🚀 Analytics - Launch Day @Birthday Week

Hello and welcome to this session of Cloudflare TV. My name is Vivek and I'm with the product marketing team here at Cloudflare. I'm joined today by Jon and Jamie. Jon and Jamie do you want to introduce yourselves real quick? Sure. Hi everyone. I'm Jon Levine. I'm product manager at Cloudflare for data and for analytics. Hi. My name is Jamie Herre. I'm a director of engineering here at Cloudflare and one of the teams I work with is the data team that worked on some of these analytics. Yeah. All right. So let's let's talk about analytics as our viewers might be aware. This week is birthday week at Cloudflare. Ten years back is when Matthew and Michelle, our founders presented the idea of what is now Cloudflare at TechCrunch. Every year since we've been celebrating Cloudflare's birthday in the last week of September. And this week marks Cloudflare's one entire decade of existence. During this week every year Cloudflare gives away products and services that benefit the Internet as a whole. A lot of times for free. So Jon, what have we announced today? What's our gift to the Internet? Yeah. So today we are announcing a huge upgrade, really like a sea change to how we offer analytics to customers. So we are going to be giving customers a whole heap of new information about how visitors are interacting with their websites. And there's going to be a big upgrade today available for customers in the Cloudflare dashboard. And we're really excited that we're going to be making our analytics available even to folks who don't use Cloudflare's kind of traditional proxy service, changing their name servers. Anyone on the web will soon be able to just add a little piece of JavaScript onto their website and be able to get free privacy-first analytics. Web analytics for everyone. So let's actually, even before we dive into the nitty-gritties of the product, what it means, why it's so path -breaking, why don't you walk us through a demo? I think a demo of the product itself is probably worth a million words. And while Jon's bringing up his screen, Jamie, you built this product. Why did you decide to work on this project? And why was this important to you? Well, first of all, let me say that I get to sit here and represent a really great team of people at Cloudflare. I work with some really exceptional people. So the truth is, I didn't build anything. A fantastic team put this together. I think the reason why it's important is that we've always felt that we could do more in presenting the analytics to people in a way that was impactful and helped them make decisions. Is their site working? How many people are visiting? We've had this data. We just haven't been able to present it in a clear and concise way. The other reason is that we operate at a very big scale. So Cloudflare is growing and we handle a huge amount of traffic. And so periodically because of that, we always have to reinvent and re-engineer how we do things in order to handle the ever -expanding nature of the Internet. So we did that and we realized after we had done it that we could actually greatly improve our analytics in terms of the kinds of metrics we expose and the interactivity of the dashboard, which Jon will demonstrate. Yeah. Let me switch over to presenting really quick. Cool. So I've gotten permission to share stats about Cloudflare's blog, which I'm really excited about because it's so much more fun to look at a demo using real data, not big data. So this is our homepage of our dashboard. It should be very familiar to customers of Cloudflare. People find this really useful. I think that was really the inspiration for wanting to build out a web analytics product. People told us that they really relied on this data that is very trustworthy because it's collected at our edge. We see every request that goes through our network and we see all the traffic and we report on it. We've made some big improvements in the firewall tab and the cache tab. If folks haven't gone there recently, I definitely encourage you to check out the analytics there. But it's been a little while since we've updated our analytics tab. So this is what it looked like before today, before this launch. And we showed you an overview of requests, cache ratios, bandwidth or data transfer, which is really useful information for understanding technical metrics about your site. But it didn't really answer how many people are coming to the website and how many page views are there. And when we realized from talking to customers, they were really relying on these metrics and they might be confused because a request is a page view. And also just like Jamie said, we're so close. We actually have all the data that they needed to get core web analytics experience, but we weren't showing that. So let's turn it on. Anyone on a pro business or enterprise plan can click that banner and turn it on. And this is what you see now. So we still have requests, we still have data transfer, but now we're really emphasizing visits and page views. And if folks haven't seen kind of our analytics UI, it's really cool. We have this big time series visualization at the top. You can zoom out for now to the last week. I'll just show you like the last three days. And what I think is so cool about this, two features I really love. One is this idea of group by. So you can change, you can kind of pivot this data and break down by, for example, where are the top countries where people are visiting from. And now you see those countries each in their own line, which is really neat. Oh yeah, sure. Sorry to interrupt. What does a visit actually mean? Are these the number of unique visitors or is it different than that? Yeah, we'll talk more about this because being privacy first is a big part of what we want to do. And there are ways that people traditionally count visits using or unique visitors, for example, using unique IP addresses, counting unique cookies. We didn't want to do any of that. So we kind of have this somewhat novel definition of a visit, which is basically, it's a page view where the referrer does not match the host name of the request. So in other words, it's a time when someone loaded a webpage, but they didn't come from the site that you're on today. So we can actually look at referrers, right? We see that people are coming directly to the site because they're typing it in from bookmark or email link or something. They're coming from Google. They're coming from some other Cloudflare sites, but none of these are coming from blog.Cloudflare .com itself. Got it. Cool. So we just look at paths, we can see the top blog posts. Okay, look at this. So there's a big, you can see Matthew's post about welcome to birthday week was really popular. And then my other favorite feature is we can actually click and drag to zoom in and actually see a smaller time range really easily. And the graph just automatically updates. And it's all super fast. So if I go down, I can scroll, I can look at various top ends. I can see where people are coming from, countries, what browsers they're using. And filtering is just so fast. So if I only want to filter traffic, say in the US, click on that. Everything that you see here just updates right away. I can remove this filter. So yeah, really, really excited about this. Yeah, it's really cool that you can actually slice and dice data based on the filters and exclusions that the different dimensions can do. So this is available for if I'm a Cloudflare customer. Yeah, I'm on the pro plan. Can I go and enable this right now? Yes. So analytics powered by our edge logs are available right now for customers on free pro business plans. And what we're working on is making an experience like this available for anyone just by adding a JavaScript snippet to their site. Ah, so in that case, I wouldn't have to be I wouldn't have to what we call internally Orange Cloud, my website, I wouldn't have to be a Cloudflare proxy customer. Exactly. I don't have to change the DNS settings on my website. That's right. So Matthew, our CEO likes to remind us birthday week is when we give a gift to the Internet. And so our gift to Internet is that we want to make this this available to anyone. And we know that there's a high bar to Orange Cloud and proxy website. And of course, we'd love to have you do that. But we know it's not possible for everyone. Let's dig into that a little more. And maybe Jamie, you can explain this to me. So what John's saying is that because of the vantage point that Cloudflare has in the end to end network, today, we look at all the requests that could go to your website. So we're automatically collecting those visits and page views on like our server logs, which is great. But there's also this other option, which we're going to make available to everyone, where you would inject a JavaScript beacon on your website, and start collecting those metrics without having to make any changes on your DNS settings. How would the product itself be different? Would I be seeing the same? How would the dashboard be? How should I as a user if I have a website that's on Cloudflare? Should I should I go and inject the JavaScript beacon or should I just get what I'm getting from Cloudflare now? So in the second case, you have control over where you put the beacon and why. You don't have to have what we call an orange clouded zone. So you wouldn't have to be a full DNS and cache customer. It'll collect slightly different information. And one of the advantages of using the beacon is that we will also get some performance information about the webpage. So this is often called like real user metrics. And our plan is to incorporate some of that into the analytics that you'll see. So there's a benefit in that case in that direction. In the other direction, you have the benefit that you don't have to configure that beacon yourself. And you get some additional performance and privacy because there is no beacon. We develop the analytics from the usage directly. So there are advantages in both directions. Our plan is to have a unified experience. So you'll see both in the same place, but they have some minor differences. And last year, at Birthday Week, if my memory serves me right, we also announced another browser product, Browser Insights, right? Speaking of the real user measurement that you talked about, it tells you the end user performance, how your website's performing from the client's perspective. Yeah, that's right. So we have that product today, and it's very successful. It's more specialized than what we call regular web analytics. So it's more of a specialized audience. So we're kind of expanding that to cover these other regular analytics cases. What's interesting about that data is you want to know, it's not just about I always think that the first thing anybody wants to know about their web property is like, is everything working? Is it functioning properly? Did I do the right things? Maybe I'm not a super expert on all of these technical issues. And the Browser Insights is helpful just in that case. So we're not just talking about the top percent who are trying to optimize some super special thing. It's for everybody, because you might want to see, is my website acting typically, normally, compared to other similar websites that use similar browsers? Is my site working in the same way? And because we have a big reach at Cloudflare, we can help people try to understand that by saying what is typical? What should your target be? How do these different parts of the site work in combination? So we're hoping to help people understand and act on these things without being intrusive. Yeah. And getting those insights on a country by country basis, seeing that your website performs very fast, maybe in the US and maybe not so fast in somewhere in Asia, is actually powerful information to know. And it also helps us guide our customers on, oh, maybe you use Argo Smart Routing. This can actually make your website faster. So I can see how that could be very helpful to our customers. Totally. I probably should do a quick demo, because I should add that today we did also launch a big update to Browser Insights. We have a whole segment on it later this afternoon, but I should give a quick preview of it now. Please do. So I'll just switch over. So over in the Speed tab is where you can find Browser Insights. And what we launched today are access to three core Web Vitals metrics. So just to take a step back, before in Browser Insights, we showed what we think of as the navigation timing metrics. So this is basically how long did it take for components of the page to load. So how long did the DNS request take, the TCP connection, the request phase, getting a response back, processing and loading that page. And that's really helpful. It's a little bit nerdy. It's kind of in the weeds. It doesn't always give like a full picture of how are, like Jamie said, like, how are users experiencing my website? Like, is it fast for them? And so we partnered with Google on this, they've done a really nice job of coming up with these three metrics, which in this afternoon, we'll be going into more detail about what each of these are, but they capture sort of high level user experience. So this largest contentful page, basically how long did it take for the web page to load, actually break it down into these three categories, good needs improvement and poor, we can just tell you kind of how many people fall into each. And then what's useful is you can actually see how different URLs do or how different browsers perform, how experience varies across countries. Something else we're really proud of in our web analytics and something that we're going to be excited to bring to the whole web, once this is available as a standalone snippet of JavaScript that you can include in your website. This is great. Is this the new web standard that Google introduced a few months back? Yeah, so the idea here is that we're trying to get folks on the same page in terms of how they measure performance. And in performances, it's pretty broad, it's right, it's not just loading speed. But it's also things like responsiveness. So if you click on something, how long does it take for the web page to respond? And that's the first input delay. And then there's cumulative layout shift, which I like to think of as like jankiness, but it's like how much did the page move around? While it's loading? I'm sure many people have loaded a website and you went to scroll or click on something and it moved and that's frustrating. And so I think they do a nice job of kind of capturing these different aspects of experience, user experience with these metrics. Wasn't AMP supposed to also be doing this? Yes. So Google has kind of announced, they've peeled back the curtains a little bit on their plans for the search algorithm. And so AMP is Google's way of kind of badging and search results that a website was fast if it conformed to certain, basically coding standards of the website. And the idea with Web Vitals is say, okay, rather than have a standard that you need to code to, like your HTML has to conform to this, they've said, if your website actually meets these performance criteria, then we can badge it and say that it's fast and give it that badge and that boost in search results, which is really neat. Yeah. So Google makes, they have a public report. They collect anonymized data from Chrome users, and that's available through the Chrome user experience report, which is interesting to check out. What's great about our browser insights and Web Vitals collection is that it's free for everyone and that the data updates right away. As you make a change to your website, you can immediately go and check out what's happened, look at the last day, last three days, whatever you want to see. Whereas the Chrome report only reports a rolling 20 day average. So if you made a change, it might take like two weeks for that to affect the average. Makes sense. It's actually really nice to see our analytics products grow by leaps and bounds to Cloudflare over the last year or so. So thank you for that, John and Jamie. Well, that's awesome. Can we give a shout out to the browser insights team? That team has also done a fantastic job and you'll see a trend in this discussion that I think we're realizing that a lot of, from an analytics point of view, a lot of the Cloudflare products all come together. As we iterate on this, we'll see more different products coming into the same views because, and the views you'll notice also start to look similar because we're using the same backend infrastructure to do these different things. So over time, I think we can do a lot for the customers by bringing these things together to give them actionable insights on things that you could change about your site or learn about the end user's behavior. So Jamie, that's such a great transition. You mentioned the common visual design, and we could do a whole segment on how we use visual components at Cloudflare. It's kind of amazing to me. But there's also some really cool architectural decisions we made on the backend that's allowed us to build these free analytics products that have like, you know, you can have a gigantic website, like many gigantic websites use us and it just works. So how does that work? Tell us about the technology behind all this. Well, what we're talking about here is this technology that we call ABR, and I won't go into all the technical details of this in this discussion because I could probably talk about it for hours. But the general idea is that we realized that the typical thing about usage at Cloudflare is that it's varied. So this seems to apply to everything that we do, and maybe it's true about the Internet in general. So that a busy site is orders of magnitude busier than a not busy site. Or a site that's under attack is getting many orders of magnitude more traffic than the same site a minute earlier when it wasn't under attack. So there's a lot of variance between things. And if you graph that, it almost always looks like what's called a Pareto distribution. In other words, like a really steep drop off. So what's busy right now, compared to what's busy an hour from now, like it could be a huge difference. And this has always posed problems for our analytics. Because building software that can handle these huge spikes in traffic versus like the steady state, it turns out to be quite difficult. And there aren't that many choices. But we came upon a pretty simple solution, which is just to provide the same information at different resolution. So we collect and record the same information at a bunch of different sampling rates. They're random samples, which have a bunch of great properties about how we can actually extrapolate metrics from those samples. And this works extremely well. So no matter how attacked your site is, your analytics will give you a useful graph. And as we extend things over time, no matter how busy your site is compared to where it was last week, or how busy one of your sites is compared to another, it will give you a useful graph. And this has unlocked a bunch of features. Because we used to have to spend quite a bit of time to program each one of these metrics collectors. And now we don't. It's kind of amazing, because we used to get all these feature requests, like, I want to see my analytics by host name, or like, I want to see my analytics by, I want to see the origin response status, or I want people to ask for these things, and we'd scratch our head, like, oh, can we do that? And now the answer is just, yeah, sure. Because we can just add this, and it's easy. We get all this dimensionality and all this cardinality really for free, which is just kind of incredible. And it's how we felt confident to be able to release something like these web analytics for free for non Cloudflare customers, for people who aren't on the service using the beacon, because we know we can handle the amount of data generated. Hmm. What I think is so cool about this is, like, there are other free services out there, but they all have, like, hard limits. Like, you can put in this many events and no more. We say, bring it on, because we know that we can only show, you know, three or four significant figures, right? And do that three or four significant figures accurately, whether it's, you know, 1,356 requests, or, you know, 13.56 million requests, right? We can accurately represent that in that relatively small amount of space, which is really neat. I think the answer is actually two significant figures is the right number for most numbers that you're interested in. And if you look at any dashboard on our site or any other site, you'll see, like, basically, it's always two significant digits. Giving people what really matters. I think that's also been one of the foundational elements of web analytics, right? Because one of the things we hear a lot of times from other popular analytics vendors, customers, is that they're so complicated to use, and they have all these knobs and whistles, which most people don't care about. And the other thing that people care about a lot is privacy. And I want to dig into that just a little bit. There's almost this zeitgeist, right, in the tech industry now around privacy. And Cloudflare obviously has an outsized impact on the Internet based on our scale, as you were mentioning, Jamie. What do you think Cloudflare's role is in helping push a more, quote, unquote, private Internet forward? Yeah, I can take that. I think what makes Cloudflare unique, we're a lot smaller than certain other well-known technology companies that provide free services. But what's unique is that we provide free services and our business model doesn't involve advertising in any way. Our business model transparently is to get people to buy our pro business and enterprise plans, and those customers put an enormous amount of trust in us to safeguard their data, to safeguard their end users' data. It's pretty foundational. It's certainly how we all think about the data that we do use, and it's kind of foundational to our business model. So when we started building analytics, that was kind of the first thing was we said, like, what would we want as end users? What do we think is the right thing to do as end users? And that's kind of led all of our decisions. And then of course, later into that are many regulations and rules about how you have to use data. That's important too, but it really starts with what we think is the right thing to do. Jamie, is there anything you wanted to add to that? I agree. I mean, and I think we all agree that privacy is an important issue and that we're dedicated to that. I think sometimes people jump to conclusions that aren't correct. And overall, I feel that there's legitimate trade-offs to be made. And I'm saying that as someone who uses the Internet and someone who provides services on the Internet and also a customer of Cloudflare, which I also have been, I feel like what we talked about earlier, for instance, was this idea, how is my website performing? Well, I think an important part of that would be like, compared to what? And how would you know that? And how do we collect that, right? And so we don't track people, individuals, but I think it is a legitimate use. There are legitimate uses of aggregated data that are helpful to the Internet, helpful to us, helpful to the customers and helpful to the end users in the end. And we're committed to striking the right balance with that. I think we have. Yeah. Yeah. I've worked at a few companies, obviously, before this, and I've seen a few companies actually really live their mission as much as we do, live and breathe it, which is to help build a better Internet. And it's nice to see that very profound idea being translated into real concrete products that we release and make available for everyone on the Internet. I think just to build on what Jamie said, I think the biggest way we live that is we say, there's this expression, none of your business. We don't think it's none of our business what individuals are doing. We have no interest in what an individual is doing over time. And we've actually kind of gone to great lengths in some ways to say, we don't want to know what individuals are doing. And so there's legitimate reasons to do that. We were just talking before this about bounce rates. There's times when you might want to know, hey, how many people place an item in the shopping cart and didn't check out? We don't support that yet. And if we do support that, we'll make sure that's something that our customers opt into and know that they're choosing to collect that kind of data. But we're not collecting anything like that by default. We view these requests as sort of independent actions of each other, which also allows us to do sampling, which is cool. And I think that's what's really most important. Our customers are in control and end users are in control. So for the purposes of analytics, we're not trying to create, some people are saying, oh, we don't use cookies, but we use fingerprint. We run some JavaScript in the browser and we've like identify what your browser is. And that's like the worst scenario, right? Because now end users can't even like clear cookies if they want to. So those are kind of our principles. We want to be transparent. If we were doing customers would have control and end users would see it really clearly and have that choice. Yeah. It's funny you bring up the idea of cookies because this morning after we released our product, after your blog came out this morning, one of my friends actually messaged me and said, oh, Kalf is doing this. Does that mean I don't have to display a cookie battery more on my website? I was like, oh, you probably still want to, I mean, it depends on whatever regulations are in your jurisdiction, but, but yeah, he's, he's excited that he doesn't have to be cooking as visitors. Thanks to Cloudflare, which is, which is great. All right. We're almost at the top of the hour. I want to take just about a few seconds to thank both of you for doing this. This has been a really, really great. Before we wrap up, John, do you want to tell people how to sign up for this new product that you and Jay have built? So if you were an existing Cloudflare customer, just head over to your dashboard, go to the analytics tab. There's a banner to go opt in. If you're not yet a Cloudflare customer, Cloudflare .com slash web dash analytics, Cloudflare .com slash web dash analytics. If you go there, there's a big signup button. You can get notified when our product will be available for anyone on the web to sign up. We'll be getting in touch very soon. All right. Thank you, Jamie. Thank you, John. This was a pleasure. Thank you so much for that. Bye-bye.