What do we mean when we talk about ‘data’?
How can you use data and insights to create design hypotheses that allow you to create and test your designs effectively and help remove bias?
What did we get up to at Christmas?
The Design Untangled Podcast
Episode – DU031 – Designing with Data
Host: Chris Mears and Carla Lindarte
(00:16) Chris: Hello and welcome to Design Untangled with me Chris Mears and joining me, in 2019, is Carla Lindarte. How is it going?
(00:24) Carla: Hello, how are you doing?
(00:26) Chris: Yes, good. So we are back from our unscheduled Christmas break, which we decided to have but did not tell anyone about. So thanks. Thanks for bearing with us and not deleting the feed, if you have not already. So how was your Christmas?
(00:42) Carla: A part of being really lazy and not doing any podcasts during the period. It is quite good. A bit busy sometimes, a bit quiet, other times. So it is a good time to disconnect and think about something different, that is not work or anything like that.
(00:58) Chris: Or a podcast.
(00:58) Carla: Or a podcast, is it not? How was yours?
(01:02) Chris: Yes, pretty good. I cooked for the family and did not ruin the turkey so it was not too bad.
(01:07) Carla: Wow. Was that the first time?
(01:11) Chris: No, it was not the first time, but it was the first time that the, I did not ruin
(01:16) Carla: Oh wow. Well done. You have to invite me next year to eat the turkey. (01:21) Chris: Well I thought you were a vegetarian or have you given up on that now? (01:23) Carla: No, I am not vegetarian anymore. I only eat chicken and fish and no red
meat. That is all I do no eat.
(01:31) Chris: Okay. So a turkey would be fine. I still remember when I had you around for dinner and cooked you like this big meatball thing and I was like, oh yes. Shit.
(01:43) Carla: Oh yes, you are a vegetarian. Oh yes, old times.
(01:48) Chris: Yes. Very old times. So what are we talking about?
(01:52) Carla: I do not know. I do not know what we are talking about. When was the last time we actually did a podcast? It was a long time ago. And it was the interview with Brendan, was it not?
(02:01) Chris: Yes. And that was pre-recorded a few weeks ahead. So it has been a long time since we actually did one of these, so probably, a bit rusty.
(02:09) Carla: Yes. Let us apologize in advance for any weird interactions like the one we just had right now. But anyway, let us talk about design. What is it that we are going to call it?
(02:27) Chris: Designing with Data, I think we are going to go with.
(02:29) Carla: Yes, Designing with Data. So what is data? How do you define data,
(02:38) Chris: Well, I think the first thing that comes to mind when you say data, is people just think of numbers, right? But data in the context of UX design is any sort of data. So it can be quant or qual. So we are talking about user interviews or a type of data equally, your analytics and stuff which you might expect to be, also forms part of that data. And they are all just inputs really that help inform the design that you are going to create.
(03:08) Carla: Yes, that is absolutely right. I like to call it rather than data. I like to call it insights. It is actually, towards the end of last year, I had run an event at Google, and we actually talked about data. And for creatives, for creative designers, especially in the advertising industry, there is a still a lot of perception when you talk data or when you say that word of data or in the world of advertising it is called programmatic, which is all this technical stuff. It is just complex and difficult, and is something that designers, creative people, or UX designers, are not used to that. At the end of the day, any design process that we do is based on some kind of insight. Right? Even if the insight is coming, from a set of interviews or it is coming from the business, and the client, telling you that they have identified that being a problem or it is always coming from an insight rather than, so you see it as an insight, it is easier to kind of visualize, than just data. Because data is a bit scary is it not?
(04:15) Chris: Yes. I think they are kind of different though. Because data, you can just say 60% of people drop out on that page. That is data, right? But it needs a layer of interpretation to turn into an insight. So I think they are a bit different. But the insight is the thing that you want to be working with rather than, just saying, oh 60 people drop off here. You need to kind of imply that it might mean, one of your CTA is not working or whatever.
(04:43) Carla: Yes, exactly. So that is what I mean. Do not think data, think of insight. What is the data actually telling you? So, there are different types of data, in the advertising world, they talk about first party data, second party data, third party data, but that is more for marketing. So your first party data is the data the company produces themselves in their website or their CRM systems, et cetera. Second party data is data you get from partnerships like Google data ias that a good example of that. Third party data is data that you buy? But that is from an advertising perspective. From a UX perspective, you have qual and quant as you mentioned. In the quant side of things, there is a lot of things that you could think of, but from a UX or design perspective of product design, I would say analytics, from a quant perspective, they are very powerful. Getting access to the analytics of existing channels. Even if you created a new one, for example, working on a brand new app, you still would like to look at some of the analytics of the website, if the client has a website, et cetera. Surveys as well, very, very useful.
(06:04) Chris: Classic. Very classic.
(06:04) Carla: Yes, they are classic and a lot of people do not use them that much I think, because the tend to be annoying, but they are still very powerful.
(06:14) Chris: The thing with those as well is it relates more to products that are already in existence I suppose. Because obviously you are not going to have any analytics if you are coming up with something completely new. So that is where, potentially, you need to look at those third party data sources, to look at things similar to what you are trying to achieve, if you do not have a live product as it stands.
(06:37) Carla: Yes, that is correct. So that is why things like social media listening, social media data is really useful if you are exploring a new market. You want to look at that particular brand and what similar people actually do. Go in and looking at social media channels or what they are interested in. So that is really powerful. Google data is actually really powerful, not just because I worked for Google, but it is actually. If you had access to a marketing team, or advertising team, or an agency, just go and talk to these people and see if you can pull up some of the Google data available for that, if you are already working with a brand. If you creating something brand new, there is a lot of Google analytics that gives you a lot of information. Google trends is another tool. So there are multiple tools that you can use from a Google perspective, to understand what people actually do, what they search for, where they have been to, where they visit. I know it sounds a bit creepy, but it is useful when you are trying to create something.
(07:44) Chris: Yes. It is all just building up as much of an idea of your target user as you can. Obviously, you always want to speak to these people when you can. But there are these other avenues which you might not have thought about, as sort of UX tools that will enable you to increase your knowledge.
(08:05) Carla: Yes, exactly. I remember as well when I was working more in UX and less in this advertising crazy world. I used to go a lot, if you do not have a budget or time to do user interviews, going to YouTube and watching people talking about the topics you are researching, if it is health related, or if it is a fashion related, et cetera, it is actually quite useful. Because you do not always have money and time to do research and get people, that is kind of one source. You can get some inside on the back of those. As we have spoken in other episodes, every research, even if it is just generating ideas or evaluating an idea, is also a quick and easy way to get it done.
(08:56) Chris: So I guess the point of all this data and ultimately the insights you can derive from them, to give you a theory as to what the right design might be for your user. And that brings us on to hypothesis driven design, which is kind of tangled up within this subject, I think. And this is where you are basically, putting forward a theory as to why you think this design is going to work, and it allows you to actually test if you were right or not. And you can only create that hypothesis when you have got enough information from the data, be that qual or quant, that you have gathered before starting the design process. You really need that to actually form these hypotheses that you can then take in, and test your designs against.
(09:48) Carla: Yes, exactly. I mean, all these data, if you have it initially, you are going to come up with a lot of ideas of what the solution will be. And all these different ideas, do you articulate them as a hypothesis? Which is, what is a hypothesis, is an assumption right? You can write them in different ways. People write them as user stories, as stuff to be done, or just sentences, you just write an assumption, an affirmation of an assumption that you have, based on the information that you have in that moment, of a potential solution for a problem. So I found that really useful, once I was doing this project for Adidas, it was really useful to get the project team to write down, together lots of hypotheses because you kind of get people involved in the process, and it is kind of a clear way of articulating people’s ideas. So you say, let us just put together some hypothesis of what we think the solutions will be. I think it is good to collaborate on those. The only problem with that is, and you end up with too many hypotheses, which is also a problem.
(11:05) Chris: Yes, but that just comes down to prioritization I think. Like what is the most important thing you need to prove or disprove and how, and maybe you can also do it against the level of uncertainty. You have gotten those different theories as well, so you want to test the ones you are most uncertain about first, usually, and as a prioritization exercise. The thing that I have found with hypotheses is they are fucking difficult to do and they take a really long time to actually get to the point where they are kind of clear enough and you are focusing on a singular thing that you believe. It is really hard to get down to that level without being too vague, because you have got to be able to test these when you take it into the lab or however you are testing the designs and it is really hard get them specific enough to actually do that.
(11:56) Carla: Yes, that is true. I also think that depending on the stage that you are, they sometimes become very high level. And as you say, if they are very high level, they are harder to prove or disprove, but what you could do is just create some kind of sub- hypotheses at the back of these high level one. That allows you to be more specific about what are the things that you want to test, to prove or disprove. Because yes, you absolutely right, if you have a very high level hypothesis it will be harder to just validate or disprove that with one, research activity or unless you put your whole research plan to validate one kind of major hypothesis and then just have multiple things at the back of it, is really hard to do. I also think that, as you said, it is really hard to get to a common way of defining hypothesis or writing hypothesis. I mean I was recently looking at a a research plan from someone where I worked for, and they were, this is the list of hypothesis of some pieces of research are they doing? And they all very, very different. Like sometimes they are just a massive paragraph with a lot of information in them. Some of them is just more like research objectives rather than hypothesis. So we want to understand that this is not. I think the way I like writing hypothesis more affirmations of something, and very simple affirmations, rather than, big paragraphs with a lot of information in them.
(13:38) Chris: It makes it quite obvious to stakeholders as well, what their internal assumptions actually are. And it makes it easier to say this is what you thought, this is what we saw, it did not work out. Whereas, if it is a bit wishy washy and we kind of think if we change this color a bit might work, that does not really get you anywhere or answer any questions. So yes, it is painful but you have really got to spend the time just getting them really nail on the head, very defined so that there is no question, once you come out of however you are testing it, whether you have proved or disproved it.
(14:15) Carla: Yes. And as you said, like prioritizing based on how certain you are about the affirmation or assumption that you making. But sometimes you will prioritize the ones that, you know the most. Sometimes development starts with the easiest thing, just to show that they are actually going fast, but it should be the other way around.
(14:42) Chris: Yes, it depends on the project and the environment. You always use a bit of common sense there.
(14:42) Carla: Also, there is always this conversation of quant and qual and what data is best, to create your hypotheses. And I think you always need to be able to try to combine both. You cannot do just one method in my opinion. I think it has to be, depending on what you are trying to achieve, one type of data would be better than the other. But always try to, throughout the duration of your project, try to always use both, when you can.
(15:20) Chris: Yes. I mean anything, the more insight you have got about the people you are designing for, the better your design is going to be. So, the more you can gather from as many sources as possible, it is likely that you are going to have a better go at your design.
(15:35) Carla: And the more frequently you do it as well. That is why it is good to test, all the time, and not just wait until the end, to do the testing.
(15:46) Chris: And also, I think having more than one sample source as well helps eliminate some of the issues you might find in data. So, as I was saying, data requires interpretation and that interpretation can be subject to bias whether kind of consciously or unconsciously. So you throw out a survey, whoever is creating the survey is deliberately asking about certain things and other things because you cannot ask unlimited questions, right. So you are inevitably applying some sort of bias to the results you are going to get. So by using more than one different source, you can try and smooth that out a little bit, hopefully, anyway.
(16:25) Carla: Exactly. I mean, biases are always there in your mind. And they are very difficult to fight. But the more different types of information and different types of people you bring into your research and your design process, the better. I was recently in a conference, a guy from Google was saying that there is a lot of like assumptions that we make based on things that we know, our biases, and information that we store in our minds. So, he was saying that gaming for example, it is always been known as something for kids. But data, and that is why data is powerful, has demonstrated that 45% of video games searchers are mobile, are over 35 especially in the U.S. So when you think about all of these things, you are working on a particular project and you have all these biases in your mind, you really need to rely on the data available, to be able to create the hypothesis and prove them or disprove them. I make sure that you include as many different types of angles to the problem, to the solution, to the problem.
(17:41) Chris: We should do an episode on biases. Actually, it is pretty interesting.
(17:44) Carla: Yes. It is really interesting. I was reading that book thing, “Fast and Slow, and Then Slow”, and it talks about the system one, system two, and how majority of your actions are based on the system one, which is the unconscious one. A lot of the things that you have learned are actually stored in there. So your reactions, your kind of primary reactions to things, before you actually think about something, are based on the things that you have experienced in life. So biases are very, very powerful and can really change the course of any design process if you do not care about that.
(18:26) Chris: Yes. And from the user’s point of view as well, right. It is going to affect their interaction with your design and all those biases that they have accumulated throughout their lives that is going to affect, how they interact with it. So yes, it comes into play both from the side of the designer and also the person using the design.
(18:46) Carla: We might find an expert in unconscious biases. There was a lady who runs these workshops at Google. She is really good. We might just do that.
(18:55) Chris: Everything happens at Google, does not it? It is has got so many plugs on this episode.
(19:00) Carla: Right. Now, that was not intended to do any plugs for Google, because anything I say has nothing to do with Google. But they do have a lot of stuff.
(19:10) Chris: They do definitely have a lot of stuff. Give them that.
(19:13) Carla: They do have a lot of stuff. It was also, in that event, as I mentioned before, I was having a conversation with someone who attended the event about whether or not designers especially in the creative space, advertising creative space should learn more about data. You know, should they learn more about analytics? Should they learn more about statistics? Should they learn about all this stuff? Or is that always something that is going to be done by someone else? And then there is just kind of the brain to comes up with the ideas. What do you think? Do you think like designers and creatives should be more data-driven or not?
(19:59) Chris: I think they should. I do not think that they should always be the ones doing the full time kind of digging into the data. I kind of believe a bit more in specialism, but with the appreciation for what other people do. So I think if you do not have the skills to drill into databases and run SQL queries and stuff, then I do not think it is necessarily your job as a designer to go out and learn that. But I think what you should be doing is working with the person who can do that for you and helping them understand what you are trying to get from the data as well. And yes, there is nothing wrong with learning new skills, of course.
(20:42) Carla: Yes. I think that is my point. I mean the more designers and creatives know about what data is available, even if they do not necessarily go out there and pull it out from the systems, as you said. They need to have an understanding, in my opinion, of the possibilities of data or the things that they can actually learn from the data. So, we were talking on that day, at that event about signals. It could be signals of like weather signals or media signals. So people who view, these, also view that. And there is a lot of data and information that is out there that designers should know more about what is there. So then when they tackle a new creative brief, they think about all the possibilities. So before, and I do not know of you have done it as well, I used to present a lot of design hypothesis and assumptions based on some interviews and perhaps some information about analytics. And I think now, there is a lot of tools and opportunity to go deeper and be more certain about the different options that you can have, as a solution, a design solution. And also the perception of just having one big solution, a big idea. I think will change with time and it is just going to be more, as personalization is actually real. Because, right now we talk about personalization but very few people are actually delivering it. As you know, personalization becomes more a thing, you could actually say that you can come up with different solutions for the same product but based on different types of audiences or different types of users. So my point is, that the more understanding you have of the possibilities of data, the more of a better designer you are going to be, in my opinion. But you are right, I do not necessarily think they are the ones who will be sitting down, and they need to just work with data scientists basically to just make sure. But they also need to understand their language and the way they communicate as well. So then they can interpret that data and get the right insight to what they are delivering.
(23:16) Chris: Yes, exactly. And I think as we move forward, this is only going to become more prevalent and I think they will probably be more user friendly ways to get at data as well. So it might be a bit less about writing database queries and you will have some sort of interface to help you gather that data a bit easier. That we shall see.
(23:37) Carla: I mean not to talk about Google again. Recently they have been introducing natural language in the way they ask question to Google analytics, which I thought it was quite cool because you do not necessarily have to go there and kind of try to press all the right buttons. You just ask a question and then the system kind of pulls all the information you need. It is very nascent still, but, I think that is where they are going. So, you are absolutely right, that with time, all these interfaces will be more user friendly and more accessible for everyone.
(24:19) Chris: Alright, we got any more on data? What did we call this episode again? Design with data.
(24:25) Carla: Design with Data and plugs to Google.
(24:29) Chris: Sponsored by Google.
(24:29) Carla: That was really bad. I did not mean it though. Recently, I have been talking about data and creativity, so that is why it is fresh in my mind, so that is why.
(24:45) Chris: That is it. First one of the year, done.
(24:47) Carla: Yes. First one of the year. So hope this was useful. If you have any questions, or any feedback, if you have any ideas of new topics you want to talk about, people you want us to interview, excluding Donald Trump. I do not want to interview Donald Trump. We can always find them, and yes, just whatever you want, just let us know. Give us some feedback.
(25:12) Chris: What did you think of the fireworks this year? Did you watch them?
(25:17) Carla: Yes. Watched them on my TV. They were really cool.
(25:20) Chris: Yes, very, very political fireworks were they not?
(25:24) Carla: I literally have decided not to read any news anymore.
25:27 Chris: I think that is a very, very good idea and a good thing to take forward into 2019. I suggest all our listeners turn off their CNN or BBC news, whatever you are watching and go and have a beer.
(25:41) Carla: Yes. Or listen to us.
(25:43) Chris: Yes. Just on loop. And by the end of it, they would all be Google clones, I think.
(25:48) Carla: Oh no, I have been Google brainwashed then. Alright. Well, I will see you next time then.
(25:57) Chris: Yes, see you later. Bye.
Narrator: Search and subscribe to Design Untangled using your favorite podcast app and leave us a review. Follow us on the web at designuntangled.co.uk or on Twitter @designuntangled. Become a better designer with online mentoring at uxmentor.me.