#065: Digital Analytics from a Psychological Perspective with Dr. Liraz Margalit

Episode Transcript

[music]

00:04 Announcer: Welcome to the Digital Analytics Power Hour. Tim, Michael, and the occasional guest discussing digital analytics issues of the day. Find them on Facebook at Facebook.com/AnalyticsHour and their website AnalyticsHour.io. And now, the Digital Analytics Power Hour.

[music]

00:28 Michael Helbling: Hi, everyone. Welcome to the Digital Analytics Power Hour. This is Episode 65. Have you ever stopped for a second as you hovered over that order button? Does anyone care how you feel right now? Data collection in digital is ubiquitous. But what lead you to this point? And how can what I know about you as a marketer help me influence others to take the same action? Certainly, optimization of the existing data helps drive the cause forward but while it may seem trite at this point, it also remains true the behavioral data we collect about consumers gives a strong signal about what has happened, that leaves a lot of murkiness around why. And so, on this episode, we should talk about that a little bit. So, obviously, I’m joined by my co-host Tim Wilson. Tim, how are you feeling today?

01:32 Tim Wilson: I’m feeling fantastic.

01:35 MH: Okay. Well, let’s analyze that statement just a little bit, because on this episode, we’re talking about the psychology of digital analytics. And obviously, to help us avoid the Jungian slips or the Freudian archetypes or whatever, we’ve got some help. Dr. Liraz Margalit, she is a web psychologist. She’s a keynote speaker, and she’s the head of Behavioral Research at ClickTale. In her work there she integrates cognitive and behavioral economic perspectives to analyze online consumer behavior and deliver actionable insights for stakeholders. She was awarded the Online Business Excellence Award for 2016 and her consumer behavior research was awarded with the Best of Neuromarketing 2016. Her research and studies can be found in top business magazines such as Entrepreneur, TechCrunch, and Forbes. And she also writes an ongoing column on the Psychology Today blog, named Behind Online Behavior. Welcome to the show, Liraz.

02:37 Liraz Margalit: Hey, thank you so much. Thank you for having me.

02:40 MH: So, right out of the gate, the first thing our listeners are gonna wanna know is, what is Tim’s deal? No, I’m just kidding.

[laughter]

02:52 TW: That’s the question for the end of the episode, after it’s [02:53] ____.

02:54 MH: That’s for the end of the… Oh yeah.

02:55 TW: Or the end of this session. Maybe that’s what it is.

02:58 MH: At the end of this session, right. Exactly. So, let’s lean back. No, but how did you get into this intersection of digital behaviors and psychology? And what lead to that point?

03:11 LM: Okay. Actually, this is a very interesting, kind of a funny question because first, when I came to ClickTale, they knew that they’re looking for someone that will help them analyze the data. Because at the end of the day, although this field belongs mostly to analyst and consultant, we are dealing with human behavior. If you think about it, all the interactions that we have, the clicks, the scrolling, this is a way to express what we think. This is a way to express different thought processes. If you scroll fast, or if you scroll slow, these are two different ways to express what you have in your mind. So then, they were smart enough to think that, “Okay, in order to really understand and not just answering the what, in order to really understand the why behind the behavior, we need to have someone that has a clue about the why.” And this is how I get into this spot, into this position.

04:19 TW: Were you recruited? Was ClickTale or a previous… Were you, was somebody convincing you that this was a space you should hop into? Or was this something that you were looking at saying, “No one is actually doing this”? Or maybe the neuromarketing was already there? It seems like an area that is not talked about so often, and yet somehow you landed in it.

04:44 LM: Yeah. So, actually, to be honest with you, I don’t know any other web psychologist in the same position, not in Israel, where I’m based and not abroad. And the way I look at the position, this is a research position and my work with ClickTale actually incorporates two different parts. First, I work closely with our consultant and analyst, try to provide answer to the question why? Why our visitors behave the way they do? What motivates their behavior? I try to uncover their decision making processes, conscious and unconscious. Because I truly believe that better understanding leads to better experience. And we all know that in today’s business, we are not selling products nor services. At the end of the day, we are selling experiences. And the other part of my work is to work closely with our innovation team, data science team. So we have data scientist and behavioral engineers and behavioral analyst. And together, what we try to do is to take all the rich experience that ClickTale has gained over the years and to productize it. So if you think about it, our clients can have actual behavioral insights in the website, as if they have their own psychologist. So this is a research position integrated with a product, a way to think differently about the product.

06:19 TW: So it’s almost like it was an economics… As you’re talking, it’s the field of behavioral economics crops up, where it’s like economists used to think that everybody behaved rationally and then they came along and said, “Oh wait, there are human beings involved,” so that whole field as it’s exploded, it is that complimentary aspect of saying, “We wish it was that clean. We wish we could just make the button larger, and because we’ve done that it is more visible and therefore people will click on it more often.” And you’re bringing… I hate to say the… It’s not the softer side, it’s just more the human psychology side.

06:54 LM: Exactly, exactly.

06:57 TW: Got it.

06:57 LM: And if you think about it today, in today’s businesses, every company that respect itself, will have a data science or innovation team, whatever you like to call it. But I think the problem today with all this innovation, is that they rely on the data to provide them with answers. And we have many, many types and sources of data, different sources of data, this is what we like to call big data today. So we have data that relates to the social media, like likes, and shares, and type of user, type of device, how many clicks. We have tons of data today. But I think that if you’re only going to pull linear regression, this is not going to provide you the insights that you’re looking for.

07:51 LM: And I can give you a very simple example. So let’s say that your data scientist discovered that all men coming from US tend to purchase more in the morning hours. So if this is actually what they discovered, this is the finding, so the recommendation will be to provide more promotions for men in the morning hours. But then, if I add another piece of information to this story, and I’m going to tell you that I observe the store, and this is actually a brick and mortar store, and I can see that men tend to purchase more in the morning hour because there is a friendly and very attractive sales person and this is actually the confounding effect, so my recommendation will be completely different. I will recommend you to add another friendly, attractive sales person in the afternoon. So this is why it’s so important to understand the why and not only relying on the data to provide us with answers.

08:56 TW: I feel like a lot of what you’re saying right now sounds a lot like what I’ve been lectured and learned over the years from the voice of customer, from the ForeSee or OpinionLab, kinda saying the same thing, that you can measure behaviorally what people are doing but if you’re not asking them why they did it… Is it sort of a cousin to that?

09:19 LM: I think there is one major difference between the voice of customer and ForeSee and the research that I conduct here, and I think the difference is to relying on the text. Meaning that I don’t think that asking our customers how did they feel in a certain interaction, I don’t believe in this idea. And I will explain why. So first, we know that it’s not scalable, so you cannot ask each and every one of your customers, but let’s just leave this reason aside.

[chuckle]

10:01 LM: But there is more fundamental reasons. And one is that there is a big gap between what people will say and the real motivators behind people behaviour. It’s not that they lie to you, of course they don’t. It’s that we just don’t know, we don’t have a clue, there are so many factors that influence our behaviour. For example we know that we are affected by environmental cues and by our emotions.

10:30 LM: Think about it, if you have a bad day, you will think differently about decisions that does not relate to the reason that you had a bad day, for example. And of course we have cognitive biases. We know that if we see an older guy we’ll think differently than if we see a younger guy. If we see certain color, if we see red we’ll think different than if we see blue color. So there’s so many factors that influence our behaviour, and people are not aware of these factors. So I don’t think and know that people are not aware for the reason, the motivators behind their behavior. So I don’t like asking people for their motivators, I like to just observe their behaviour. I think observing their behaviour can give us the perfect answer.

11:21 LM: And there is one more reason, and that is that recently, many studies that were conducted on this area found that most of our day-to-day interactions are based almost entirely on non-verbal signals, on non-verbal cue. It’s not about what you say, it’s about the way you say it, it’s about your body language, your facial expression, the distance that we have from one another. So let me give you an example, a quick example. I can ask you guys if you like to have dinner with me tonight. And of course, you will say yes because you are being polite. But from the way you say it I can infer to actually mean that you don’t like to have dinner with me. So it’s not about what people say, it’s about how they say it. What I discovered in my studies of the online interaction is that the same way that in face-to-face interaction we have non-verbal signals, we have those non-verbal cues in the digital arena. And I think that if we’ll be smart enough we’ll be able to capture those signals. I mentioned if visitors scroll faster or if visitors scroll slow, this make difference and if we’re going to nail these differences, what I like to call digital body language, then we’ll be able to understand the visitor’s mindset and how he actually feels without having to ask them.

12:51 TW: It’s behavioral, but looking at behavioral with more of a nuance from, instead of, “They did click”, but “How quickly did they click?”, or, “How quickly did they scroll?”, or, “Where was the mouse hovering?” That’s still inherently imperfect. Like in a perfect world you’re getting… It’s so funny that the founding technology guy, I’m blanking, it was Steve something at Eloqua. I think he wrote a book called Digital Body Language.

13:18 LM: Really? He did?

13:19 TW: But in that world, was trying to say we need to go beyond… I’m trying to think what his… He was talking about lead scoring and B2B. I think it was still much more mechanical of the specific actions, whereas you are really saying you wanna be able to look at… What can you do without looking at somebody’s hands and face and expression, but their behavior with more subtlety? And you’re still trying to look at it in mass or aggregate. Segmenting, yes, but you’re still trying to use a proxy for what would, in a perfect world, you’d watch everybody and literally observe… You’d be tracking their face and every aspect of their body, is that right? It’s still a…

14:08 LM: That’s what we do in the face-to-face interaction. We have a special part in our brain that all its job, all its functionality is to infer what other people’s feeling, about other people’s emotion actually. This is what this part of the brain aim to do. And by the way, just an FYI, we know that today with all the screen time that kids are exposed to there are problem for kids, in the younger generation to actually detect facial expression. Because in order to develop this part of the brain we need to be exposed to face-to-face interaction. What I’m saying that this is a really major part of the way we understand other people. And we think that it’s about what we say, but the most important part is how we say it.

15:08 LM: So, I’ll just say, okay, so you try to develop chat boards and you try to develop AIs but you don’t capture one important thing, and that’s the mindset of the visitors. Just imagine that you are going to develop a chat board that will guide you in a flower website. So you want to buy flowers to a funeral and then he wishes you a fantastic day. I think that this is something that is very wrong about how we try to develop chat boards. We try to base all our understanding, all our knowledge on the text, on how people, and on what people say, and we should actually emphasize how they are saying it. And I can tell you that only with this type of information, with pieces of data of how people interact with each other, I can infer a lot about their actual mindsets. I can automatically detect their intent after a few seconds that they land on the page.

16:13 MH: No, I think that’s really good. And actually, one concept, Liraz, that you put into some of your writing is this idea of a conversion cycle versus a conversion point. How do you suggest that marketers start to configure or build that concept into how they do measurement?

16:34 LM: First, we need to realize that conversion is not an action. This is not something we measure. I have to tell you that when I was last week in the US, we had different round tables and there is one question that they started with, and this is, “What is your average conversion rate?” And we know that typically for most clients, and I’m not only talking about retailers, we are talking about between 3% to 10%. And when I’m talking about conversion, conversion is defined as, “A prospect that completed the whole process from beginning to end, whatever this process may be, it doesn’t have to be transaction.” You can define conversion however you wish to define it. Then I have another question, so if the average conversion rate is about 3% what about the other 97%? Can we actually say something about their experience, what a negative, what a positive, can you actually infer from this piece of data that 97% of the visitors that didn’t convert had a negative experience? No, of course not. And the discussion about conversion is always, “Yes and no, black and white, converted, didn’t convert.” And I really think that we have to think differently about conversion, and here is why. Let me share a personal story with you.

18:05 LM: My husband wanted to buy me a gift for our seven years anniversary. So he entered this really fancy jewelry website and he sees there this amazing watch, but the watch was extremely expensive. So he decided to wait until I come back home and in the meantime he left the website. At the evening I came back home, and we both entered a website together, and I saw this watch that was truly an amazing watch. But then I told him, “Listen, tomorrow we are going to take the day off anyway to celebrate our anniversary, so why don’t we go to the store together, to the mall, and purchase the watch there?” So we agreed of course because he has to and…

[laughter]

18:56 LM: The next day we went to the mall and we purchased the watch. So if you think about it, what do we have here? We have two perfectly successful interactions, but if we are to measure those interactions using the old classic conversion rate, we will analyze or classify those interactions as negative, only because the visitors didn’t convert online. So we know that sometimes it takes a few iteration for the visitor to convert. And sometimes the visitor starts his journey on the website and complete his journey or his transaction on the brick and mortar store. So we really have to start thinking about conversion cycles and not about conversion rate.

19:43 MH: And in your work… Are you working with people to actually break that down? ‘Cause obviously the concept I think is really great, the technology is a little bit problematic still, in terms of actually allowing that kind of measurement that happened. Is that some of the stuff that you’re working with or doing today, as sort of…

20:04 LM: Exactly. This is what I’m working on right now, because you know what’s the problem? The problem is that we always talk about experiences. Think about it. We talk about user experience and customer experience, and we already define the error of experiences. But do you ever stop to think what is an experience? What is the definition of an experience? So in order to actually measure and evaluate experience, we have to think about the definition of an experience. And we know that the formal definition of an experience from a psychological perspective is the interaction between a subject and a stimulus that is influenced by a personal interpretation. So it’s not about the visitor or the customer, and it’s not about the layout or the design, it’s about what happens in the interaction between the two. It’s about what the person brings to the situation with him, his personal background, his set of expectation, his current mindset, his personality trait. And that’s why two people can interact with the same stimulus and they can have two completely different stories to tell about their experiences. So for example, two people can go to the exact same party, and for one person, he can describe the experience as fun and enjoyable. And for the other person, the room was too smoky, too noisy, he didn’t like the people and the food was bad.

21:45 LM: Two people can interact with the exact same stimulus and they can have two completely different stories to tell about their experience, and the same goes for their online experience, for the online interaction. Two people can interact with the exact same site and they can have two completely different stories to tell. So this is one aspect that we need to measure. And there is another aspect that is even more important, and that is that one single experience is made up of many, many smaller experiences. Take a visit to Disneyland for example. We can go Disneyland, we can go an roller coaster, we can stand on line, we can shake hands with Mickey Mouse, we can visit the Disney store.

22:33 LM: So one single experience is made up of many, many smaller experiences and we cannot possibly remember all the experiences, because we don’t have enough capacity in our brain. Think about it in one day, even in one hour, we have so many experiences. So there was an evolutionary mechanism that was developed, that its functionality is to compress the whole experience to one general feeling of how we felt about the experience. So after awhile if you are to be asked about your experiences, you are going to say, “We had a nice experience, we had a wonderful… ” You’re not going to remember it all. What I’m doing now in my work is to find a way to differentiate the different experiences. So I discovered that every page triggered different type of experience for different type of people. So as I mentioned previously, if I’m able to automatically detect a visitor’s intent, I can also understand if their interaction was positive or was it negative. And I’m able to come up with an experience score that can measure how visitors felt about the experience, but even more important, it doesn’t matter what happened in every single moment.

24:00 LM: What matters is the remembered value. What matter is what you are going to remember from an experience. And there is an example that I really like to use that Kahneman, he’s an Israeli-American psychologist, and he once told this example. He was telling about a man listening to a symphony and he really enjoyed himself and he has been listening to the symphony for 20 minutes. And at the end of the symphony, there was an awful scratching sound. And this man described this as it ruined his whole experience. But he had had the experience, and still it doesn’t matter. What matters is what left, what carry over after the experience already ended. So we need to take all these pieces of information, all this knowledge that we have to the interaction with the website.

25:00 TW: So back to earlier, it does seem like that’s the sort of thing that you would want to say towards the end of the experience. I totally get that if you ask people, “What’s your likelihood to… What’s your purchase intent?” Asking them to predict what they’re doing in the future is maybe… Or might do in the future would be unreliable but would it still be unreliable to say, “How was your experience today?” To get that sort of level. To be able to then go back and segment that other data that you have. To try to figure out what was the equivalent of that scratching sound for this group that reported they didn’t have a good experience? And I guess that still doesn’t factor in why they were coming or what their emotional state was when they arrived. It seems so tough to get to that good or bad experience based on just observed behavioral data. I’m still struggling there.

26:00 LM: Yeah, I can see what you say. My answer has two parts. First, I don’t think that we can actually ask people how you felt, just because they don’t know. This evolutionary mechanism, the remembering self, it takes time to the remembering self to figure out and to consolidate the memories. So, it’s useless to ask people. So this is one part.

26:28 TW: It’s useless to ask them right in the moment, but in theory if you could ask them two days or a week later, “How was that experience?” I definitely recognize that in myself. We had a vacation. What we’re saying the day we get back may be directionally where we wind up, but it’s really six months later, “How was that trip?” Or, “How was that experience?” That’s where you’re saying that evolutionary mechanism has consolidated it?

26:53 LM: Exactly.

26:54 TW: Is that part of it? Okay, got it.

26:55 LM: Yeah. Because we have two different selves. We have the experiencing self and we have the remembering self. And the experiencing self is the part of the brain that can answer the question, “How do you feel right now? Are you happy right now?” And there is the remembering self, and it can answer the question, “Are you happy with your life? Are you satisfied with your life?” This is the part of the brain that keep scores and documents the story of our life. Exactly as you said.

27:29 TW: Okay. And I cut you off. You had a second more important point that I interrupted. [laughter]

27:34 LM: Yeah. The more important point that I know that it sounds very difficult, but let me describe you something. Let’s say that you see someone scrolling up and down, you can see many direction changes and you can see very few stops. So you can actually understand that there is a visitor, for example, that is highly motivated to find something but he just can’t find it. So this is one of the mindset, one of the behavioral pattern that I call the disorientation or the confusion. And in order to validate it, I took this recording or session replay, because I track the mouse move of the visitors or in touch-enabled device, the scroll. And then I take this recording and I ask different people to tag this recording, to tell me what they think. And they are not psychologist, they’re just regular people, regular students. And it turns out that different people, and we’re talking about hundreds of people, will have the same interpretation of this specific recording. This means that we can actually build different algorithm that will capture the intent of the visitors. So this is exactly what we did.

28:58 LM: We observed many, many visitors’ interactions across different industries, across different verticals, different devices. And we find out that there are five behavioral patterns that repeat themselves over and over again. And then we figured those behavioral patterns, those are the users way to communicate their intent with us. This is what we like to call the digital body language of our visitors. And then we have developed five different algorithms. Five different models for each behavioral pattern, and in order to validate those model, we took these algorithms and we ran them on thousands of entirely new interactions. And then we found that we can automatically understand and discover the visitor’s intent in an accuracy level of 85%. So this is possible.

29:56 TW: Okay, that’s cool. And so, where you were starting was if… It’s like sentiment analysis when one of the ways they’ll test it is say, “Let 100 people read this tweet and see how many of them agree that this tweet is negative or positive.” And then, if you’ve matched that up to the machine sentiment analysis and say, “Look, if this many humans… ” I guess, it’s the same as if you show them just static pictures of people’s faces and say, “Is this person angry? Sad?” And then, if you can get to where the machine is smart enough. So the pipe bulb’s starting to come on for me that it’s a much more nuanced version… As you’re talking about scrolling and how often we get the request to say, “Oh, can we measure scroll depth?” But it’s literally a check mark of where we’re gonna capture in Google Analytics or Adobe Analytics, that what percentage of the users got to the 80% spot on the page.

31:00 TW: The problem is, what that’s totally missing is, did they get there ’cause they couldn’t find anything and they scrolled back up? Because they scrolled to the bottom and they weren’t finding what they looked for, and then they scrolled back up. And somebody got there ’cause they were totally engaged, and they were like, “This is awesome and amazing, exactly what I’m looking for.” Somebody else hit that same micro conversion because they couldn’t find what they were looking for at all, and they wound up being incredibly annoyed. And if I was just watching those two people who both hit that little threshold, I’d say, “Oh yeah, clearly this person is annoyed and it’s a mess. And this other person looks like they’re totally, deliberately scrolling down and consuming the content.”

31:42 LM: Exactly.

31:42 TW: I think.

31:43 LM: This is exactly my point. And I really happy that you said it, because there is a tendency of analyst today who analyze behavior in isolation. Meaning that they try to correlate different metrics with conversion. Just to give you an example, they can say, “Higher click through rate in a certain area correlates highly, or positively, or negatively with conversion.” And I think that this is a wrong attitude, a wrong approach to analyze interaction because behavior cannot be analyzed in isolation. As we say, we need to add the attribute and we need to add the following action. And a quick example, let’s imagine that we are browsing in Amazon in our smart phone. And then we detect scroll followed by another scroll, or the exact same scroll followed by a click. So, in the first example, the visitor just kept on scrolling because there was no interest in the content. But in the second example, there was an interest because after the visitor scrolled, he clicked on something. So, it’s all about analyzing the whole behavior, the behavior as a whole and not relying on different metrics. Because metrics are not the behavior. Human behavior or cognitive processes are much more complicated than to be analyzed in metrics.

33:20 MH: One question I had was, to what extent does there need to be aspects of that individual’s behavior that is known about that person? So things like, who they are, ’cause so much of digital is anonymous. Do we need to understand things like, was it a man, is it a woman, is it someone who is wealthy, is it someone who is uneducated? Are those attributes things that are necessary for us to make appropriate conclusions about someone’s motivations at that point in time?

33:50 LM: I like to differentiate between demographics that were found to have very little correlation to conversion and between personality traits. So, what we talked about first, the mindset, if the visitor is goal-oriented, if he is a browser, these are mindsets. These are a state of mind that can change from hour to hour. But there is another factor that is extremely important here, and this is the personality traits. So actually, I developed a model for personalization that is called The Online Behavior Model, and it integrates between three different elements, between the context of which the decision was made, and in the digital platform, it’s the type of visitor, it’s the type of website, type of industry. And there is the mindset, if the visitor is goal-oriented, if he just came to browse. And there is the personality trait.

34:49 LM: And there were two personality traits to greatly affect visitor’s purchase decision making. And these personality traits are the maximizer versus the satisfizer. When I’m talking about the maximizer, for example. If a visitor has a list of options, so the maximizer will go through, search through all the list, one option, one by one, and conduct a cost-benefit calculation. And he couldn’t decide until searching through all the different options. And even after he’s going to decide, he’s still never satisfied with his decision and he keeps thinking that, maybe there is another option better for him, somewhere. And if you think about it, you can actually detect these personality traits if you detect visitor’s interaction. And there is the satisfizer, and the satisfizer is the person I love to call the “good enough”. Because if he has the list of option, he’s going to choose the first option that meets his criteria, and he’s going to feel good about this option and about himself. And these are the people in general that happily live their life without being bothered about the fact that there might be a better option for them somewhere.

36:09 TW: There’s not the third option, that is the “seven year anniversary trying to get my wife something appropriate watch”, that’s not a third type?

[laughter]

36:21 TW: This all makes a ton of sense from a coloring and informing… I’m just trying to think through what the pure data scientists or somebody focused on reinforcement learning could say, and I don’t know what the response is. All that stuff doesn’t matter, if we’ve got the machine pointed at just the behavior and we can figure out there’s reinforcement learning, or deep learning, or something going on where we’re swapping stuff out just to figure out what works better, then it doesn’t really matter why. And part of it feels like well, you may not get to that point, it would be way more helpful to actually have an understanding over the why to guide that learning.

37:10 TW: Because what you’re talking about does have very much a heavy analytical component. When you talk about these are these five models and we were able to build the machine to recognize and identify these patterns, that was informed, starting from everything we’ve been talking about. I just wonder, will there be listeners or other people in the industry who would say, “All that doesn’t matter, just keep changing stuff and figure out what works, and we don’t need to know about why.” I feel like I’ve heard those cases made, not necessarily as a counter, and is that horribly flawed in some way that you can put your finger on?

37:51 LM: Yeah. I used to hear that a lot actually, and I think that today if we are going to focus on personalization for example, and about the data science development in this specific area, because this is the holy grail, how to personalize the experience in order to actually adapt different type of experiences to different type of visitors. So today we have two main approaches for personalization that is driven, that those different approaches are driven on data. So the approaches are the user-user approach and the item-item approach. So very quickly when we are talking about the user-user approach, this algorithm tries to find similarities based on visitor’s preferences. So if, for example, I ordered the movie Dirty Dancing and the movie Pretty Woman, and there is another user that ordered the exact same movies. So if the next movie that I’m going to order is going to be Ghost, this algorithm is going to offer the movie Ghost to the other visitor. It tries to understand the visitor’s preferences, and there is another approach that is the item-item approach.

39:13 LM: So this algorithm tries to find similarities in different product characteristic. So if I ordered a melon and an orange, this algorithm thinks to itself, “Okay, they are pretty much the same price, they are both tasty and they are both single served, so the next item that I’m going to offer her is going to be an orange”, for example. So those algorithm are partly successful. They were able to make some changes in the conversion rate, but the change is very, very limited. And the problem of these algorithms is that they are based on a wrong assumption. And the assumption is that our past purchases, our past behaviour can predict our future behavior. And as a psychologist I can safely assure you that this assumption is wrong, because in order to really understand and predict visitors’ behaviour there are different elements that we have to take into consideration. And one very important element is the context of which the decision was made. So, for example, I’m not going to behave in a party the same way that I’m going to behave in my work place. Different context equal different behaviours. And another thing is the visitor’s mindset that we already discussed. So that’s why if you are going to ignore all these psychological factors, you are going to make some changes in the conversion, but only to a certain point.

40:50 TW: So you’ll be optimizing within an unfortunately narrow and possibly slightly flawed… You may get to say. “Oh we’ve reached optimal, because we’ve watched whatever this thing is get better, better slow down and get better. Now we’ve achieved this state and now we’re gonna… ” It sounds like the risk is that you wind up perceiving your maximum success of the experience you’re delivering being a lower point than it could be if you factored in the… Got it.

41:23 LM: Exactly.

41:23 MH: Okay. So this is really good. One of the things I love to try to do is try to bring it back to, where can people do some things tactically? How do they get started down this path? What are the first few things you would recommend that they start to try to do to move toward a model that’s more encompassing of thinking about the person’s experience as opposed to the point of conversion?

41:46 LM: Okay. So first I would suggest to analysts to start thinking like a psychologist and to doubt everything. Meaning that very quickly we found out that people usually… And you started with this theme, people are not rational. Human being are not rational. So we need to understand what type of factors involved in their decision making processes. And we know that if you’re talking about purchase or purchase decision making, purchase decision making are found to be based on emotions. It’s how the product make you feel and not if you really need it. So there is two level mechanisms. So, the first thing is that your limbic system decides that it wants this specific product. And the second thing is that there is a rational mechanism that start making excuses why do we actually need to purchase this product, and they are two different things. So, when we optimize the experience, we need to think about how people are going to feel about our product, how people are going to feel about the layout, about the design. And then if we’re going to actually try to filter the way the decision making process works, we then need to provide them with ways to rationalize their decisions, their ideas. So we need to have emotional content and rational content, and we have to keep thinking like psychologists. Asking the why behind the behavior, and not settle for only the what.

43:29 TW: So on that front, is there a variation of the degree that the emotional and feeling side comes in when comparing types of industry? I could see buying a luxury watch or apparel or retail consumer stuff all the way to maybe a very considered purchase, B2B mechanical. Are there cases where there is much less potential? I’m thinking about some past companies I’ve worked for, as well as some current clients where it is a much more, “I need a product to fulfill a need.” And is there a level where that plays in less? I could see there’s no place that the experience and the mindset and the emotion and the feelings don’t matter at all, but does it vary in degrees as to how much, based on the industry and the type of product and the target customer?

44:28 LM: Of course. Of course. Of course when we’re talking about insurance or financial product, of course we are applying a more rational cognitive process. But we have to understand that there are different types of products, and there are nice-to-have products, and there are functional products. So when I’m talking about nice-to-have products, so for example if I have a closet full of dresses, I don’t really need another dress. I will purchase it because of how it makes me feel when I picture myself wearing it. But when I’m talking about functional products, we purchase them only because of the function. They don’t make us feel anything, like light bulb or USB or different type of connections. So, we really have to understand the type of products we try to cater. And another thing is that if we really understand that we have different types of people, so we really have to stop use the concept best practices, because if we think about it, what are best practices? Best practices actually means that there is a common denominator that can be applied to different types of people. But we try to actually think differently about people, and personalization actually is the realization that we have different types of people, so we need to provide them different type of experiences.

46:05 LM: So if we actually agree on this, and I think we all can agree on what is personalization and why we really need to be in that place, we really have to stop use the term best practices, and we have to understand the type of our audience, and for different type of audience, we like to provide different type of experiences. And another thing is the different type of people will differ in how they treat different type of products. So, for example, for me, I can think of computers as a functional product, and you can think of something to wear as a functional product. So we really need to understand how different visitors interact with different products.

46:51 MH: No, that’s really good, and I like that statement about best practices as well. It’s all too often people tend to try to shoot down the middle and they leave a lot of good things out because of that. But one of the best practices that we have here on the show is to go around and do a last call where we talk about things that we found in the last couple of weeks that we think are interesting. And this has been a really great conversation Dr. Margalit, and thank you so much for that. But why don’t we kick off the last calls. Do you have a last call that you wanna share?

47:24 LM: Sure. So, actually if there is one book that I really enjoyed and I think that can be really relevant to all the people that actually want to better understand the psychology and the decision making processes, is How We Decide by Jonah Lehrer. And I highly recommend this book, and I think there were different negative things that people said about this book, that there are some inaccuracies there, but I think in general, if you read this book thoroughly it can change your perspective about how we make decisions.

[chuckle]

48:08 MH: Nice.

48:09 TW: So Jonah Lehrer had a little bit of a… You’re saying it’s good content, even though… He got in all sorts of hot water based on his poor behavior from a journalist perspective, right?

48:20 LM: Yeah, I know. But you know how people, they try to generalize the whole book because… Yeah, I agree, what he did was very bad. But you can find very good content in there. I don’t think that…

48:35 TW: No, no. If it’s good content… I actually remember hearing him or reading him, something, as he reflected on what had happened. ‘Cause I think he wound up being… Yeah, it was tough. He was young to have his industry or his career kinda trashed, but How We Decide… I almost was gonna ask, is there a good book somebody could read? ‘Cause all I could think about is the Predictably Irrational, which is more the behavioural economics side of it. The How We Decide sounds like an awesome way to think a little bit different about this whole area.

49:08 LM: Right. And also, Thinking Fast and Slow, by Kahneman.

49:12 TW: Oh, I think you’ve got it.

49:13 MH: Oh yeah. That’s on my list.

49:15 LM: Great.

49:15 MH: Which actually brings up my last call, which is a book I just finished recently, called The Undoing Project, by Michael Lewis, which actually talks a lot about the life of Amos Tversky and Daniel Kahneman and their work, also a couple of Israeli psychologists. So that book is really interesting because obviously it walks through a lot of what they came up with, but it also goes into a lot of depth about their relationship and collaboration over the years, which is really profound. So, anyways, really enjoyable book, somewhat tangential to what we just discussed, but still very interesting. And so, if you’re interested in that, that is definitely a great book to read. ‘Cause they, basically they’re founding fathers of behavioral economics and the work that they did in the ’70s around proving that we are not rational in our indecision making. So, anyways, really good stuff. Anyways, Tim, what about you?

50:15 TW: And they were two very, very different personalities, right, who worked very well together.

50:17 MH: Oh, extremely different. Yeah, very, very different from each other.

50:22 TW: Oh my God, I really need to finish my last reading assignment so I can move on to this other stuff.

50:27 MH: I know right? [chuckle]

50:30 TW: So just in case people are saying, “No, I don’t wanna read six books. I’d rather take a course.” My last call, which maybe is a little ironic or it is… I’ve came across a post on Medium, it’s called Every Single Machine Learning Course on the Internet, Ranked by Reviews. It’s a long post but it’s, as I’ve keep thinking I need to just understand what this whole machine learning world is. It’s nice because it actually has a pretty good little write up of what’s the approach. Is this coming at it from, “Oh, this covers analytics and machine learning”? Of course, I’m like, “I don’t fully understand the difference between the two.” They even talk about what tools are used in the course. So I’m not claiming that I am going to sign up and take one of those, especially now that I have four books that I need to read that are not down that path. But it’s a good post for people who are maybe looking to dive into some open course work on that topic.

51:28 MH: No, that’s great. And thank you for that. And again, this is a really interesting conversation because so much of what we do in digital analytics, we often describe as art and science. But actually, there’s science to apply to the what we would call sometimes the art, which is knowing who the customer is, and applying good thinking to themselves. Thank you very much Dr. Margalit, it’s been a pleasure having you on the show. As you’ve been listening to the show today and if you’ve heard things that you are interested in learning more about, or would like to ask questions, please don’t hesitate to get to us on our Facebook page or Twitter account, or on the Measure Slack. And we would love to hear from you. We’d love to gather your questions and I’m sure whereas you probably would love to hear from folks too if they have questions about any of this.

52:23 LM: Yes, of course.

52:25 MH: So I don’t know if you are on the Measure Slack at all but we highly recommend it as a place to understand the psychology of analysts.

[laughter]

52:37 MH: But anyway, it’s been great having you on. Thank you very much.

52:40 LM: Thank you. Thank you for having me. I really enjoyed the conversation. Thank you.

52:44 MH: Yeah, and for my co-host, Tim Wilson, remember, keep analyzing.

[music]

52:55 Announcer: Thanks for listening. And don’t forget to join the conversation on Facebook, Twitter or Measure Slack group. We welcome your comments and questions. Visit us on the web at Analyticshour.io, Facebook.com/analyticshour or at Analyticshour on Twitter.

[music]

53:15 Charles Barkley: So smart guys want to fit in, so they made up a term called analytics. Analytics don’t work.

[music]

53:24 LM: It’s okay. But people actually start telling me things about their personal lives and about their issues and their problem. And I’m like, “I’m not your psychologist. This is not the appropriate time.”

53:36 TW: I think Michael gets the same thing. It’s just he actually has no professional training where he’s qualified to hear them for that.

53:43 MH: Exactly.

[laughter]

53:44 MH: So of course I just dispense it by its Willy Nilly and it’s no problem.

53:50 TW: We could ask Liraz that. Do most people expect to have a search box on the homepage of their website?

53:56 MH: I don’t think that we need to have that conversation right now, Tim.

[chuckle]
[music]

54:03 TW: So, you know how normally, if you go for somebody who’s like, you have a formal and then the more you get to know them as you talk to them you get to a little more informal. Somehow, Michael, you started as Liraz and by the end you’d shifted to Dr. Margalit. So, I feel like you might’ve just gotten more intimidated throughout that.

54:20 MH: No, no.

[laughter]

54:21 MH: That was not it. I have a purpose for why I do what I do.

54:28 TW: I’ve tried to read your body language to understand…

[laughter]

54:31 MH: No. You can’t do it.

[laughter]

54:36 TW: Rock, flag, and psychology.

[music]

 

Podcast: Download | Embed

Subscribe: RSS