Chicago Booth Review Podcast How ‘Hidden’ Features Shape Shopping Habits

Episode Transcript

Hal Weitzman: When companies sell products, they only put some of the features front and center and have hide others. How do they know if they're highlighting the right things? Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking academic research in a clear and straightforward way. I'm Hal Weitzman, and today I'm talking with Chicago Booth Giovanni Compiani about his research on salient and hidden features. Economic models traditionally assume that consumers consider all the facts before buying. So how can we work out what they really care about?

Giovanni Compiani, thank you for coming back on the Chicago Booth Review Podcast.

Giovanni Compiani: Thank you, Hal.

Hal Weitzman: We are delighted to have you back because we're going to talk here about hidden features. Everyone always wants to know what is being hidden from them, right? So let's start with this. Some economic models assume that when people shop for products, they weigh all the facts before buying. I think we know that's not quite how people buy stuff, but why is that a dangerous assumption for companies to make when they try and understand customer behavior?

Giovanni Compiani: Yeah, so I would say most models of consumer choice really make this assumption that consumers are omniscent, that they essentially have a full knowledge of all the products in the category that they're looking at. And for each product, they know everything about them, all the attributes, their price, but also their non-price attributes. Now, this assumption is made for simplicity and we make a lot of assumptions for simplicity in economics and marketing, but as you already hinted, that is not an innocuous assumption in many cases. So just to take a simple example, if we think about this podcast mic that I have in front of me, and you think about the wealth of information that is available about all podcast mics on the internet, I'm sure people at the Chicago Booth Review have done a good job at sort of researching the best podcast mic, but I suspect they probably haven't read and internalized all of the information that is available on all the products available in this category.

So the danger particularly for a company is that if you look at the data that you have or what consumers choose at what price levels, and you incorrectly assume that people know everything when they're making their choice, you're going to draw the wrong conclusions from these choices. And so you might, for example, see that people don't seem willing to pay a higher price for, let's say, a longer warranty for a podcast mic. And you might infer, oh, this is because they don't care about warranty, they don't value this. And in fact, it might be that they're just not aware of this attribute when they make their choice. They're not paying attention, but if they knew, they would actually value it a lot. So you might draw the wrong conclusions and then make the wrong decisions as a result.

Hal Weitzman: Okay. And so similarly, your view is that standard data analysis often understates how much people care about certain features. Why is that? And what are the wrong conclusions they might draw from those assumptions?

Giovanni Compiani: That is related to what we're just talking about, which is if in the data you see that people don't pay attention or don't respond to a given attribute. So you see that people don't seem to shift as let's say the length of the warranty for a mic varies in the data. They don't seem to change their choices. There's two explanations for this. One is people don't care about this. They really don't value the warranty, and that's explanation number one, but there's also separate explanation, which is they care a lot about warranty and they are just not aware of this, at least not for all the products that are available.

And so the standard approach is really sort of essentially ruling out the second explanation and loading everything onto the first one and saying it's all about preferences. If people don't respond to something, it's because they don't value it. When in fact, it's probably going to be a mixture of the two. There's going to be some preferences, but it's also going to be a little bit driven by, this pattern is going to be driven a little by attention or lack thereof. And again, if we don't take this seriously, we're going to then get a noisy or wrong erroneous estimate of how much people value the attributes. And typically it's going to be an attenuation. It's going to say people don't care about a given attribute. That's the conclusion we're going to draw when in fact they might care about it, but just not be aware of it when they make their choices.

Hal Weitzman: This is sounding so sort of familiar because we probably all worked in organizations where a piece of news is so big within the organization. We're going to extend the warranty. We're going to double the... And outside, nobody knows. They have no idea what you're talking about. And they didn't know what the warranty was in the first place and they still don't know what it is once you've doubled it. So it is more meaningful to you than it is to them. So you are basically saying, let's look at what's meaningful to them.

Giovanni Compiani: Exactly.

Hal Weitzman: As you say, there's some salient attributes and there are some hidden attributes is what we talked about right at the beginning. So talk about those characterizations, salient and hidden and how we work out which is which.

Giovanni Compiani: Right. So as salient attributes, we define this as attributes that are easily accessible to consumers when they shop. So for example, things that are immediately visible on a category or listing page on Amazon. So it could be the product prices, the brands, for example, and some of this kind of key information that is prominently displayed on a website or in a brick and mortar setting that is sort of prominently visible to consumers. The hidden, what we call hidden attributes are attributes that are typically a little bit harder and to learn during the shopping process. So things that, for example, warranty information that a lot of the times are sort of buried down a product page. So you need to kind of click through, scroll down and pay attention in order to internalize this information and use it when you're making your choice. And so these hidden attributes are the ones where you're likely to draw wrong conclusions if you're not paying attention and you're not accounting for the fact that consumers might not be fully aware of them when they're making their choices.

Hal Weitzman: Okay. And typically, so how you know if it's salient or hidden is where it's placed on the product description.

Giovanni Compiani: Right. So this is typically, this is a-

Hal Weitzman: So it's like small print versus big print.

Giovanni Compiani: A little bit, but our approach actually allows you to test this using our methodology. So you can look at the data and we have a method to really allow you to say, okay, this attribute is actually something that consumers know pretty much across the board. It's very salient to them. And now there's this other attribute, maybe warranty, something that instead requires a little bit more effort. And so sometimes it's going to get ignored by consumers leading potentially to suboptimal decisions for them.

Hal Weitzman: So Giovanni, I know it's complicated, but can you try to explain to us in plain English how you know if beyond the product description location or whatever, whether the information is salient or not for a consumer?

Giovanni Compiani: Well, we have in our paper a method that recovers how much people care about an attribute and that is robust to whether they know about that attribute or not. So it's something that works regardless of how much information consumers have. So if we compare the results that we get from this method that we developed to the standard approach, then instead assumes that consumers have full information about everything, we are able to essentially say which one is which. So if we get the same answer using the two methods, that means that consumers are likely to have full information about this given attribute. If our method gives a different answer relative to the standard method, in particular, if the standard method gives an attenuated answer, like it's underestimating the value of the attribute for consumers, then this is a sign that consumers at least some of the time are ignoring this attribute.

So that's how we learn which one is which from the data

Hal Weitzman: What they can really care about.

Giovanni Compiani: Exactly.

Hal Weitzman: So I know that the particular application of this methodology is that companies spend millions on A/B testing things like website layout. So how would your approach offer a different, more efficient alternative to the current way that people do that, A/B testing?

Giovanni Compiani: Right. So a key question for a lot of companies, particularly for e-commerce platforms is how to display information, how to sell their products, which information to display prominently and which to sort of relegate to maybe the product page and make it a little bit harder to find. And sometimes this is a decision that is dictated by just space constraints. There's only so much information that can be fit in the results page or in the category page.

So one approach to answering this question is A/B test. If you're a company, you could say, well, I'm going to, again, going with the Mike example, we're going to randomly assign some of our customers to a version of the website where the warranty information is made very prominent. It's right on the results page so everyone sees it for all the products and see what happens, see what the purchase rates are, what the returns rate are, how happy consumers seem to be.

This is very straightforward, intuitive, but it does hit a roadblock in terms of scalability because now we're talking about warranty, but you can imagine there's so many other attributes that one could do this for and the number of combinations of configurations of which attribute to make salient and which one to demote to the product page is very, very large. And companies just don't have the ability or the willingness or the resources to try all of these A/B tests. So what our method allows us to do is to essentially learn from data what consumers really care about. And once we have this result, we are able to answer essentially any number of what if questions. What would happen if this attribute was made more salient? And we can just run the model and produce an answer for that scenario, but we can consider many other, any number of these scenarios and then just pick the one that gives us the best outcome, maybe the highest purchase rate or the lowest return rate for consumers.

So it is a way to essentially answer what if questions without running the A/B test or before running the A/B test. Now we don't see this necessarily as an alternative to an A/B test or as something that completely replaces A/B test because ultimately our recommendation is still to say companies should be running a couple, a few A/B tests for the most promising configurations that we have selected. But it is a way to essentially zoom in on those configurations that seem more promising down the road.

Hal Weitzman: Giovanni Compiani, in the first half we talked about your research about hidden features. Basically there is limited real estate on things like webpages or on Amazon products. And so companies have to decide what's going to be upfront, what do they really want to tell the consumers about and what's going to be more hidden. And then based on that, consumers will buy or not, but sometimes the things that they think aren't less important are actually more important to consumers or vice versa. So based on your research, if we think about platforms like Amazon or Expedia, how should they decide which of their product features, shipping speed or hotel location, which of those should be on the front page versus relegating them to the details page?

Giovanni Compiani: I would say it depends a little bit on the company's objectives, right? So of course the companies want to make money, that is one of their key objectives, but they also want to keep customers happy. In the long run, again, Amazon cannot probably afford to make customers too unhappy because there will be a threat that they will switch to some other retailer, at least in the long run. So it depends a little bit on this, but if we focus on the sort of more long run perspective and we assume that companies want to make customers and increase customer satisfaction, what I would say is they should highlight those attributes that are currently hard to find, that are hidden attributes in our setting and that are valued by customers, that consumers really care about.

So for example, again, going back to our podcast Mike example, maybe warranty. Maybe warranty is an attribute that consumers care about, but they don't really pay attention to it when they purchase in the status quo. And so maybe they purchase something that has a very short warranty and then after the fact they realize, ah, this is not quite as good as I wanted it to be, and they maybe regret their purchase. So if that is the case, making this attribute more salient and making sure that consumers take that into account when making their purchases can substantially improve customer satisfaction and down the road purchase rates as well.

Hal Weitzman: Okay. So I'm thinking about advertising. Advertising, we definitely get big print, small print, particularly if you think about drugs in America, there's a lot of small print. So I'm just wondering, your model, could it help businesses think about advertising? What are the features? That's a great example. We have very limited space, right? You got 10 seconds or whatever to talk about the attributes of the product. Is that an example where they could use your kind of technology for thinking about, or your methodology rather, for thinking about things that customers simply aren't aware of that they should be aware of that would actually get them to buy?

Giovanni Compiani: Yes, that is absolutely another interesting use case. So we've talked about e-commerce and how to display information in e-commerce, but advertising is exactly sort of the same logic applies to an advertising setting. So the idea here is maybe you're selling a product that has a really nice new feature and that you think consumers would like. So maybe a new noise-canceling function for your podcast mic. And if it is the case that consumers value that feature a lot, you probably want to advertise it and make sure that they are aware of it so that that can drive their choice and drive your sales as a result. And so the idea here is always to look at the data, look at what consumers really care about in the product. And if as a producer or a manufacturer, your products are strong in that dimension, that's what you should be advertising and emphasizing in your communication to consumers.

Hal Weitzman: Okay. Just one thing as you're talking, thinking about wondering about consumers, the information that they have is not perfect, but should we assume that they know what they're really looking for? Because there is an assumption built into what you're talking about that they know what salient features are important, whereas really they might be fixating on something that really might be that the warranty information on this microphone is meaningless and they should be caring about something completely different. I mean, is there an opportunity to think about educating consumers more about what the salient features should be in the purchase and why they're important?

Giovanni Compiani: It is an interesting question. I would say it's complimentary to what we're doing here. Here we are assuming that sort of people know their taste and know what matters to them. And it's just that they haven't collected information or haven't paid the cost to collect all the information that is potentially available to them. But you're right that in some, especially in newer categories where consumers are still learning about what it is that they like and don't like, we might want to also allow for some learning about consumer preferences. And so this would be a dimension to extend the model and our approach and say, okay, it's not just about giving people information about the attributes, but also teaching them which ones they should be paying more attention to and which ones should they care more about.

Hal Weitzman: Okay. Now, first of all, I want to say we do not endorse any podcast microphones or anything else. We're not selling anything, but everything we've been talking about has been products relatively low cost, even a podcast microphone. I'm just thinking about big ticket items people purchase once in their lives or very rarely like say going to college or health insurance or even a home. Does the same kind of dynamic apply for a big ticket item like that?

Giovanni Compiani: This is a very interesting question and the answer is yes. And it is a little bit surprising because we might think, well, okay, maybe it's sort of not that big of a deal and it makes sense that people don't spend too much time researching podcast microphones or cereal for the breakfast because the stakes are pretty low. But it turns out that there is research showing that even for very big ticket items and when the stakes are really high, consumers still are exhibiting sort of inattention. So for example, my coauthor has done work on the health insurance market and he finds that in the US where, as we know, this is a very complex choice oftentimes to make, consumers are leaving a lot of money, hundreds or even thousands of dollars on the table. And so there is definitely scope for providing information and information interventions that would sort of help people and nudge them towards what ultimately is going to be in their best interest.

Hal Weitzman: Okay. Even for things like buying a house?

Giovanni Compiani: For a house, yes. I think that would be one very interesting area to look at. Yeah.

Hal Weitzman: A lot of what we've been talking about has been digital purchases or digital research in the case of buying a home. I'm just wondering how that would translate into the physical world. Let's imagine a grocery store, could they think about how to organize products in the store in a way that, I don't know, brings out the salient characteristics of the products and so the consumer more likely to buy them? How would it affect that kind of situation?

Giovanni Compiani: Yes, for sure. Anytime consumers have limited information about the products that they're facing, I think the methodology could be applied. In fact, in a follow-up work with our colleague, JP Dube and Joonhwi Joo from UT Dallas, we are looking exactly at this in the context of grocery retail. So we're looking using Nielsen data, we're looking at pricing attention, and we're looking at cases where basically consumers can buy the same product in different sizes. And each different sizes have different unit prices.

So for example, different per ounce price of breakfast cereal. And find that sometimes people seem to make what appear to be mistakes. So they buy a given quantity of a product when they could have at a given price, when they could have bought exactly the same quantity at a lower total price. So why do they do this? There's a few explanation, maybe it's convenience if there's a bigger pack that is just more convenient versus many smaller packs. But another explanation that we find evidence for is pricing attention. And we use exactly the machinery that we developed in this paper to answer this question and then suggest potential policy interventions. There's price transparency laws already in some states and in the U and other states are debating them. And so this project is going to speak to that conversation.

Hal Weitzman: I'm wondering also if there's a distinction between the way that people search for products online. So some people will just look, be very focused. They'll look at a product and they'll look at another product. Other people like me will open a thousand tabs and they'll look at a thousand products and then if you're like me, you forget where the tab that you wanted to find is. So does your approach work for different kinds of searches like that?

Giovanni Compiani: That is correct. And it is an important feature of the approach. So when we set out to work on this project, we really didn't want to have something that was going to work only for a very specific type of search pattern because the truth is we just don't know how consumers search in many settings and it probably depends on the category, depends, it varies across people, and it's just something that is very hard to even to measure and to exactly pin down. So we want it to be somewhat agnostic, somewhat robust to the different search patterns. And to your examples, the approach allows for both a simultaneous search process where you just consider a few good options or a few sort of seemingly good option and then just pick the one among those that is best for you or that you prefer. And that is sort of a simultaneous, if you will, search pattern.

But we also consider a sequential search pattern where you might start from one product, you investigate it, you decide whether or not it's good enough for you. And if it's not, you continue searching. If it is, you stop and purchase that. And so the approach does not hinge on a specific way in which consumers are searching or are learning about the product attributes. And I think this speaks to a more general point about the fact that when we do research, we need to make assumptions to gain any traction, but we really want these assumptions to be sort of not too restrictive and not drive and not dictate the answer that we're going to get because then that means that if the assumptions that we make are wrong, then the answer is also going to be wrong. So we want to be somewhat agnostic about things like search patterns that are just very hard to pin down exactly.

Hal Weitzman: Okay, great. Well, Giovanni, this has been a fascinating discussion about what's hidden and what's seen and what drives our shopping habits. Thank you very much for coming back on the Chicago Booth Review Podcast.

Giovanni Compiani: My pleasure. Thank you, Hal.