One in three consumers say they abandon brands because of a lack of personalization.
Statistics like this convey the gravity of personalization — a term that’s often narrowly understood. Personalization is more than displaying a user’s name, it’s more than sending them relevant content.
Personalization can cut deep to the level of personality. In this piece, we’ll explore the latest research on how we can use the Big-5 traits to tailor interfaces.
Beginning at the End: The Five-Step Process of Personalization
Personalization is often made out to be a complex process, and it certainly can be. That said, there’s value in cutting through the noise to get to a simplified set of steps. For personalization using traits, here they are:
- Identify your user’s level of a personality trait either with a survey or through observation of their behavior.
- Pick a design element to manipulate, like the wording of a homepage banner.
- Create at least 2 versions of the design element: One for those high in the trait and the other for those low.
- Present users with the element version that aligns with their personality
- Track engagement metrics like user satisfaction.
In other words, measure users’ personality then mirror it; that way, they find apps easier to use and more fun.
That’s our foundation.
Now let’s step back and build on it as we define key terms.
What Is Personalization?
Personalization concerns individuals, not a crowd. To clarify, it helps to distinguish between two related concepts, personalization and personas.
Personas are fictitious individuals. They feature needs, goals, and demographic characteristics that represent a market segment. While personas are essential for communicating user needs across departments, they have an Achilles’ heel:
We describe a group of users’ average characteristics with personas. Within that group, there’s variability, which leads persona-based designs to miss the mark more often than people realize.
Personalization overcomes this limitation. It’s a set of techniques that adapt designs so they meet the needs of real individuals.
Note that personalization is not just using a person’s name; it’s a spectrum, as illustrated below…
Overall, the spectrum has two categories: conversation characteristics and psychographics.
- Conversation Characteristics: These qualities often appear in small talk, “Hey! I’m Joe and my background is in engineering.” They’re powerful attributes for personalization. For instance, emails that use the recipient’s first name have 40% higher click-through rates.
- Psychographics: These describe what makes a person tick. Two of the most important types of psychographics are emotions and personality traits.
Generally, the farther we go on the spectrum, the more difficult it becomes to engage in personalization. Theoretical problems start to crop-up like: how should our designs relate to a user’s traits?
Competing Approaches for Tailoring Designs
Two schools of thought — heavily influenced by social psychology — compete to dictate the optimal personalization strategy.
First, there’s the complementarity hypothesis. It argues we should match users with page elements at odds with their personality.
As an example, imagine we have a programming whiz with a fatal flaw: they’re low in conscientiousness and carry with them an unremitting sense of apathy. Using our company’s intranet, we could present them with messages that nudge them to strive for greater performance.
Our aim? To use complementary messaging so we can curb aspects of an employee’s personality that stifle their productivity.
On the flip side, there’s the similarity-attraction hypothesis. Here’s the premise: we like people and webpages that share our values, beliefs, and character traits.
Researchers found support for this theory in a strange yet illuminating study. Shockingly, the participants were college students.
(Said no psychologist ever.)
Researchers paired subjects with either a dominant or submissive computer as they solved a problem.
The dominant computer reliably talked first and made commands, not suggestions. The submissive computer, by contrast, talked second and made statements with hesitation.
Can you guess what happened?
When users interacted with a computer that matched their personality, they liked it more, expressed greater satisfaction, and even rated the computer as more competent.
The results of this and many other studies support people’s intuitions: generally, we should create interfaces that are like our users.
At this point, it’s unclear what it means to be ‘like’ a user. There may be dozens, even hundreds, of personal attributes that an interface could mirror.
How, then, can we simplify? How can we go from a messy pile of traits to a set that has the orderliness of a well-maintained library?
The Big-5 Personality Traits
The Big-5 personality traits, also known as the “Five-Factor Model,” is the most widely accepted personality theory. The theory states that personality consists of five core factors, which are:
- Openness to Experience: A person’s tendency to be interested in ideas and aesthetics
- Conscientiousness: The degree to which a person is hard-working and organized
- Extraversion: An individual’s level of sociability, energy, and assertiveness
- Agreeableness: A person’s tendency to be compassionate and polite
- Emotional Stability: This trait describes the volatility and intensity of a person’s emotions
An individual’s level of a Big-5 personality trait is like the face of a mountain. It tends to be relatively stable over a lifetime.
To that point, suppose we measured a group of users’ personality, waited 50 years, then measured their traits again. The correlation between the first measurement and the second would be about .40.1
That’s what’s meant by “relatively stable.”
In contrast, attributes such as emotions change more often. If we use these qualities for personalization then our interface may become too unpredictable. One day a menu has 5 items, then 3, and before you know it, the labels have changed.
That’s a jarring experience. In fact, this unpredictability can make a user’s experience worse than if there was no personalization at all.
A better option is to adapt our interface based on personality traits. That way, we can offer just the right amount of consistency like a pizza bagel that’s not too hot, and not too cold.
What is Personalityzation?
Oded Nov is a leading researcher. He studies human-computer interaction and coined the awkward term “personalityzation.” It’s a cross between personality and personalization.
Definition: Personalityzation refers to the tailoring of an interface so that its compatible with a user’s personality.
Personalityzation has its roots in the interactionist perspective. It’s the idea that a person’s personality interacts with their situation to cause their behavior.
We can readily apply this theory to design. A user’s engagement with an app emerges from how their personality interacts with the interface.
To illustrate this concept, consider an analogy to medicine.
Personalityzation is like dispensing medicine that’s tailored to a patient’s individual genetics. In contrast, showing everyone the same design is like giving patients the same treatment, regardless of their genetic makeup.
Let’s extend the metaphor one step further. Patients who get tailored medicine are more likely to make a full recovery, just like tailored interfaces are more likely to get users full engagement.
In other words, personalityzation should outperform generic designs. The technique accounts for the dynamic interaction between people and interfaces allowing us to create experiences that are much more than ‘meh.’
With that, we have under our belts the theory behind personalityzation. Let’s now get concrete as we discuss 3 thought-provoking experiments from Oded Nov’s lab.
Experiment 1: Displaying an App’s Popularity Isn’t Always Desirable
Oded Nov and colleagues put personalityzation to the test. They ran an experiment with 3 parts using a site that helped graduate students plan their coursework:
- Extraversion measured: Students completed a two-item measure of extraversion on the homepage.
- Banner manipulated: Researchers randomly assigned participants to an interface — one where a banner showed a low number of users who viewed the app, the other a high number (see below).
- Engagement assessed: The researchers tracked whether students left a course evaluation. In another context, we might define engagement as the number of likes, sign-ups, or sales.
They found that extroverts engaged more when the banner suggested the app was popular, while introverts engaged less.
Here’s one perspective on these findings. Popular apps may be appealing to extroverts because they enable receiving a large amount of attention from others. They’re apt to find this deeply rewarding, almost like biting into a chocolate bar of social approval.
Many introverts have virtually the opposite experience. Apps they perceive as having a large number of users are the digital equivalent of a crowded bar. Introverts may be uneasy about engaging in this setting, in part, because of ‘evaluation apprehension,’ a fancy word for fear of being judged.
This account defies a supposed best practice: trumpet an app’s popularity when possible. Rather it suggests going with a more nuanced approach. If your app is popular and your users introverted, consider leaving view counts out of the interface.
Experiment 2: Conscientious Users Provide Reviews If There Aren’t Many
This time around Oded and his team used a clever ruse. They told participants that they were testing PetMatch — a fake service that pretended to pair users to pets based on their personality.
After hitting the app’s landing page, they completed a two-item measure of conscientiousness.
They then received a supposedly tailored “match” for a pet, not realizing that the animal was selected randomly (sneaky, huh?).
Shown above is the page where users received their fake match. Notice the average rating of 3.5 stars? That’s for the quality of the match. Researchers randomly assigned participants to either see a low number of these ratings or a high number (26 versus 2,127).
The dependent variable was whether participants provided feedback on the match’s quality.
Users high in conscientiousness were more likely to leave a review, but only if they saw a low number of ratings. The idea is straightforward. Conscientious people are responsible, so when they see there aren’t many reviews, they lend a hand.
These findings have implications for the cold start problem, the kryptonite of growth.
The dilemma, in our case, works as follows: digital platforms struggle to grow early on because they lack users, but to acquire users they need reviews, which only users can leave.
It’s a classic catch-22.
A potential remedy would be for new companies to systematically recruit conscientious users then ask them for reviews.2 The aspiration is simple: hit the ‘review escape velocity’ and begin growing more rapidly.
Experiment 3: Unstable Users More Prone to Anchoring Effect
Anchors are everywhere, and I’m not talking about pirate ships. Instead, I’m referring to our point of reference when making predictions.
To clarify this concept, let’s consider an example from a memorable study.
Researchers brought participants into a lab and presented them with an anchor using a random number.
It would be one thing if they did this with a mundane computer program. Instead, the researchers used a spinner, think Wheel of Fortune, to generate a completely arbitrary number.
Researchers then asked participants to guess the number of African countries in the European Union. Here’s the kicker: participants systematically skewed their estimate towards their blatantly random anchor.
Put succinctly, arbitrary reference points bias people’s predictions, aka the anchoring effect.
We’re all prone to this effect, but what if some of us more than others? This is exactly what Oded Nov and his team set out to investigate in our third and final study.
They had participants from Amazon Mechanical Turk complete a two-item measure of emotional stability.
Emotional stability captures the intensity and volatility of a person’s emotions. Individuals low in this trait experience life as a roller coaster with thrilling ups and harrowing downs.
In fact, they often feel out of control — a tendency that makes them more prone to social influence.
That’s where the experimental manipulation comes in, again on the phony PetMatch app. Researchers adjusted an anchor: whether users saw low or high average ratings, from other users, for the quality of their pet match.
On average, everyone was prone to the anchoring effect.
No surprises there.
Interestingly, those low in emotional stability were more susceptible to the anchoring effect. They tended to provide ratings closer to the average researchers exposed them to than those high in emotional stability.
What does this mean for UI design?
Using design elements that indicate popularity, unstable individuals can be nudged to take action.
For instance, imagine we’re creating the FitBit app. We could recommend workouts to unstable users by accentuating their popularity (e.g. “Join 4,335 of your peers”).
It’s a small nudge, but at scale, it could have a sizeable impact: 100s of people might engage in self-care that otherwise would not.
Personalityzation 5-Step Process Example
We’ve now covered 3 experiments that serve as proof of concept for personalityzation. That’s a solid foundation. Let’s put into practice what we’ve learned with an example.
Imagine a fictitious streaming service named Pure Fit with on-demand workout videos, including a daily featured class. A problem shared by physical gyms plagues them. About half of users who sign-up for Pure Fit attend virtual classes less than once a week — a dynamic that ultimately leads to greater abandonment, i.e., churn.
With the problem stated, let’s walk through the 5-step personalization process.
- Identify your user’s level of a personality trait either with a survey or through observation of their behavior.
Pure Fit chooses to measure extraversion. Their selection is based on a mixture of empirical research and pragmatic considerations:
- Extraversion has a .27 correlation with the number of exercise sessions a person engages in each week.
- Extraverts are more likely to engage in group workout classes while introverts solitary ones.
- From a more practical standpoint, extraversion is a more socially acceptable trait to ask about than, say, emotional stability.
To acquire a user’s extraversion level, the Pure Fit team administered a 2-item measure of extraversion. They add the survey to an optional set of questions that users can answer when filling out their profile.
Above the items it says, “Get Recommended Classes Just for You.” This approach was inspired by the way Amazon Kindle includes statements like “Upcoming Releases for You” above its recommendations.
The team logs the data from the scale in a database with users’ personal information. Pure Fit either labels the users as high or low in extraversion depending on if their score is above or below the median.
2. Pick a design element to manipulate.
Pure Fit chooses to manipulate their graphic for the featured workout. They base this decision on analytics data, which shows that users who engage with the featured video are less likely to cancel their subscription.
3. Create at least 2 versions of the design element: One for those high in the trait and the other for those low.
The team hires a graphics designer to create two images — one for extraversion and the other low extraversion, aka introversion.
The introversion graphic features a single person engaged in yoga, a more quiescent activity. Looking more closely we can see cool shades such as blue, and the muted color grey — both of which get at the more low-key vibes of introverts. Lastly, the word ‘reflect’ is intended to connect with the contemplative tendencies of introverts.
The extraversion piece, by contrast, makes no reference to the inner life. Instead, it emphasizes group fun both in the copy and imagery. Using the color red hits upon extraverts’ preference for hues that are warm and saturated.
4. Present users with the element version that aligns with their personality.
Pure Fit has to identify users after they sign-in then dynamically render a featured workout graphic that aligns with their personality.
5. Track engagement metrics such as satisfaction.
Pure Fit works with a data scientist to track the results of their personalization.
Here’s what they found…
Users who received a graphic aligned with their personality would tap through 30% more than those who saw a graphic misaligned with their personality.3 Also, they found evidence that the increased click-through rate caused lower churn.4
The team is satisfied with the results, and so are users. Riding off the wave of this success, Pure Fit decides to test personalizing their homepage banner in the future.
Future Directions and Conclusion
So far we’ve looked at studies that center on extraversion, conscientiousness, and emotional stability. Let’s now delve into future studies that could use the last two traits in the Big-5 model, agreeableness and openness.
People high in openness crave novelty, those low, stability. As such, the higher a person is in this trait, the more likely they are to enjoy an unpredictable interface. Let’s apply this nugget of knowledge to YouTube’s UX. They could monitor a user’s behavior to estimate their level of the Big-5 traits then feed that data to an algorithm that recommends videos.
In theory, the user experience should get a lift. Those high in openness get the novel video suggestions they desire, while those low in the trait receive more reliable recommendations.
In this case, we looked at just openness much like how in other studies we only considered one trait at a time. That’s a simplification. In the real world, nobody exists on a single dimension of personality like a character from flatland. Instead, we navigate life with a constellation of traits.
For this reason, designers would benefit from considering the interaction between traits.
For example, conscientious people who are low in agreeableness tend to have worse job performance, especially if the job requires frequent social interaction.
Think about it this way: people low in agreeableness are brusque, and conscientious people work hard. If you combine these two traits, you might be left with a large number of meetings where there’s more bickering than problem-solving.
A potential remedy would be to create a tailored intranet reminder. We could present workers that are both disagreeable and conscientious a message that says: “Remember, it’s company policy to be respectful of your teammates.” This message is tailored to two traits at the same time:
- Conscientiousness: Individuals high in this trait tend to follow rules. That means by invoking a company policy the conscientious user will be more likely to pay attention.
- Agreeableness: Those low in agreeableness often have a blunt style of communicating. The message mirrors that tone by ending in a period, not a smiley emoji.
Hopefully, this little push would cause some employees to give a second thought before leaving insensitive comments. Admittedly, it would be a complex endeavor to put this reminder in effect. After all, we’re talking about a three-way interaction.
But no matter how many traits there are to juggle, no matter how many statistical tests there are to run, the underlying process will be the 5 steps we started with.
Measure a user’s personality, then adapt the interface to align with it. It’s a promising way to create experiences that go beyond displaying a user’s name, beyond accounting for their interests.
The 5-step process can truly delight users.
Top Takeaways
- Personalization exists on a spectrum from name and interest to emotions and personality.
- A user’s character traits interact with design elements to shape how they engage with apps.
- Users vary in how they respond to ‘best practices’ based on their personality.
Endnotes
1 If personality is stable across a 50 year period then there’s a high likelihood that it’ll be stable across, say, the 12 months a user engages with your product.
2 It’s possible to recruit conscientious users through Facebook advertising systematically. The platform doesn’t allow marketers to directly target potential customers based on their level of the Big-5 traits.
They don’t have to though.
Marketers can indirectly reach conscientious users by targeting interests, like accounting, that are conceptually related to the trait (see Matz et al., 2017).
3 Consider a hierarchal logistic regression with user personality, graphic personality, and their interaction as predictors.
4 To test the effect on churn, run a Hayes Process Model using moderated mediation. In this model, graphic personality interacts with user personality in predicting click-through rate, which, in turn, explains churn..
References
Alves, T., Natálio, J., Henriques-Calado, J., & Gama, S. (2020). Incorporating personality in user interface design: A review. Personality and Individual Differences, 155, 109709.
Damian, R. I., Spengler, M., Sutu, A., & Roberts, B. W. (2019). Sixteen going on sixty-six: A longitudinal study of personality stability and change across 50 years. Journal of Personality and Social Psychology, 117(3), 674.
Experian-Marketing-Services-study-finds-personalized-emails-generate-six. (2020). Experian-Marketing-Services-study-finds-personalized-emails-generate-six. Experian Plc. https://www.experianplc.com/media/news/2014/experian-marketing-services-study-finds-personalized-emails-generate-six/
Lee KK. An indirect debiasing method: Priming a target attribute reduces judgmental biases in likelihood estimations. PLoS ONE. 2019;14(3):e0212609. doi:10.1371/journal.pone.0212609
Matz, S. C., Kosinski, M., Nave, G., & Stillwell, D. J. (2017). Psychological targeting as an effective approach to digital mass persuasion. Proceedings of the National Academy of Sciences, 114(48), 12714–12719. https://doi.org/10.1073/pnas.1710966114
Nass, C., Moon, Y., Fogg, B. J., Reeves, B., & Dryer, C. (1995, May). Can computer personalities be human personalities?. In Conference companion on Human factors in computing systems (pp. 228-229).
Nov, O., & Arazy, O. (2013, February). Personality-targeted design: theory, experimental procedure, and preliminary results. In Proceedings of the 2013 conference on Computer supported cooperative work (pp. 977-984).
Nov, O., Arazy, O., López, C., & Brusilovsky, P. (2013, April). Exploring personality-targeted UI design in online social participation systems. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 361-370).
Pazda, A. D., & Thorstenson, C. A. (2018). Extraversion predicts a preference for high-chroma colors. Personality and Individual Differences, 127, 133-138.
Rhodes RE, Smith NE. Personality correlates of physical activity: a review and meta-analysis. Br J Sports Med. 2006;40(12):958-965. doi:10.1136/bjsm.2006.028860
Willott, L. (2019, November 25). Customer Service Stats for 2019. Customer Thermometer. https://www.customerthermometer.com/customer-service/customer-service-and-satisfaction-statistics-for-2020/
Witt, L. A., Burke, L. A., Barrick, M. R., & Mount, M. K. (2002). The interactive effects of conscientiousness and agreeableness on job performance. Journal of Applied Psychology, 87(1), 164–169. https://doi.org/10.1037/0021-9010.87.1.164