#301 What the Like Button Tells You About Your Customers with Bob Goodson, Inventor of the Like Button

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com.

When new things get developed, especially when they're as successful as something like the like button and now emotional reaction buttons. It's never one person, it's never one company. It's always a long string of little tweaks and innovations that culminate in something that's very popular and successful. So it's really more of a process of evolution. The like button paved the way and conditioned people to get used to reacting to content. And that made it much more likely that we would react to an email or a text or a call.

Welcome to Data Framed. This is Richie. The like button is a ubiquitous feature of social media, messaging and content platforms. It's such a simple user interface feature, but one that generates a powerful data set for understanding both your customers and your content. And it's not just simple number crunching. Like More likes means Better. The like button opens up an exciting world of graph analytics, analytics that I'm fairly certain most companies are failing to take advantage of. Our guest today is Bob Goodson, who in the stint as a product manager at Yelp 20 years ago, was one of the inventors of the like button. He's also the author of a book about the history of the like button called like the Button that Changed the World. These days, Bob is president at market intelligence company Quid, where he continues to help companies understand their customers better. Have a listen to this episode. I hope you like it. Hi Bob, welcome to the show.

Hey Richie.

Wonderful. So, just to begin with, tell me, how did the idea for the like button come about?

Well, it was 20 years ago and May 2005. I was working as a product manager in San Francisco for Yelp US review site and we were the early days of Yelp and we were trying to figure out how to make this feature that was called Send a Compliment even more reactive and quicker and easier to use because it took a few page clicks, which at the time, if you wanted to give any kind of feedback on the Internet. That's what was involved, believe it or not. So our co founder Russ came over one day and said, I think I could use JavaScript to keep people on the page so when they look at a business listing, they could click multiple reviews and they wouldn't leave the page to communicate whether it's useful, funny or cool. And that would allow us to do more with the reviews and show better reviews to people near the top of the page and other useful things. So he asked me to come up with a few potential designs, one of which I put the sketch in the book, was the word like in a thumbs up symbol, which we actually didn't use at Yelp. We ended up choosing useful, funny, cool as three separate buttons without an icon. And we became the first company to put multiple emotions on a piece of web content and the first to allow like a single click without leaving the page to be able to react to content. And we were just trying to solve that day's design problem. We really only worked on it for a day and yeah, it kind of went from there. But the book is all about the fact that when new things get developed, especially when they're as successful as something like the like button and now emotional reaction buttons, it's never one person, it's never one company. It's always a long string of little tweaks and innovations that culminate in something that's very popular and successful. So it's really more of a process of evolution, which is what we're pointing to and trying to bring light to in the book.

That is very, very cool. It was a one day project just solving a little thing. And this is now a ubiquitous feature of just many websites. So I'm always curious, whenever something is just everywhere, you think, okay, well, this is an obvious solution to a problem. But I guess at the time it was brand new, it wasn't obvious at all. So did you try other things beyond just the idea of the thumbs up like, what else was tried beyond what we now know as the like button?

Yeah, I mean, even that day, you can see from the notes, we came up with various ideas for how to present this and how to interact with it. It really wasn't an obvious choice. The thumb wasn't an obvious choice at the time, neither was the word like, which is why it's interesting that it was in the notes. And when we trace back the origins of other sites and we talk to founders, like through this book, we talk to the founders of YouTube and LinkedIn and Twitter, and recently interviewed the founder of Pinterest. And we talked to so many people, and around that time, in the sort of early 2000s, so many things were being used. Like the first icon on YouTube was actually a heart symbol, which wasn't used much at all at the time. But when I spoke to Steve Chen, one of the co founders, he was saying that they started with a heart partly because YouTube started out as a dating site, actually only for two or three weeks. Most people don't remember that era, but they thought it would be a good place to add videos for dating purposes. But it didn't last very long, and they quickly realized there was a bigger need and opportunity. But instead of any kind of thumb or like, they actually had hearts on those videos that quickly got changed. Friend feed, which later got acquired by Facebook, actually had just little smiley face, even though they. They were the first to put the word like on a social media news feed, which has now become a natural part of all social networks. But at the time, it wasn't even for them with the word like. It wasn't obvious the thumb would go next to it. They had a smiley face. So many different icons were being tried. And it seems like the thumb has some magical properties to it that made it a very accepted icon in the West. Absolutely.

That's fascinating the way the idea just went from one company to the next. I also have no idea about YouTube originally being a dating site. That's quite a pivot, really. So it's interesting. So the thumbs kind of won it, I guess. You occasionally see things like hearts as well. Do you want to tell me a bit more about how the progression went then beyond that, like, how did it take over the world?

It's been interesting for me to watch because on that day we put these buttons on the content and none of us thought that something important was happening. Everyone we spoke to, by the way, that made important contributions to the development of the like button today have said the same thing, that they were solving that day's design challenge. And. And the thing evolves, right? And now there have been many thousands of people that have worked on this feature and many different companies. And so mostly the thing that I've. I noticed over the years is that first of all, we didn't think too much about it. Within a year or so, I saw it start to spread to other sites. I remember thinking, really, every few years, for the last 20 years, I've witnessed, because I have a background in user interface design, I'm interested in these things and all the apps I use, I'M curious, like, how they're evolving and. And I would see this feature kind of gradually spread and become more and more popular. It was about three years after we did that at Yelp that Facebook added the like button to its site. So it even took several years before it reached Facebook, which most people think of as sort of the. The iconic implementation of that feature. And yeah, and I suppose I just kept thinking, every few years someone's going to write a book about this or someone's going to write like an amazing Wired article about this feature. And I would search Google every few years and look for that thing. And no one had written it, which was really surprising. So I ended up chatting to a good friend of mine, Martin Reeves, and we've written some things before a few years ago, and I mentioned this fact to him and he was so curious about the origin of the button. He asked questions like, you have, like, where did it come from? Or who really did invent it? And I couldn't answer these questions. So we started pulling the thread and as we researched it, we just found it a really interesting keyhole. When you ask this question, who invented the like button? Where did it come from? How has it changed the way we interact online? You actually uncover all sorts of new and deeper narratives that have lessons broadly when it comes to understanding tech, how companies can innovate, and what it means for future waves of technology as well.

Lots of great questions. I think we're going to get some insight onto some of those in the rest of this episode. All right, so I guess from a technical point of view, it's like, okay, as a user of a website, you click like, it's not immediately obvious what happens next. So when you click the button, what goes on behind the scenes?

Yeah, those of you interested in data and analytics will enjoy thinking about and hearing about some of this. First of all, before we get into some of the interesting data that flows from likes, I would point out that liking is. And reacting to content is what I describe as the atomic unit of user generated content. So Web 2.0 was all about the era of user generated content. Essentially, the web went booped from being something that around 10% of people on the web would contribute or create content. About 90% of users would read the content. And we can tell this from anyone that's developing apps in the early 2000s would look at their analytics log and they would see they got all these visitors coming in, consuming content and very few people actually creating the content because the barrier to create it was quite high. There wasn't really the level of, like, trust in creating content that there is now. I know there are other types of trust issues, but at the time, it was almost like a niche activity to create content, and almost everyone was reading it. What Web 2.0 did is it effectively brought 90% of people that use the web into content creator roles. And now that number's probably much higher. What companies were doing at the time is trying to figure out ways to bring more people in, ways to engage people into content. And of course, photos. And uploading photos is part of that. Uploading videos, commenting, sharing, forums, all of these are things that brought down the barrier. But nothing then or now is a simpler way to generate content than simply clicking on something to give you a reaction. And most people don't think of that act as an act of content creation or an active data contribution. They're merely trying to signal to someone else that they appreciate or like they're expressing something to another person. But of course, when someone clicks, like, then they are putting their name to that content, a face to that content, and they're also contributing to the underlying graph. And much has been said about the social graph, which of course was very powerful and, and is behind a company like Meta, ultimately, because of the social graph that Facebook mapped and created. Meaning, we know that you and I are friends because we're connected on a platform, for example, so it draws a line between those nodes and the network has more information that it can use. And I would argue that the second wave of graphing was actually from like data. We can think of that as the like graph. It's the fact that you and I liked or reacted to a piece of content. The third wave that's come since we can think of as the attention graph. You and I looked at the same piece of content or watched the same video. That too can form a graph or part of a graph. So the latest wave of social media is mostly driven by that. With the rise of TikTok, which is really the main graph they care about. The only graph they seem to care about is the attention and what that tells them about what other people would want to see.

Yeah, I hadn't really thought of the like button being around the same time as all these other Web 2.0 companies. So you had Wikipedia started out about, I guess, a similar time where it's like, okay, you can write encyclopedia articles, you've got things like WordPress maybe a little bit later, and, well, the rise of many blogging platforms where you can write your own blogs. But, yeah, just being able to click a button and, yeah, a very simple form of creating content. So I definitely want to get into the light graph and the attention graph shortly, but maybe we'll back up and talk about what. Why you should care about, like, buttons. So can you spell it out? Like, what is the benefit to the user of clicking, like, on something?

Yeah, that's a good question. Because we have to ask ourselves, with something that's now clicked, we estimate more than 7 billion times a day. We estimate there are more than 7 billion reactions to web content on apps and online. You know, in the browser, that's a tremendous amount of activity. It might be that this feature is the most successful user interface element of all time. And so that leads us to ask, why is it so successful? What is it telling us about ourselves and about our needs to make it so successful? Because there was never a manual, no one ever had to read how to react to a piece of content. None of you have checked the instructions for how to use this. It's incredibly intuitive, and it's something that is so widely used. And there are a couple of elements. We talk about this at length in the book. We interview a number of prominent neuroscientists and psychologists as well to inform our perspective on this. And two main things we took away is that it is a powerful way to connect with the dopamine response. There is a sort of anticipation element of, will, I get likes for this? And then there's a reward element that you do get likes. Similarly, when you give likes, there's a hit involved in that as well. So much has been said about dopamine and social media in general, and certainly liking ties into that. The other is around social. The fact that humans are fundamentally wired for social learning. We learn from those around us. We want to surround ourselves with people we can learn from, and we have a craving for a sort of mild hierarchy. As humans, we don't have such attraction to sort of stark hierarchy. We like a certain amount of freedom and space. But we are drawn to what we describe as sort of mild hierarchy. We want to be around people that we think are slightly cooler than us and slightly smarter than us or slightly more attractive than us. So we have that interest. So we need ways to kind of level up with those around us. And social media provides an effective way to do that. And liking is a sort of currency of interaction in that type of what we describe as sort of learning sociality. So those are a couple of things that humans are very, very hardwired for. That people are hardwired for. And the like button seems to tap into both of those in a pretty smooth and elemental way. Okay.

Yeah, I suppose it makes sense that something that's going to be successful has to kind of align with people's psyches in general or cultural preferences. And I suppose there's a politeness angle here as well. It's like someone creates some content, then just click. Like is a very simple way to say thank you.

Yeah, exactly. There's certainly a reciprocation or a gratitude mechanism at work as well.

So that's kind of the. One of the mechanisms for, like, people creating content and enjoying the implications of like buttons there and also for people who are consuming content. I'm curious, from a business point of view, like, why would you want a like button feature in your product? Like, how are you going to benefit from that?

Yeah. One of the things that we realized as we researched the book is that the customer experience industry, which has been very successful for businesses over the past 20 years and successful for consumers as well, customer experience has led to no doubt a rise in the experience that we have interacting with brands and with services. However it gets measured, there are many studies on this that measure over time how satisfied consumers are. It certainly seems to be the case that the customer experience industry has been a huge contributor to the success and rise of that. If you think about one of the cornerstones that customer experience is built upon is on feedback. And customer experience has. Has made it much easier for any consumer to give feedback to a point where it's almost overwhelming these days. Right. You go for a coffee and you get an email from the cafe afterwards saying, could you give us a rating because we comment on our staff. You were served by this person. How did they do? It's kind of getting silly how much of this kind of interaction we have.

I don't think they spat in my coffee. It was probably fine.

I can see your review, Richie now. But if you look at the timing when this rise happened, the popularity of the like button slightly preceded the success and popularity of these features. And we argue that the like button paved the way and conditioned people to get used to reacting to content. And that made it much more likely that we would react to an email or a text or a call. These ways that companies can now reach out and say, how was that experience? And so if we get used to and conditioned to reacting to content to our friends, they were much more likely to be willing to do that with companies. And so we don't think it's been observed anywhere before the book. But we suspect that this feature is one of the things that made the success of the customer experience industry possible in the last 10 or 20 years.

That's really interesting. From speaking to product managers, they're always desperate for feedback on how good is their product, what needs changing, all that kind of stuff. And it's really difficult to just get people to give you useful information. So that's interesting that you're trying to condition people to give feedback on your product and therefore you can make your products better.

Yeah, and I don't think anyone planned that, by the way. I just think that's just how it panned out. Again, it's the customer experience company's probably reacting to the fact that people are more prone and willing to react. And so why not gather that data and find ways to make it simpler for customers to react?

All right, wonderful. I'd love to get into the analytics a little bit. So is the simpler thing, is it just about counting likes on things or what's the easiest way of analyzing, like, data?

Well, I founded a company called Quid, which is analyzing social media data, media data and other forms of public content. Without personally identifiable information, we don't manage it at an individual level. We're looking at. We help brands and companies find trends that are important that they can then act on. So for the type of analysis that we do and others in our industry do that, we look at a combination of the content itself. So we use natural language processing and now forms of large language models that allow us to read the content because we're dealing with things at scale. Right. So if someone reacts to a trend like flared genes are hot at the moment, I'm told. So companies, whether the retailers or in the fashion brands, are reacting to that. Right. And trying to fulfill that customer need and interest in flared jeans. So what they need to know is some demographics, but they also need to know, like, why, what is it about the flared jeans, what other ideas are being related, what other trends are connected to that, who are the influences in that trend? And all of that is going to come from analysis of the text in the posts themselves. In social media posts involving flared genes, what the like data does is it allows an understanding of the magnitude and level of interest in these, whereas other types of data is not visible. The attention, for example, is only going to be known by the company that's providing that page. Whereas liking is actually a form of public content that can be utilized by companies such as Quid to understand and quantify These trends.

Oh, wow. So is it basically like this? It can then feed into I guess like your, your supply chain, like working out like how much stock you're creating. So I said, yeah, this is going to forecast planning and yeah, are there any other sort of related business use cases here?

Certainly, as you mentioned, some retailers are able and do utilize social media content as a way of informing demand. And as algorithms become more sophisticated and more predictive, the window in which they can look ahead becomes longer. So there was a period where we could tell you kind of what's happening right now through the analytics. And this isn't just light data, this is number of posts, we call it post volume, but number of posts on the topic, which platforms, how people are sharing, interacting. There's many metrics that can be used. But it is interesting for those of you interested in data analysis that there was a time when you could kind of look backwards at trends and then we got to a point where you could look at what's happening now, like real time analysis. This is what's happening now. We're now moved into a territory, thanks to large language models where we can look ahead. And that means that several months before the summer we can start to provide an understanding of which trends are going to be hot. In fashion, for example, this summer we look ahead to big events like Easter. You might be wondering like which children toys are going to be big this Easter. And if you know two or three months in advance, that is enough time if you're a reseller, to manage your stocks and inventory and even set up campaigns and other assets to be able to serve customers interest in that trend and give yourself a competitive advantage.

Absolutely. The soles are incredibly useful, especially if you've got physical products that take a bit of time to create and ship and all that kind of stuff.

Exactly.

Okay, so you mentioned the idea of a light graph as well. So this sounds like it's maybe the next level up in terms of analytics difficulty. So talk us through, how is the light graph generated and what can you use it for?

So any company that has this data has the opportunity. Some, some use it, some don't. I mean, if you, if you're not a social media platform, then you probably wouldn't get so much value from creating a graph out of it. But if you have a platform where users are communicating with each other and you can understand that two people like the same content and you can draw a line between them and ultimately a graph gets built based on interests and that interest data can then be used to inform what to show a user. For example, if enough people have liked the same piece of content and you like other things that they like, then there's a good chance you'll like the thing that they've liked that you haven't seen yet. So that can be applied in shopping environments, it can be applied in brands to show which other items within the catalog that you might like. And then of course, it is used within social media platforms already to help guide what content to show that person next. One of the things that we realized this came out in an interview I conducted with Max Levchin, One of the PayPal co founders is he pointed out that Meta has an incredible and he didn't say this because he's involved with Meta, but rather just as an outsider looking in, he made the observation that Meta must have this incredible wealth of like data and this massive graph of like data between content and reactions to content. And in a generative AI world, this could be a real advantage for Meta, that is proprietary data that can't easily be collected or scraped from the outside like as text data can and indeed has been by other AI companies. Meta has this graph of how people have reacted to text and content. And yeah, Max suggested the fact that this might be one of the most valuable assets that Meta in fact has in the race for more effective generative AI and chat applications.

Wow. So this is incredible. So it sounds like actually based on this light graph, there are going to be all these other, I guess Richie clones or people with the same interest as me, like the same things, but I just don't know about them at all. Is that kind of the idea?

Yeah, absolutely. And that's how you provided content. If you were to join TikTok right now and start swiping, it will quickly tune into the fact that you left your attention on that item for longer. So it knows, okay, that's a signal, right? It's almost like a thumbs up. But you didn't have to even click anything. You just spent more time on that video and then the application is capturing that like and or rather that attention and it's feeding it into a model that says, well, what did people that like this also like? Where did they spend time? So from the moment you open the application, the moment you first pause and spend time looking at a video, it's already building an understanding of your interests. But based on hundreds of millions of people who are on those applications viewing content, and this graph is just in real time getting built and informing. And of course this can all be Very simply, ab tested in terms of the effectiveness of those models. And those models just get stronger and stronger, which is why, you know, attention is such a powerful way of building these graphs. And this is ultimately how these graphs are used. I just saw in the news that OpenAI may be building their own social network. And the speculation is that if they do so, they will be using their models to create the most compelling content possible. And then my assumption would be that they will tuny algorithms in a similar way that TikTok has, which is based on the tension, because now they'd be the only two things you need to build a social network. Originally, you know, in web 2.0, you had to build a graph of people. People had to invite each other to the app to know that they're related and connected. And Facebook was built out that way. But what TikTok has proven out is that provided you have a great source of content and you can get eyeballs on it, and people are willing to then share that content with others, then it's quite feasible that you could build a huge social network. But what would be your content advantage? Like, right now, like pre LLMs, if you wanted to build another Instagram, like, how would you do it? They're really hard to displace. Once that network is there and the attention is there, it's very, very hard to create another Instagram. Which is why there's so much attention from like a, an antitrust standpoint on these companies. Because, like, why is it so hard to display something like Instagram? Well, the content is naturally going to go there. If you put out content, you put out a snippet about this podcast, you're not going to take your time to put it somewhere where there aren't eyeballs. You will naturally put that content, a snippet from this podcast, in one of the platforms where there are eyeballs. And so if you're a competitor, how do you ever attract that content in the first place? But LLMs can create content. They can bypass that altogether. They can just take all the content ever created. They can create new, compelling content. They can hone that model based on the interactions with that content. So the model gets even better at creating compelling content. And if content's compelling, you want to share it with people, well, then you've got the virality coming from sharing. So anyone with a, with a powerful and distinctive large language model right now that's good at handling images has the potential to create new social networks. So we'll see. We'll see if OpenAI do this, we'll see if they succeed at it. But it's certainly a very interesting time as we see this intersection between the power and value of social networks and large language models, which Elon Musk has clearly demonstrated in the acquisition of X by Xai. I mean, Twitter's just been acquired twice in a matter of a few years, right? It was acquired by. By Elon and from being a public company to being owned by and controlled by Elon and now it's been acquired by Xai, an AI company. Well, why, why is there such synergy between social media and social networks and large language models? It's because of the data and it's because they provide a rich environment for people to interact. LinkedIn is another amazing asset which Microsoft have, and I think LinkedIn is going to provide an interesting piece in the chess game between these companies too, because it's the place that people are interacting, connecting around content and a particularly refined sort of content as well. Wow.

Lots to unpack that. So I think first of all, the idea of this attention graph seems like from a user interface point of view, that's genius because the user doesn't have to do anything at all and is still providing like signal to the companies on what customers like or not. In terms of the social media network, OpenAI creating a social network. So I just assumed it was because a lot of the sort of the data that is needed to create a large language model, it comes from X, from Reddit, from all these social networks. And actually that's kind of the secret sauce that meta and the llama model and XAI with Grok have and OpenAI just wanted to sort of hedge against that. But the idea that they can just create a load of content, attract people to it, and you're not even having to get new kind of tweets or images from users, you just using the like data from that. That's also very fascinating, like a very interesting take there. Okay, all right. So it seems like this stuff is incredibly valuable and I almost feel like the sort of network analysis like graph data is one of those underappreciated areas of data science. Can you talk me through more generally? Like, who needs to care about network data? Why do you want this?

I'll pull back a level and talk about why relational information is so important and valuable or perhaps illustrate it for people. So if you look at the last three major waves of value creation in tech, but certainly in the Internet, you have search, exemplified by Google, you have social media, exemplified by Facebook, and you have Large language models exemplified by ChatGPT and OpenAI. So what has led to the creation of three trillion dollar opportunities? I mean, I remember in the early 2000s, the Holy Grail was to create a company worth a billion dollars. Something that level of value is what people aspired to. Well, in these three cases, Google, Facebook, and I think soon to be OpenAI are at trillion dollar levels of value creation. That's staggering. And there is one simple thing that connects all three of them, and that is they each pioneered a new way to relate information. They each found a new way to create a graph and relate information. So with Google, it was PageRank, which at the time was simply. It's become more sophisticated as other layers of metrics got added onto it. But for the cornerstone of it is still, and was at the time, to look at which web pages are pointing at each other through links or hyperlinks. And if a page has 20 hyperlinks pointing at it and it has content about a certain subject, well, we can weight that more strongly than a page that only has one link or has no links pointing at it. And so it was a very simple way of just graphing the content of the early web and being able to rank content so when you search, you don't have to trawl through a million search results. The most important ones are at the top. And so that was a simple innovation that ultimately unlocked search and creation of a trillion dollar opportunity. With Facebook, the graph is the social network. It was knowing that, it was knowing who knew who, that was it. It was just that data was always out there in the world, but no one had ever captured it before. And Facebook built a mechanism and scaled a mechanism to effectively capture the relationships between hundreds of millions of people. And then with large language models, the underlying innovation is just a way of relating content. And by putting content into these models, we can now see how they're related. And this doesn't just apply to text, it applies to images or indeed any kind of information. So it's incredibly scalable across all types of information, which is why it's so important scientifically in another way. So I encourage people who are interested in data to think a lot about relational information, because it's clearly a key unlock in each massive wave of value creation. And that just goes to illustrate how much power there is in relating information. Understanding how it's related is critical to making the most of whatever data you have.

And it sounds like there's going to be a lot of businesses now sitting on untapped sort of Piles of relational data and nothing's happening with it. Do you have any advice for, like, any businesses, like, what should you be doing to, like, even discover that this exists or to get some value out of it?

Yeah, any, any data scientist is going to be all over it in terms of what models and approaches could be used. So if you're listening to this and you're on the business side and you're thinking about the data that your business is accumulating, then, yeah, highly recommend reaching out to your friendly data scientist if you have one in, in the company, or indeed bringing in a consultant and asking them to explore your data. Because often hidden business models that are even available for some companies, if they are attracting and acquiring information in unique ways, then there can be new business opportunities to serve customers you may not have even thought about in the past.

Yeah, certainly it seems like from the things you've been talking about so far, that, like, social network data of, like, maybe existing customers, that might be a good place to start. Or even just like. Yeah, if you've got, like, buttons built into your product, you probably want to be looking at what's going on there and it should bring you something useful. All right, so I'd love to kind of go back to where we started about the like button as a user interface feature. Now there are quite a few other sort of related ideas. So you've got things like, as well, just thumbs up. You've got the idea of a thumbs down button as well. You've got things like review stars. Can you talk me through how do those other user interface ideas relate back to the like button?

Yeah, it makes me. You mentioned five stars. It makes me think about the early days of Yelp because we had to choose what the scale would be to rate a business. And again, that wasn't obvious. And five stars has become kind of ubiquitous on the web. And certainly there were sites out there that had already used five stars, but there was nowhere near as much consensus as there is now around that convention. You'd see numerical rate 3 stars, you'd see 10 stars, you'd see various kinds of pluses, minuses. You can plus one on something, you can plus two on something in some environments. So there were a significant number of ways of rating content. And five stars has sort of emerged to be one of the most popular ways of doing that. And Yelp, that's what we went with from the start, but we had to consider lots of different options and it was a combination. I think it was attractive to Yelp at the Time, it was mainly our CEO and founder. I seem to remember he was sort of leading the way in thinking through what Yelp should use. But I remember it was this, what was discussed as this combination of simplicity, but with enough granularity, like three stars did not quite enough granularity to distinguish between, say, excellent and good. In that scenario, two stars is mediocre, three stars could be excellent, or it could be just good. Whereas five gives you like, five is exceptional, four is very good, three is okay. So it still allows for that kind of sort of segregation, I suppose. And then stars were also attractive to us because it was a convention that was popularized by Michelin guides. So you have a one star Michelin restaurant and a two star Michelin restaurant and so on. So the idea of a star being attached to a really quality business, the way had been paved by the Michelin guides and by their restaurant rating system. And so I credit the Michelin guides a little bit to giving a little prestige to stars. And I actually saw an academic paper that came out on this topic just a few weeks ago, and it was fascinating. It was in. I don't know why I ended up reading this, but I think it was in nature. But it was a paper that had. Where they had studied, they were assessing different ways to rate employees and seeing whether that brought racial bias into the employee rating process. So that was their question of the study, is that if HR teams use 5 stars versus a 1 to 10 rating scale, they were the two key ways they were comparing 5 stars rating system versus 1 to 10 scale. What they found is that in an HR context, if you rate people from 1 to 10, it actually brings in significantly more racial bias than if you use five stars.

That's completely wild. I have to say it's not very intuitive either because you think, well, equivalently, these are all just like arbitrary scales from like lowest to highest, and they should give the same effect. But the fact that one system introduces racial bias and the other one doesn't, that's very unexpected.

Yeah, it's a fascinating study. And it just goes to show that different rating systems condition users to. To behave differently. If you're interested in user interface, you have to get very interested in the mechanism you're using to collect responses, because it turns out it makes a big difference. This paper is just one example of that fact. And in a, in a different context, in an HR context. But also, if you listen to this and you're in HR and you're into data, please look up that study and please make sure you use five star rating systems on employees and not one to ten.

Yeah, this is very interesting. And even the idea of the Michelin stars, even though it's a star system, that's very different from your consumer review one to five stars because their one star means absolutely amazing. And then three star is unheard of best in the world, whereas the one to five stars is like, well, four stars just means average. So, yeah, out of subtle changes in the universe interface, it's going to have very, very different implications. Okay, so for anyone who's interested in user interface design, do you have any advice on how you make sure that you're doing it well and all these kind of subtle cultural issues are being taken care of for rating systems?

I mean, I suppose the number one thing in UI is, and anyone that does UI knows this and has probably been trained in it, but it really is this simple. It's use the convention that is most familiar to your users. That's it. You never want to try and invent new elements or design new things. I know that's ironic coming from me because I have created a few new user interface conventions over the years, including our company. Quid was the first to put network exploration visualizations into the web browser. So there were applications in chemistry and physics that allowed you to map information, protein graphs, for example. There was an application called Cytoscape, which I don't know if it's still around, but back in the day that's what was used by scientists and. But it was a desktop application. You could visualize a thousand nodes on a graph and navigate them visually. But it was a desktop application and no one had ever been able to make those kind of graphics work in a browser. And Quid, we were the first to do that and we made it work with 3000 nodes to interact in a browser. So there have been some conventions that I've, I've pushed on. But even when creating new things, like an element interactive element that you, that has never existed before, you always have to start with what are the users already using? That is as close to this as it exists. So you look to the applications that are most popular, that are most popular with your users, and you go and you look around the interface and you say, how is that happening? So Excel is an example of an extremely popular data interface. So before you do anything in terms of data, you kind of want to look at Excel and say, what do their dropdowns look like? What do their sort options look like? What do the bars look like at the top of the page and wherever possible you want to sort of mirror those because that's what people are used to. So when they look at your interface you just, you want to make it as intuitive, possible. The mantra is don't make me think. If a user has to pause and think what will it do if I click that? Then you've messed it up. You just want to take all that friction out of it for people. And in web 2.0 that's what everyone was obsessing about. There was a book called Don't Make Me Think which became a sort of bible for user interface designers on the web. But that was the simple idea underneath it. Like always look for the closest possible pre existing paradigm and then use that if you can.

Yeah, Don't Make Me Think is a great book. I remember reading that earlier in my career, but also I have to say I have used Cytoscape back in an earlier phase of my career.

Serious props.

Yeah, I was looking at protein networks with that and yeah, that's one of the challenges with visualizing graphs is like when you've got a small graph looks brilliant. As soon as you start having hundreds or thousands of nodes, it's a total mess and visualizing it becomes a real challenger. So yeah, using Cytoscape for that.

If you've ever listened to, if you've ever used Cytoscape, put it in the comments. Tag me and Richie because we want to hear from you.

It's not widely used, probably like two, maybe three other people. Exactly, yeah.

But if you're listening, you know it's you we're talking to, so get in touch. Wonderful.

All right, so do you have any final advice for people who are interested in working on like data and analyzing it?

There are a couple of paths to it. Right. If you have some within your organization because you work at a retailer and you've got a like feature within the retail environment, I think that would be very interesting to explore and to see if you can then use that to inform what shows up next or what related products you can give to to the users. If you are in a social media environment, you're probably already making very good use of it. And if you're just a business that's trying to figure out how can we make better use of social media data in our day to day operations we see a lot of companies right now trying to be more in tune with what's trending. Whether you're a brand, whether you're a service, whether you're a retailer. Social media is only becoming more and More important, you can see that from the growth stats, from the usage stats, and it's the place where people are going there to search for new ideas, new products, and it's a huge source of influence, but it's also an amazing source of data to understand what people want, what they need, what they think and feel about the world around them. At quidd, we call that customer context because it's about understanding what's the context that your customers are operating in. And companies that can more quickly adjust and operationalize around these trends do have a competitive advantage. So you could learn more about Quid, or there are obviously other companies in our space that you can read more about and connect with to adjust to the trends as they happen.

Yeah, I think like everyone's aware that social media is everywhere, and I'm sure every company has a strategy for posting on social media and then maybe doing bits of things, but I feel like not very many companies are tapping into the full potential of what social data can do to help them with their own business.

Definitely. And I'll just add to that that even those of you that have sort of explored social data maybe 10 years ago, five years ago, what was possible then is completely different to what's possible in a world of large language models, because the insights that you can now draw from social media content, and media content and other forms of public content are far deeper, far more meaningful, and far more cost efficient to access than they were before. So if it's something you looked at in the past and maybe got bored of it or thought there wasn't enough value, highly recommend people get interested in this topic again. Because in the era of LLMs, a whole new swathe of things are possible and can be set up and automated for the business. So, yeah, it's a pretty exciting time to be in data and data analysis in general. And certainly there's a lot of new potential with social media data.

Certainly exciting times right there with you on that. All right, finally. So I always want new follower recommendations. So can you tell me whose work are you most excited about at the moment?

There's a book that I'm reading right now that I absolutely love. It did come out a few years ago, but I only recently recommended it. It's called Range, and the subtitle is why Generalists Thrive in a World of Specialists. And I love this book because it does deliver on what it says in the title. It is explaining and advocating for why a diverse set of interests and following your passion and curiosity across your lifetime creates meaningful advantage over time. But it's more than that. It actually offers a kind of framework for great ways to think and for great skills to develop in a world of AI. It's not trying to do that, it was trying to do something else. But I think it's a very important book to read when it comes to sort of plotting your career development. Many people right now are thinking about what does AI mean for me, what does it mean for my job, my industry? And rightly so, I think we're all thinking about that. And Range is perhaps a slightly lesser known book for contributing to your thinking about that and thinking about thought patterns and things you can develop that will be more robust and more long lasting than some more specialized skills.

That's very interesting. And just from your description of it, it sounds like I can now justify my complete lack of career direction. Going all over the place and it's like, oh, I've got more range. That's wonderful. Maybe that has benefited me. So yeah, maybe don't over plan your career re range and go where the wind blows with your career. Wonderful. All right, thank you so much for your time, Bob.

Thanks for having me on. Richie, great to see you.