What do you get when you put together: Drupal 8 + AI + UX? Drupal8's content management features and integration capabilities, AI, for storing and interpreting data and building a predictive model and UX for anticipating user behavior while adding a “human touch” to the equation? You get predictive UX in Drupal!
Is it possible? Can we implement predictive UX in Drupal and thus create anticipated user experiences that:
- help you deliver meaningful content only
- simplify user choice
- simplify users'... lives?
But how does machine learning actually power these predictive user experiences? What's the whole mechanism behind?
And how is predictive analytics UX any different from... personalization?
Are there any “traps” to be avoided when using the same event data to make informed decisions on the customer's behalf?
And last but not least: what makes Drupal 8 the best fit for predictively serving content?
1. What Is Predictive UX More Precisely?
“Less choice, more automation.”
Or: Anticipating users' needs and delivering them precisely and exclusively the content they need (when they need it). In other words: creating those predictive user experiences that anticipate and meet your customers' needs...
Which one of these 2 possible definitions do you prefer?
Or maybe you'd like a more “elaborate” one:
Predictive UX means leveraging machine learning and statistical analysis to make informed decisions for the customer.
And if we are to turn this definition into a mathematical equation, it would go something like this:
machine learning (predicting) + UX design (anticipating)= predictive UX (based on a predictive or anticipatory design)
2. But Isn't This Just Another Word for “Personalization”?
As compared to personalization, predictive UX goes beyond tailoring content to users' past choices:
It actually makes decisions on their behalf.
It's not limited to leveraging data in order to deliver dynamic content. Which would automatically call for heavy manual work.
Instead, predictive UX is AI-driven, thus automating decision making on the user's behalf.
3. How Does Predictive Analytics Benefit You and Your Customers?
Here's an empathy exercise for you:
You're a mobile app user who's being constantly “flooded” by heavy streams of disruptive information through push notification, by text or by email. Or you're an online customer faced with a discouragingly “beefy” set of options as you're about to order food for lunch. There are so, so many irrelevant options that you're striving to make your way through till you find the dish that really suits your preferences... that you just feel like closing the app and hitting the closest resto instead...
So, what if:
- your app could... tell what you want to have for lunch and display the most relevant options only?
- you would receive app alerts or push notifications in precisely the most appropriate moments (time of the day, of the month)?
- make your life so much easier
- improve your overall user experience
As a company, by leveraging predictive analytics to deliver relevant user experiences only, you're winning your customers' loyalty.
You're simplifying their lives, after all...
4. Leveraging Machine Learning to Create Predictive User Experiences
What's the whole mechanism behind the creation of predictive user experiences?
How is the machine learning technology/tool leveraged to predict user behavior?
It's no more than a 3-step sequence:
- you first define the problem (using machine learning terms)
- gather data in a suitable format
- put together a model
For instance, here's a machine-learning-based recommendation system deconstructed:
- content-based recommendation: recommending items based on similar characteristics
- collaborative filtering: recommending items/services based on other customers' preferences (customers with similar past choices)
Note: more often than not it's a mix of these 2 types of recommendation systems that you'll find.
5. Predictive UX: 4 Common Sense Principles to Consider
5.1. Simplify the UI: keep the most relevant design elements and meaningful content only.
Instead of forcing customers to make too many choices, to scan through chunks of content, go for a minimal interface! Trim down the “irrelevant fat” and keep the essential.
Leveraging machine learning and statistical techniques, you should know by then what's essential and meaningful in terms of information and functionality for your users.
5.2. Disrupt the all-too-familiar patterns now and then.
In other words: don't get trapped in the “experience bubble”, where you keep recommending the same familiar options and encourage the user to make the very same choices over and over again.
Consider adding disruptive layers, now and then, “tempting” them to try something new.
5.3. Avoid forcing those most relevant options on the user.
OK, so you have the data at hand, you're leveraging that machine learning algorithm that anticipates:
- what the user needs
- what the user wants
- what the user's going to do next
That doesn't mean you should overlook that:
It's always the customer who makes the final choice!
So give them enough options to choose from! Put him/her in full control of the final decision-making process!
5.4. Create predictive user experiences that are helpful, not annoying
In other words: when it comes to push notifications, choose the most appropriate time (if you're a retailer, you can't possibly anticipate that anyone would read about your promotion during work hours).
6. Predictive UX in Drupal: What Makes Drupal 8 the Perfect Fit?
There are some particular characteristics that make Drupal the perfect “teammate” for a machine learning tool:
- its content management features and (huge amounts of) data storing and maintaining capabilities
- its API-first approach, which makes third-party integrations conveniently easy; you can integrate Drupal with any system providing an API and an interface
- the “decoupled architecture” approach, which enables Drupal to serve content in various ways
Now, just think about it:
Analyzing that huge volume of data, stored on your Drupal website, and leveraging it, using a machine learning tool, to create anticipated user experiences! Think of all the emerging possibilities of implementing predictive UX in Drupal!
7. And How Do You Implement Predictive UX in Drupal?
First of all: choose your machine learning tool.
Let's say you will have chosen to go with Apache PredictionIO for obvious reasons:
- it's open source
- it “spoils” you with a set of customizable templates
- a full machine learning stack
- the tool's also conveniently easy to deploy as a web service
Now, let's have a close look at the Drupal & machine learning tool interaction:
The Event Server collects data from your Drupal app/website — provides it to the Engine —this one reads it — it uses it to put together a predictive model, by leveraging machine learning — one that it then sends over to your Drupal app/website — upon a query via REST
Et voila! A predictive result is sent to your Drupal website or application, one that will power a predictive user experience.
Now, since we've been talking about the event data that's being sent from Drupal to the machine learning tool and further “exploited” for building that predictive model, you should know that it comes in “2 flavors”:
- explicit: the user will have already rated or bought an item, so you have explicit information about his/her preferences
- implicit: the already available information is being leveraged, since there's no past choice or user feedback to analyze for anticipating his/her needs
The END! What do you... predict:
Will we be witnessing more and more Drupal 8 websites leveraging predictive UX and, implicitly, machine learning technology, to create anticipated user experiences?