The design process is often treated as an art, and intuition usually is the way to go. Unfortunately, designers can’t read users’ minds. That’s why this approach may lead to a design that is out of alignment with the needs of a user. This is where the data-driven design may help.
This approach of data-driven design helps to create a user-centric design and a better user experience. It enables you to make better design choices based on real evidence about the user’s behavior, attitude, needs, etc.
However, there’s still a lot of confusion about data-driven design and a lack of understanding of why it is important.
To clear things out, we prepared this complete guide to the data-driven design process. This guide will cover what data-driven design is, why it’s important, what counts as data, how to use it, how to get support from stakeholders, and the first steps for implementing the data-driven design process.
Table of Contents
PART 1: Explanation of a data-driven design concept
What is a data-driven design?
Data-driven design can be defined as a decision-making approach to the design process that heavily relies on collected data about customers’ behavior and attitude.
Information about how customers interact with your design acts as feedback that informs you whether your design fulfills its purpose.
In the case of a landing page, is the CTA button visible enough, and does it get enough clicks? Does design capture the attention without overshadowing the main message? Are all the steps in the buying process clear to the user if it’s an e-commerce page?
What counts as data?
When people hear the word “data,” they almost instantaneously think about quantitative data, which comes in numbers.
Data is not only numbers. Qualitative data, which refers to things like feelings, opinions, and observations that can’t be expressed in numerical value, is also data.
You can use both quantitative and qualitative data to inform the design process.
Quantitative data comes in numbers and answers questions on how many, how much, how often. A/B or multivariate testing, website analytics, heatmaps from eye tracking studies, large-sample surveys – all of these are examples of quantitative data sources.
Qualitative data, on the other hand, focuses on the “why.” It gives insights on user motivation and intent. You can acquire qualitative data via interviews, competitor analysis, usability studies, focus groups, diary studies.
Both types of data are valuable because they supplement each other.
Why should you care about data-driven design?
Trusting only your gut and not tapping into the data for real-life feedback can be a dangerous approach. It may lead to ineffective design, which in turn can result in lost revenue, wasted time and effort for redesigning, or even some harm to the brand image.
Effective use of data can increase conversions and drive your business to overall success. There are quite a few success stories on how data-driven UX methods significantly contribute to the growth of a business.
For example, a customized workplace solution provider Continental Office needed a new website and a rebranding. They wanted to integrate buyer personas to provide an engaging user experience with relevant content marketing.
The results of their redesign based on data about their customers:
- 103% increase in traffic year-over-year
- 645% increase in net-new contacts
However, such major changes are not always necessary to reach comparable improvements.
Data is there to help you.
Designers often don’t like this data-driven design concept because they fear that their creativity will be limited or even replaced by data and numbers. Not to mention the false assumption that they would need to deal with numbers. That is far from the truth.
Data acts like a tool that helps the designer in the process of creating the best user experience. Designers can validate their instinctive choices with evidence. Also, they can better understand the customers’ needs and motivations from qualitative data and adjust the design accordingly.
There’s no need for designers to start crunching numbers and learning statistical analysis. They would still be focused on creative work but in cooperation with researchers and data scientists who would provide beneficial feedback backed by data.
All data can be presented in a simple way and even visualized to make it more clear. So, mathematical literacy is really not necessary for designers.
Also, data-driven design techniques can save time and resources. Proper user research and some testing can decrease the number of iterations needed to get the final design, or you may even get it right from the first time.
It’s the designer’s dream – less dealing with revisions and more time available for creative work.
Data is not everything
But there’s a dark side to this data-driven design process.
Don’t fall into the trap of trying to optimize the numbers while forgetting a broader picture. It may lead to a worse user experience and damage to a brand image in the long term.
You need to maintain a healthy balance between your intuition and empirical evidence. It’s a subtle art of knowing when to rely on which. Basically, when the data doesn’t give a very clear answer, or you need to stitch many different pieces in one harmonious ensemble, you can trust your expertise and gut feeling.
Be informed by numbers but not a slave to them.
After all, data can inform, but the designer needs to add that secret ingredient and bring the design to life. This human factor is what drives real innovation, while numbers can only inspire or give some useful insights.
PART 2: How to get stakeholders’ support for data-driven design?
Facing resistance from above
So far, we have explained what data-driven design is and what value it can create. Also, we addressed some misconceptions about it.
The data-driven design process is backed by evidence about the users, which is the central pillar in creating a user-centric design. As we have shown, this customer-oriented and data-based approach to design can create significant value for your business.
Despite that, the data-driven design is often misunderstood, and its implementation may create a lot of friction in a company. You may face some resistance from the stakeholders who may have a different opinion. Suppose the designer can create a decent design all by himself without any data, and revenue is flowing. Why bother with all the research, analysis, data interpretation, and additional people for all of that?
Thus, it can be challenging to get stakeholders on board with this concept of data-driven design. However, if you apply our tips, this task might get a lot easier.
Explain the basics and provide context
First of all, if you want to get stakeholders’ support for your vision of a data-driven design approach, you need to make sure that they clearly understand what a data-driven approach means. One of the essential points to address is what counts as “data.”
Usually, when most people think about data, they think about numbers. It is especially true if they are not too familiar with UX research. But some information you can’t express in numbers. Sometimes it comes as opinions, observations, or feelings.
You can think of data more like evidence. A good chunk of it comes as numerical data and gives answers on What, When, Where, and How often. Although this quantitative data is highly valuable, it can’t answer why people behave that way.
We need to tap into qualitative research to get insights into the motivations of the user. By observing users, listening to their opinions, and empathizing with them, we can bring to the table that missing piece of information, and that can be a game-changer.
That’s why we should treat this less tangible and less structured information with the same respect as numerical data. Take time to make sure that you and stakeholders are on the same page about this.
Also, you can’t treat data as unquestionable truth. Even though data provides us evidence about customer behavior, we need to remember that this is just a partial representation of their experience. No matter how big your data is, we are talking about one or a few aspects of the whole story. And to see the complete view, we need to broaden and diversify data. Consider including as many different data sources as you can, and even then, keep in mind that this is an approximation of user experience rather than the ultimate truth.
Having this conversation with stakeholders about what we can and can’t accomplish with data is necessary for the successful implementation of a data-driven design approach.
Show data-driven design value for business
While many companies still don’t see the design as a priority, facts suggest that they should. Take a look at the DMI Design Value Index that includes a list of carefully selected design-led publicly traded US companies. As the graph below shows, companies who make design a priority over the ten years outperformed S&P index companies in the stock market by an outstanding 228%.
If a design can make such a difference, make sure to back it with data, not just guesswork and intuition. Research supports this notion that a data-driven approach can give you a business advantage.
As Andrew McAfee and Erik Brynjolfsson from MIT stated in Harvard Business Review:
Moreover, data-driven design principles can help increase the productivity of designers. User research and some testing can decrease the number of iterations and revisions needed to finish the design. So, a designer can do more in the same amount of time.
It’s quite clear that by understanding the importance of design and integrating data-driven philosophy as the core principle of your company’s work, there’s a lot to gain in the long term.
Show success cases of data-driven design
When you try to convince stakeholders to adopt the data-driven design, you may want to use success cases to demonstrate the return of investment (ROI) value of data-driven UX research methods. It’s easier to sell the idea when you show how others benefited from evidence-based practices.
Let the facts speak for themselves. It’s hard to be ignorant about data-driven design when you see what value it can create.
We can apply this idea when talking to the stakeholders.
People tend to digest information better when you present it in a few different ways. A 2016 study shows that having visuals when introducing new information helps people to absorb information.
As researchers conclude in that study:
So, consider including some visualizations and diagrams when preparing your presentation for stakeholders. A few convincing graphs can go a long way.
PART 3: Getting started with data-driven design process
Confusion before the start
In the two previous sections, we have covered quite a lot.
We explained what data-driven design is and why it is essential in today’s business world, and we gave you some tips on how to get stakeholders’ support for implementing it in your company.
However, we haven’t touched on the topic of getting started with this data-driven design process yet.
Introducing a data-driven design model to your existing working framework might seem like a daunting task. With so many “moving parts,” it’s even hard to get this process started. This section will give you a better understanding of where you begin with this data-based approach and vital elements for implementing this model in your workplace.
Ensure access to data
First and foremost, key people will need to access the data for this data-driven design process to work. There’s often a limited information flow between the departments in larger organizations, which can be a real drag in implementing this new process.
It’s quite an outdated view that analytics specialists deal with quantitative data, and UX researchers and designers take care of the experience. Little by little, these boundaries are blurring.
Designers also need access to the quantitative data, especially that which characterizes user behavior. Of course, they need this data represented in an understandable way, not just raw data, which can look like a foreign language to non-data people.
Discuss with your team and decide how you will share this information in your organization between the data people and designers. Consider who needs to know what, how you will structure the whole process, and what tools you will use. Make sure that acquiring all the necessary information is as easy as it can get to everyone.
Get on the same page about data-driven design
Ensuring effective information flow is not enough. Colleagues also need to understand each other. That takes us to another challenge you’re likely to face from the very start – no common language between the data and design departments. Designers do not necessarily need to learn all the technical slang that data people use, but they at least need to understand each other.
An excellent place to start is to define the basics. Touch the subject of quantitative and qualitative data by explaining how they relate and why it’s crucial to use both of those data types in the design process. We’ve already covered this in part 2.
Early on, clear out misunderstandings about this quantitative/qualitative issue to avoid unnecessary conflict related to doubts about qualitative data’s credibility and to improve communication between team members. Also, define the terminology that you will use while communicating or in some data extraction tools.
It’s also essential to have a clear understanding of what success looks like and make sure everybody in the team agrees on it. There might be many different goals that you might aim for, but you can achieve everything at once. Double-check whether everyone who’s involved in a particular task has the same goal in mind. That leads us to the process of defining goals.
Make your goals clear and realistic
Before you start gathering the data, you need to know what you want to accomplish with it. Making your goals as clear as possible is necessary for implementing this data-centric approach effectively.
Do you want to introduce a completely new product or just modify an existing one and create a new iteration of it? The data collecting process will be different in these two scenarios.
Also, make sure that your goals are realistic. Preferably, you want data-driven techniques applied right from the beginning. That’s why it may be tempting to remake the product from scratch using data-driven design principles. Next to the data, other important factors should influence your decisions, such as cost, time, and feasibility.
You can consider these additional variables and ask yourself whether a redesign is a practical choice in that concrete situation? Sometimes you might be better off just modifying the product.
Create a hypothesis
Once you know why you are gathering the data, then you can focus on creating a hypothesis.
Hypotheses are used in science experiments, but the same principles apply when you create a UX experiment.
Let’s look at the definition of hypothesis, provided by the biology department at California State University, Bakersfield:
Do not confuse a hypothesis with a theory. A hypothesis is more like a predictive assumption, while a theory is a body of knowledge backed by data that explains phenomena.
An example of a formalized hypothesis would be:
Notice that this statement has two parts: a proposition of testable relationship and a prediction of expected results. If A relationship is true, then B will happen. In this way, the hypothesis assumes a cause-and-effect relationship between the independent variable and the dependent variable. The same principles apply when you create a hypothesis for UX research.
Here’s an example of a formalized hypothesis of a UX experiment:
A hypothesis like this would be a great starting point for a UX experiment. You could A/B test a few designs – one with a high contrasting CTA button and another with less contrasting CTA. Numbers will show you the truth in this case.
Choosing a data-driven design strategy
Since you have guidelines for creating a hypothesis, the critical question is, how do you choose what to test? Where do you start the research?
Testing everything that can be tested is not a good idea. Also, avoid stabbing in the dark, hoping you will hit the target. In both cases, the chances are that the results won’t suggest any impactful changes. What’s worse, you may lose credibility in the decision-makers’ eyes, and it might be hard to get a green light for another test.
The good news is that you already have some data about the behavior of your customers. Either you do some simple user testing or have the data from analytics. If you’re using this data to make decisions, it can be called a simplified version of a data-driven approach.
This is a good start, and it will make it a lot easier to introduce a more sophisticated data-driven design process into your decision making. You can start applying data-driven techniques with the data that you have. After you got used to it and exploited this data available at your fingers, you can start thinking about other data collecting methods.
Get to know your customers using data
There are a few approaches you could try. The first approach is excellent if you’re not confident about your ideal customer profile (ICP). You can use data to get to know your customers better. Check page analytics and analyze behavior flow to understand what they are doing on your page. Analyze demographic data and audience analytics to get a more detailed view of your clients.
Using this data, you can start looking for confirmation or disproof for your guesses in the real world with real persons. Get more information on what people use your product, how they use it, and why via customer surveys and interviews. Analyze this new information and create your ideal user personas using recurring themes and patterns that you uncovered from this user research.
Now that you have user personas, you can run tests with users matching your ICPs. Get feedback from them in the early development stage. Involve these users in beta testing so you could find as many bugs and other issues as possible before the launch. As you develop your product or website, get back to these users to get valuable feedback.
Search for anomalies in data
The second approach is great if you have a good grasp on your ICP, but something else seems off. By analyzing quantitative data from web analytics, you can spot some anomalies in user behavior. Those strange patterns of your customers can manifest as some metric which is unusually high or low. It can be an immensely high bounce rate, very short average dwell time, or higher than average exit percentage on some subpages. It’s not easy to figure out what those indications mean because there may be many different reasons.
Suppose you have a SaaS platform or an app that you sell to customers. Your landing page leads users to the trial registration form. While examining analytic data, you observe that conversion from traffic to trial users is quite good. Still, you discover that just a few hours after registering, most of them stop using your platform and become inactive users. What’s happening?
There might be various scenarios. Maybe some bugs annoy the users too much, or inconvenient user interface just finally gets at them. Maybe they get lost in the tutorial or your animated tutorial freezes at some time? Or is it the problem with communication and your customers expect one thing, but when they register, they do not get what they wanted exactly. Quantitative data will get you this far.
Numbers can inform you of what is happening, but they won’t tell you why something is happening.
Now you need to dig deeper with qualitative data. Get in touch with your customers and look from their perspective. Gather more data using surveys or user interviews or get session recordings.
What’s even better, it’s observing a person matching your ideal customer persona using your product in real-time and asking him some questions about what he thinks while in action. It may be expensive but might result in some valuable insights. When you pinpoint the problem, you can develop a few potential solutions and run an A/B test to figure out which of them solves the problem better.
Get good at data
If you’re learning how to harness the power of the data, make sure you do it correctly and avoid mistakes that could lead you to lower quality data or even incorrect conclusions.
Here are a few guidelines to follow if you want to make the most of your data:
- Gather enough data. Make sure that the data sample from which you draw conclusions is big enough so that your results are statistically significant. If you collect too little data, it may show a skewed view of reality. You won’t get much value out of such data. Also, try gathering data over extended periods to minimize the influence of transient effects and random fluctuations.
- Use reference points for your data. Suppose you’re checking your page analytics, and you see that 500 users visited your site in a day. Is it a good or bad result? You wouldn’t know until you compare it to your weekly or monthly average. It might also be useful to compare such stats among the competitors in your niche.
- Test one variable at a time during A/B testing. For example, when you’re testing how the contrast of your CTA button affects conversions, it’s not a good idea to change its position because you wouldn’t know which of them made the difference
- Use both types of data – quantitative and qualitative, when there’s a possibility. Those two different data types reveal different insights about your users. While quantitative data shows customers’ behavior, only qualitative data can show the reasons behind that behavior. Try including qualitative data in your analyses despite the negative preconceptions you or your colleagues may have.
- Keep in mind the context when you optimize. Successful optimization might not look the same for different types of pages, content, and visitors. Some pages are all about conversions, and others are there to inform users or serve some other purpose. Remember that the needs and goals of new visitors might differ from returning visitors. Also, consider the differences between the users from different sources such as organic search, email marketing, and ads.
It’s not enough just to get the data and know the methods on how to analyze it. Having patience is crucial and might be one of the most challenging aspects of establishing a fully functional data-driven design process in your organization.
After you analyze your data and make adjustments accordingly, you have to wait and watch. That is when another challenge occurs. People tend to try to understand and categorize everything as fast as they see, and in this case, that tendency is not in their favor.
Monitoring the effects that your changes on the design had on the customers is important. However, remember that there’s always some period of adjustment. If you change the design of a website or interface of an app, it will take some time for the users to relearn how to use it and get used to it.
So whether you see positive or negative effects early on, don’t rush to conclude because you might not yet see the big picture. When the initial hype of new design passes, users might see that it’s not that functional, affecting their usage. Vice versa, after an unpopular change, customers might see more benefits and find it actually useful as time goes on.
Next time you implement design changes, give your clients some time to get accustomed to it before interpreting the new data.
Here you go. We’ve covered all the points crucial for you to start successfully implementing a data-driven design process in your organization.
With time the data-driven design is going to become even more popular, and it is here to stay. So the best strategy is to embrace this evidence-based approach to the design and also help stakeholders see the value it can create for the business in this digital age.
Understanding the data-driven design process can give an edge to the designer as a professional in his career. User research, analytics, A/B testing, and other techniques enable a designer to create better designs with more ease and some evidence to back it up.
This approach can also create significant value for the business by creating a better customer experience, increasing conversions, and maximizing ROI in the long term. Learning to implement these techniques might take some time, but it’s worth the effort. It will benefit significantly in the long run.
Use insights provided by data wisely, experiment, and be patient.