, deliver a more targeted experience to users who already use the product:
Of course, web personalization doesn’t necessarily mean content personalization. You can personalize fundamental user experience elements on a site, such as product recommendation engines in E-commerce. Amazon is the king of this technique:
We can personalize email subject lines, ad creative, product interfaces, landing page headlines, marketing automation, and even product functionality.
We can also target personalization using many different criteria, including:
- Demographics (e.g. 50+ year old males who make $100,000 or more per year)
- Psychographics (e.g. people who like action movies)
- User Behavior (e.g. those who scroll 50% down a page)
- Technology (e.g. using Chrome on an Apple iPhone)
Here’s the thing about website personalization: you can do it in an infinite amount of ways. I could keep giving examples all day, but that would only serve to emphasize the tactic. In reality, we should approach personalization like any other product or marketing investment, with an eye towards ROI.
“The most fundamental thing people forget about “personalization” is that I can “personalize” an experience in almost infinite ways. I can change copy, I can change workflow, I can change layout or features of the experience. Even better, I can do this for the same user in a thousand different ways. I am a returning user to your site… but I am also a user in the afternoon, who came from Google, who has been on the site 12 times, who has made 3 purchases, and who is using FireFox. So the question is not CAN I personalize an experience, at this point there are a thousand different tools and ways to do so. So the simple act of creating an experience is not the goal, the goal is to do so in the way that generates the greatest ROI for my organization. The question needs to be, how do I discover the most valuable way to change the experience?”
But how exactly do you discover the most valuable way to change the experience for visitors?
How to Get Started with Personalization (What You Need to Begin)
The main point with website personalization you should keep in mind is this: just because you can personalize an experience doesn’t mean you should personalize an experience.
Personalization should be subject to the same decision criteria and rigor as A/B testing or any design decision.
I’ve seen countless articles calling personalization “a better alternative to A/B testing,” which it simply isn’t. You don’t get to skimp out on the math and decision criteria because you call it “personalization.”
“You can frame personalization as just another arm in a test. Here the hypothesis is that there is a marginal return to a many to many mapping of customers to experiences, vs a many to one mapping or a random assignment, depending. By default, we pipe in the results of any predictive, or fixed targeting into this type of AB testing.”
So there’s nothing magic here. Rather, web personalization is just a dedicated experience delivered to a subset of the population with the belief that it will result in a worthwhile ROI.
To make personalization work, we need a few technological and resource elements in place. According to Matt Gershoff, these elements are:
- Customer Data
- User Experience (Content)
- Targeting Rules
You’re likely collecting data through one method or another. If you’re not, start there. Ask yourself, which user behaviors may be important to know about, and is it possible to measure those behaviors?
Customer data that may be important for personalization includes:
- Device type
- Scrolling/clicking/browsing behavior
- Session count
- Time on site
- Time of day
- The weather
But that’s just the beginning. There are a countless number of data points we can capture with analytics solutions nowadays.
User Experience (Personalized Content)
Content is fairly straightforward. Given a population of users, which creative or content or experience is the most suitable to deliver them? Here, you enter the land of copywriting, design, advertising, engineering, etc.
Content is the actual execution you see with personalization, such as copy that looks like this:
Finally, the unique piece of personalization: targeting. Once we have the data, and we have the resources to manage the creative, under what rules do we decide to target a subset of users with a given experience?
Generally, there are two ways to go about this:
- Business Logic
- Machine Logic
Business logic is somewhat straightforward, and it’s probably what you’re used to in regards to web personalization. Here are two business logic web personalization examples:
You can collect the IP information of visitors and determine what company they work for, so you display a unique message to visitors with that IP:
Or let’s say know that a person has already signed up for an e-book in the past, so you preclude the need for them to fill out form fields that they’ve already completed and instead ask them more interesting questions. In other words, you use progressive profiling, something that you can accomplish out of the box with many form builder tools:
Whatever the context, in using business logic, you set the targeting rules using your own reasoning and rules to do so.
Machine logic is different.
In this path, you rely on predictive targeting to find and/or exploit the segments for you. Here’s how Reid Bryant put in on the Brooks Bell blog:
“A better approach is more algorithmic: an analyst uses unsupervised learning techniques, such as clustering or latent class analysis, to determine the number and definitions of each group. The algorithm will only partially answer the question—the business will need to determine if it’s profitable to actually direct distinct campaigns to each group.”
To get started, it may be easier to begin by crawling rather than running. Business logic can help you pick off the biggest low hanging fruit. It doesn’t have to be arbitrary, either.
In personalization, as in other areas of marketing, it helps to have a framework or a process by which you operate. I’ll outline one of my favorite straightforward processes here.
A Simple Framework for Starting Out with Website Personalization
- Define your goals and KPIs
- Do both quantitative and qualitative research
- Identify, select, and prioritize segments to target
- Prioritize your segments and experiences
- Design and orchestrate experiments
- Measure and iterate
1. Define Your Goals
All great plans start with clearly defined goals and KPIs. What do you want to happen? Where do you want to move the needle?
In practice, this goal setting looks quite similar to what you’d do with any type of conversion rate optimization. You build a growth model, look for areas of high impact, and map out what results you expect by working on key parts of your user journey.
I’m glossing over this step a bit, but largely because it’s been written about a ton in relation to startups and specifically for marketing goals. Essentially, what you want to do is audit your current growth marketing funnel and see where you can best apply your efforts — whether that is to acquisition, monetization, activation, or another goal.
2. Do Qualitative and Quantitative Research
User research isn’t only for A/B testing. With testing, you should know which problems you’re trying to solve before you dive into them. With web personalization, it’s no different.
Now, this step may be longer or shorter depending on your context. If you’re an agency or consultant coming in to work with a new client, you probably need to ramp up on user insights. Therefore, your research process will be more comprehensive.
If you’ve been working at your company for a while, you probably know a good bit about what problems your users are having and where the bottlenecks are located. Still, a bit of user research can probably help you in most cases.
This, too, is a rather big topic, so I’ll just outline a few good research methods I’ve found to be helpful:
- Digital analytics analysis
- User surveys
- On-site polls
- User testing
- Heat maps
- Session replays
And here’s a great article covering conversion research.
3. Identify and Select Segments to Target
If you conduct user research properly, you’ll walk away with tons of ideas on which segments you can target and how you can do so.
As for these segments, some you can look into include:
- Demographic segments
- Acquisition segments
- Contextual segments
- Behavioral segments
- Historical segments
- Algorithmic segments
Demographics segments include variables such as gender, age, and income.
Acquisition segments are all about where your visitors are coming from — browsers, devices, channels, firmographic data, etc.
Contextual segments may include geographic data such as city or country. Further, you can take into account the time of day or the weather to further personalize the experience for visitors.
Behavioral segments are my favorite to think about. Here, you’re looking at the actual action your visitors or customers display. Think about activation metrics like Facebook’s “seven friends in seven days,” or even simpler, something like a visitor scrolling to the 50% down a blog post to trigger a scroll box:
Behavioral cohorts can also refer to new vs. returning visitors or even how much users typically spend on your site.
In reality, the classes and types of segments could go on and on for days, and we could argue about where they’re delineating, but think about this:
What’s a meaningful segment to target with a dedicated experience?
Obviously, some could surprise you in their importance (which is why predictive targeting is so great). But some segments are clearer than others in their pliability. For instance, whether a visitor is returning or new is probably important. Similarly, if a visitor is on mobile or desktop is an important distinction.
In addition, these two segments are large. If you aim too small — for example, Russian mobile visitors who have more than 6 pageviews per session — you’re probably not going to get much ROI on building a dedicated experience for them, regardless of how well it would perform.
4. Prioritize Your Segments and Experiences
Next, we need some way to prioritize these segments and experiences. Otherwise, we’d just have a chaotic and unmanageable list of ideas.
I’m always a fan of balancing two variables: impact and ease.
If implementing a targeting rule affects very few people, the impact is low. Put it at the bottom of the list. If the idea will take months to set up, the ease score is low. Similarly, put it at the bottom of the list.
The best ideas you can have are high impact and high ease. These should be shuffled towards the top, as you can test them quickly and move on.
5. Design and Orchestrate Experiments
Finally, we can test out our website personalization rules!
Here’s the important part: you don’t get to skimp out on the statistics you’d use for A/B testing. You still need to validate decisions empirically before moving on and declaring victory to your team.
Testing is a huge topic, obviously. Here’s a huge article I wrote on A/B testing years ago that is pretty helpful.
So, you’ll want to run a test to its course, and if it wins, launch it to the segment in question.
6. Measure and Iterate
Finally, you’ll need to manage your targeting rules. The more rules you have, the more complex your organizational capacities will need to be.
Think about it: ideas are perishable. They don’t last forever. What works now may not work in 6 months. If you run A/B tests, you know that for a fact. Matt Gershoff riffed on the idea of perishability here:
“I think the world has drift. We’re in a world in which things change over time. Our solutions are perishable in many situations. I think that’s another key thing to think about – the perishability of the problem. It’s something you should ask before you start trying to solve a problem. If it isn’t very perishable, that’s a situation where an A/B test approach is going to be effective.It could be that, for movies, they’ve seen lots of films, new ones are coming out – let’s say you’re Netflix and Amazon has its own products – there are lots of things outside of your control in the real world. Your users could start to look different over time. Just in general, the world changes from underneath you.”
So what do you do about that?
Well, you can’t remain static. Optimization never ends. You’ve got to keep testing and improving and keep your eye on the ball.
That’s another reason you don’t want to just go and set up a thousand targeting rules: you have to manage them and continually be assured they’re actually optimal.
Take Two: A Discipline-Based Targeting Approach
The process above that I outlined (from Ismaël Sow’s CXL article) isn’t the only way to do things.
In fact, one of my favorite ways of finding “exploitable segments” is through the standard process of A/B testing. However, after analyzing your A/B test, you should dig further into key segments to see if there were any wildly varying behaviors that are worth focusing more time on.
This, in essence, is the approach Andrew Anderson, Director of Growth at American Addiction Centers (and one of the smartest minds in optimization) recommends with regard to targeting. It’s a process of discovery, weighing the ROI and feasibility, and then exploiting through targeting.
- Create multiple executions of the message or experience
- Serve all the offers to everyone
- Look at the results by segment and calculate the total gain by giving a differentiated experience.
- Push live the highest revenue producing opportunity found
Basically, run a test with several variants and determine, based on the behavior of individual segments, whether it is worth creating individual targeting rules. If you deem it worthwhile from an ROI perspective, then you can create a test for a targeting rule (make sure you still do a controlled experiment using proper statistical methods). If that wins, then you can set it live to the full segment.
As Andrew says, this is fundamentally different than those who fail at website personalization, which looks like this usually:
- Push the single piece of creative to the repeat purchaser segment
Andrew has a great article on how exactly he calculates the ROI of implementing a given personalization rule if you want to read up further on the nitty-gritty.
Personalization is hard. It’s not a magic bullet. But it’s not impossible, and it can help any company achieve its conversion or user experience goals.
The main point is this: chase ROI and good decision making, just as you would with any other marketing method.
In addition, it helps to have a process. I outlined two possible processes here. Clearly, there are other ways to do things. I haven’t even started to cover predictive targeting (a whole subject in its own right).
To get you started, take one of these frameworks and get running.