Name: Office Hours: Ecommerce Analytics in Action
Uploaded: 2026-08-27T17:21:43.632Z
Duration: 59 min 12 s
Description: Office Hours: Ecommerce Analytics in Action

Transcript for "Office Hours: Ecommerce Analytics in Action": Great. Alright. I think we'll give it a minute for folks to kind of trickle in, and we'll get started in in a second. Alright. Thanks everyone for joining. Really excited to have you here. So welcome. Thanks for spending the hour with us. I'm Peter Lautengard, principal customer success or customer value architect at Mixpanel. And I'm joined by my colleague, Reuben Moss, who's a senior solutions engineer. And, today's session is called, office hours, ecommerce analytics in action. Most of the hour will be real life use cases with, two ecommerce leaders, that are Mixpanel customers and walking through decisions they've made with the help of Mixpanel. After that, Reuben will showcase how to make the most of our newer advanced features like our Mixpanel agent and root cause analysis tools, as well as MCP. And then we'll close with live Q and A. So, kind of as we're getting started here, if if you're comfortable with it, it'd be great to hear in the chat, you know, where all in the audience you're coming from and and maybe what your role is as well. A few housekeeping things. One, please drop your questions in the Q and A panel as they occur to you. And then we have some pre, preceded questions that that some of you shared, during registration. And two, we are recording and you'll get the link afterwards. So if you want to share with your your colleagues or team members. And then finally, we do have some prizes to give out if you stay through Reuben's demo. We'll ask a couple trivia questions on what was presented and, have something for the first slash most accurate answers, so, stay tuned for that. And now let me intro and sorry. Here's the agenda here. And, without further ado, let me intro who you'll be hearing from today. So, Dani Smith is a senior manager of digital product at Bollin Branch, the luxury, bedding brand. She owns core commerce, cart checkout, search, product recommendations, and she's also the AI champion of her team, leading AI integration across the site. So, before Bolling Branch, she was a senior PM at Spanx and a consultant at Bain. And, Danny has a hard stop at 12:20. So we'll start with her and then get on to our next speaker. Our next speaker will be Sierra Chen, is a senior analyst on the data science and analytics team at Weil Alaskan Company, a D2C sustainably sourced seafood, company where she's been for four years. She has deep data expertise in conversion funnels, AB testing, and so feel free to ask her more questions in in our Q and A at the end. So, Danny, let's bring you on stage and start with you here. Great. So, Danny, excited to hear from you about how you've been using Mixpanel in your decision making at Bolling Branch. We have a few questions to go through, so let's just jump in. Luxury bedding is something of kind of a more infrequent purchase. So someone buys a great set of sheets, and if you did your job, they're they're not shopping again for a while. So would love to hear how does your team think about identifying and driving high value re repeat buyers when the product itself, doesn't necessarily naturally bring people back. Yeah. I think it's a great question. And to your point, luxury bedding is definitely meant to be more of a long term investment, but I think that's where our category depth comes into play. So a lot of times, sheets are the entry point for our company. It's not necessarily the end all, be all. The goal is to sell you a set of sheets, convince you on its value and its quality. And at that point, you become obsessed with the brand, hopefully, and you switch your whole bed, your whole house. You move into our products as well. So we do see it as a long term investment in the customer rather than just Yes. trying to get you to come back and buy another set of sheets. And I think where the mixed panel data and where our customer journey data really helps us with that is, again, we can have two customers that have the same exact first purchase of just a plain set of white sheets because they're trying the brand out and they wanna see what we're about. But those two customers can have completely different discovery journeys and paths before they hit the checkout flow. Whereas one might be a very utilitarian buyer who's coming in. They know exactly what they want. They they just look at the sheets, maybe browse a couple of fabric types, but then make their decision and purchase. Whereas another might explore everything that the brand has to offer, and they're browsing multiple categories. They're browsing multiple collections on our site. They're really digging into the brand as a whole before they decide to make that purchase. And, obviously, based on those signals, that helps us kind of separate out the really high intent long term value customers from the folks who maybe are gonna be more of a light touch customer long term. That's great. Yeah. Definitely makes sense. And so it sounds like multi category customer is kind of a key cohort that you that you track? Awesome. Yes. Yeah. Great. So, yeah, looking at those behavioral signals, keep an eye on those different key cohorts and build experiences for for those users. Awesome. So going to our next question, you tracked the full purchase funnel in this panel. So, kinda walk us through the last time one of those numbers actually really changed meaningfully and and what your, based on what your team shipped, and what did you see and do about about that? Yeah. So, I mean, in terms of the last time that our funnel data influenced a decision that we made on-site, probably yesterday, we're constantly looking at it and basing our decisions on that. But in terms of, like, a really tangible example, that had significant impact, I would say, maybe a month ago now, we did a pretty heavy design and functionality overhaul of all of our upsell touchpoints across the site, whether it's a product carousel at the bottom of a product page, our, product add on feature that's in kind of our standard product page that's just meant to be an easy complimentary ad product for our cart upsell. We did, again, very significant design overhauls there, and we saw really significant impacts through Mixpanel, basically. Like, Mixpanel was the source of our performance data for this design overhaul. And it was just it was really great to be able to look into the data and see exactly how much those changes were driving shifts in the product ad mix across sources. So, again, like, how many products were being purchased as a direct result of cart upsell, as a direct result of PDP upsell now versus beforehand. And that it it was just really insightful. I mean, I think we were able to drive nearly a 30% lift in average unit retail price for those placements because, again, the design overhauls unlocked new functionality. It unlocked our ability to showcase new products there that we hadn't before. Again, all of this was gonna be anecdotal data if we didn't have the mixed panel results to back it up. Got it. Makes sense. So it sounds like daily, weekly, monthly ritual, essentially. And and I'm in Mixpanel all the time. and and is that pretty, evenly mixed between insights report, funnels report, kind of our our whole suite of of tools? Yeah. I would say insights, funnel, flows, and then, obviously, we have been leveraging the cloud MCP a lot, with our enterprise cloud accounts just to help marry different data sources together across platforms as well to build an even richer picture. Awesome. Well, that is a great, transition to our next question. We we made it two questions in and didn't didn't talk about AI yet. So incredible in 2026. Okay. But let's kind of kind of get into that. So you're the AI champion of your team. What's something you've actually tried recently that you're excited about operationalizing? It sounds like you gave a little teaser of that, in your last answer. Yeah. I mean, part of our weekly ritual at Bolling Branch is we have weekly performance calls across the digital team just to understand what worked well last week, what maybe didn't, and what levers we need to pull this week. And mixed panel is a huge source of data for all of us for that reporting session. And in previous like, in our previous life, I would say we spent probably 80% of our time just pulling the same metrics over and over again from the same reports and Mixpanel, but just having to do additional data transformations on top of it or, matching it up to different sources to build the full picture. We might be spent 80% of the time just digging into the trying to pull the exact same metrics that we were pulling every week, and then we could only spend 20% of our time digging into the why and digging into, like, another layer of detail behind what those trends actually meant. And with the Claude MCP, at least for the reporting that I've been able to do, it's completely flipped where, you know, now I can only I only have to spend 20% of my time pulling those same metrics again because I have Claude fully I have Claude co workflows fully built to pull the exact same metrics and replicate my mixed panel dashboards. And I can then spend 80 percent of the time just prompting Claude, like, okay, but why am I seeing this trend here? What what other trends are we seeing? I maybe pull an additional report in Mixpanel and feed the export into Claude there to, again, just try to dig deeper into why is this happening, what does it mean tangibly for our company and for our team. Very cool. Yeah. I you you kind of answered, part of what I was gonna ask, but it sounds like you have some automation with that setup and co work. Do you does has your team built any, like, Claude skills around the mixed panel MCP or is it more, you know, individuals building their own co work setup to do that kind of thing? A lot of us right now are in the test and learn phase, and so I would say a lot of us are kind of experimenting alone and then sharing out any best practices that we find. And so, like, one example is I was using, like, some of our AB tests in our platform. Like, our testing platform, we don't necessarily get the full layer of funnel detail that we would want to dig into those results. So we also analyze AB test results in Mixpanel. Obviously, mixed panel does not have, like, a statistical significance calculator baked in. So I did build a skill in Claude that mirrors the same statistical significance calculator from our AB testing tool and just applies it to any mixed panel data. So that, again, I can try to, like, create one cohesive view of here's all of the test metrics that we had, and here's all of them kind of broken down and analyzed in the same methodology, which previously was like me copying and pasting numbers into an online stats and calculator, which is just right there. Yeah. Yeah. So that has made my life a lot easier as well. That's that's awesome. Kinda taking it from experimenting, testing, which I feel like, so many companies out there this year has, been a big bulk of the effort. And then, yeah, kind of up leveling those the integrations there and and starting to automate it with cloud skills and co work. That's that's really exciting to hear. Great. So one more question for you, and then we'll we'll get you out at your hard stop. But, Yeah. So your team, you mentioned, owns, Agentic Commerce, which is obviously as hot hot of a topic as exists right now. What do you think kinda stays constant about understanding your customer, no matter how much the interface between you and the customer changes? Yeah. That's a great question. I think at the end of the day, it's just that the the way that we think about agentic commerce at the very least is that the the tools that the customer is using are different, but the questions that they're asking are not, Like, at the end of the day, it's the same questions that they've always had about our brand and that we've always needed to answer. It's just a different journey that they're taking to get those questions answered. Whereas previously, we might have been trying to direct them to really heavy content landing pages to try to explain, here are all the differences in our fabric types. Here's all the differences, between down and down alt and the benefits between those two materials. Here's why Boll and Branch is different in how we source and make our products. And now all of that has been collapsed into these LLM tools, where, again, they're asking the same exact question. It's just a different way of presenting that information and, hopefully, in a way that is much more condensed and, like, easy for them to digest. And so a lot of our focus on agentic commerce I would say, like, if you were asking us a year ago, we were very focused on the technical side of the house of just making sure that, again, our basic product schema was set up correctly so that agents could read and understand our product catalog, agents could read and understand our general structure of the website. And where we've really pivoted this year is making sure that that rich content is also being fed into LLMs so that, again, they can speak correctly about our brand. They can understand how we would talk to a customer, who have these kinds of questions and help them with their decision making in that way. So, again, to boil it down, it's the same exact questions. It's just a different way of presenting the information and a different entry point once they get to the site. And, hopefully, they've already had those questions answered, so we don't necessarily need to go into the same level of detail once they get to our site. Got it. Yeah. That it sounds like you, Bolling Branch, is very, very at the forefront of of this stuff and really exciting to hear, moving yeah. Kind of some of the marketing shifts from SEO to AEO and where where the customers find you and stuff. But, that's all. really exciting. Thank you so much. I know you got a hard stop, so thank you for making the time and for sharing all that detail. I think it's really helpful for everyone here to hear how how a company like Bolland Branch operates. Great. Thanks for having All right. So next we'll bring our next speaker, Sierra, to the stage. And yeah, I think really great to have you, Sierra. It's you're obviously solving a version of the same problems as Boll and Branch, but with a completely different catalog. So, yeah. We'll we'll kinda jump into the questions here next, but, yeah. Just to repeat, Sierra works at Wild Alaskan in, the data side. And so, yeah. Really excited to have you here, Sierra. Thanks for having me. Awesome. So let me, change the slide here and get into our first question with Sierra. So, Wild Alaskan's whole catalog is essentially, three boxes and two sizes. With with a product mix that's kinda that tight, where where does your team focus, in terms of the user journey to, continue to drive growth? Yeah. So our, we that's catalog that kind of limited catalog is for our new member offerings, kind of that first box experience. We keep our that first box experience, fairly small in the offerings. They're a little on the smaller side to reduce decision fatigue during the sign up process. There are additional items that you can purchase, as sort of add ons to that base box experience, but, you know, the way we look at it, like, the user drain doesn't end once the member or this new member has converted. It's really just begun. And so once a person becomes a member, it opens up a much broader, set of offerings, that they can get from Fin Fish to Shellfish, etcetera. And, you know, it allows them to, you know, they can customize their boxes. They can kind of mix and match. We have a wide variety of curated boxes that we offer. So, you know, there's there's I would say there's no one place that we hone in on the most for the user journey, because it's it's there's so many different things. You know? But one area we really do focus on is trying to get members back to, our website and our app, to just keep them engaged kind of with our digital product. We see that that tends to help with just overall retention, and AOV, etcetera. You know, we want our members, you know, inactive members to come back, you know, even if they're not buying, we do want them to see, like, what's on offer, what's changed, what's new since they were last here, to keep up to date about recipes or news in the kind of fishing in Alaska, space. You know, the membership to us, the way we look at it and the way we are trying to craft it is it's not just about providing people with fish. It's about providing members with the tools they need to, like, sustainably and successfully integrate seafood into their diet. So it's a very broad it's a very holistic approach. Awesome. Yeah. Makes a lot of sense. So kind of building on that, can could you maybe walk us through an example of how, data helped you improve one of those key retention or engagement metrics, whether it's, you know, cohort analysis or AB testing or what have you? Yeah. For sure. So last year, much like Boll and Branch, as Dana just mentioned, we also overhauled our member shopping experience. It's, it's all the, you know, the rave right now. So, you know, just talked about them pretty much everything changed. And in a perfect world, we would where we had infinite time, and a ton of resources, we would have AB tested every single component and reached those statistical significance and put it on the site and then waited. But, you know, we work in the real world where we don't have time for that necessarily. So we use Mixpanel to track engagement with the kind of key points of the website that we know tend to lead to our higher retention, higher AOV, you know, better experience for the members and did a a just a ton of pre and post analysis on, you know, are they clicking on this? Are they going to this page? You know, are they finding what they need to find? You know, have we caused friction with these changes? And using Mixpanel to determine that with, conversion rate charts, the flows charts, etcetera. One more specific example, we noticed pretty quickly data. There's just a banner that's intended to, spotlight a specific species kind of like, hey. You know, this is back in stock. And, the click rate on it just dropped off. You know, the style had changed, and people just weren't engaging with anymore. But it had been a really large driver of sending members to our marketplace where they can, you know, make changes to their plan and and update their offerings or update what their their choices are. And that was a really it was a low hanging fruit. You know, just it's not working, fix the design, monitor afterwards, and we could see it's better now. You know? And and we're seeing people go to the page. We're seeing high engagement with that banner. And, yeah, mixed panel was made it really, really easy to to catch that. Awesome. Yeah. So, sounds like it was really that overhauling the member shopping experience, looking at some of those those key banners and and root causing when when there's a big drop off, like, what what happened and and looking at all that mixed panel. Great. Mhmm. So, yeah, I think this next question, it might be kind of a continuation, and it's similar to one of the questions we asked Danny, but, again, with a totally different product, might be interesting to hear. You track the shopper purchase funnel in Mixpanel. So, tell us about kind of the last time one of those, numbers changed a decision you made, and what what did you see and what did you do? Yeah. Great question. So, we we actually just recently ended, an AB test on our plan editor page, and this would kind of be, like, where you're, yeah, you're you're updating your plan and kind of saving, and that's what you're gonna get every month unless you make more changes to it. And we were made some visual changes, hoping that this would lead to higher rates of box customization, higher AOV, you know, all that good stuff. And, you know, the the test won. We ran this through our testing platform, the variant won. So, you know, it was that's all good. But I was digging more into the kind of experience of the users we're getting in the variant using Mixpanel, using the flows to kind of see when they go here, what are they doing afterwards? Like, you know, what are they clicking on on this page, and how is it different? Like, trying to understand more specifically what was it about the new experience, that caused the change, so we can possibly replicate that in other places on the site. And one just really simple thing was, in the new experience, the nav top nav bar was sticky and in the old experience, it wasn't. And realized that a lot of people were engaging with that sticky nav bar to kind of, it acted as sort of a filter so you could see different. offerings without having to look at everything all at once. And when, you know, when it was available to them, in a sticky version format, they clicked it. And that just made it a lot easier for them to navigate the new page, to make the decisions they wanna make, and to save. And that was probably a big driver of the improved performance that we saw there. Awesome. Yeah. So sounds like pretty much you similar to Danny using funnels, flows, insights, all to see. how behavior is changing and and make decisions off of it. Mhmm. Great. So going on to our last question for you. Given your data analyst background, what's what's been kind of most effective or exciting for you in in using AI with product data and decision making? On our side at WACC, we are where we haven't turned on the, MCP with mixed panel, yet for everyone. It's something that we're kind of playing around with. We're being taking a more cautious approach. But me personally, how I'm using AI kind of in my work when I'm doing this, is, you know, it's a lot of product research, competitive research, as well as just helping with writing SQL queries. We ingest mixed panel mixed panel data into our database, to help marry some of our first party, you know, purchasing data that we do have. We have kind of data on both sides, and sometimes it's easier to pull in, the kind of session and click event data into our database and work with it there. And so I do that pretty often, and that's that's where it's been really helpful for me, is making sure that everything's good, speeding up the process, and getting to results a lot faster. Got it. So, yeah, still kind of, testing the the Mixpanel AI features, Mhmm. but largely using it for, yeah, I've heard that use case a lot, SQL query writing, it just reduces time and, product research as well. So, awesome. Well, Mhmm. thank you so much, Sierra. That was all really great to hear. And, yeah, it's it's great to hear how real live Mixpanel customers are are using us in in their day to day. So really appreciate your time. And, yeah, Sierra will stick around for, Q and A if at the end, if you do, put any of those questions in the chat and we'll we'll get to them after, Reuben's demo. So now that we've heard some of those great stories, the the kind of real question is how do you practically apply some of these learnings to your own analytics practice? For example, Dani told us how she's saving so much time with Mixpanel MCP, to automate her weekly reporting pools, or Sierra shared how she, discovered the real unexpected reason behind why an experiment succeeded. So these are all the the kinds of insights that our latest releases are really built for. And so that for that, I'm going to hand things off to Ruben here, one of our senior solutions, engineers to showcase some of the great stuff we've been building and and how you can use it live in a project. Thank you so much, everybody. I'm excited to get this going. So can you all see my screen? And it says everything demo, all project data, and now you can see a screen within screen. Thanks very much. Excited to walk through a couple of the new offerings, that, our product has rolled out recently and how they can apply to a lot of what you just heard specifically, as well as, you know, what might be going on in your in your own case. In general, I would say it used to be enough, to learn from a product analytics tool, what happened, right? Checkouts went up by this much on this day, up by this much on this day, this is what the trend has been. I would say increasingly the expectation from folks we talk to, what we hear in the space, is that tools like ours need to be able to help you understand not just what happened, but why it happened and what you might wanna do about it. See, a common theme between what you heard from Danny and Sierra is that numbers moved. They found out that it happened. The reason behind why it happened, though, might have been counterintuitive. And in any case, it required some digging. That digging is has always been designed to be as easy as possible. We're breaking this down by plan type. We're breaking this down by platform, what have you. But when the actual answer is counterintuitive or unexpected, we have been working on building out some new AI powered approaches, to kinda make this digging a little bit less manual, a little bit easier, and a little bit more substantiated. If you happen to be in a report like this, all you need to do is click on root cause analysis if and when you see a spike or a dip in the data. This is one click away, and it's going to give you a dashboard that looks a lot like this that is intended to tell you not just that something happened, but what was responsible for it. So the volume tripled. It was primarily driven by these components. Here's what you might want to do about it. Here are the specific things we found. Don't believe my words. Look at these reports that automatically get generated. We can see here that US traffic is what is primarily responsible for this spike. We can see that it is, primarily a a free plan type of spike even though you can see a little bit across the board, etcetera, etcetera. So you can imagine that especially when the the reason behind the dip or the spike is counterintuitive that this can save massive amounts of time and really give analysts a jumping off point, to go from here and have some theories. Again, not just what, but why, not just why, but what you might want to do about this. Now as is the case with all of our AI offerings, they are powered by context. That context can live in your lexicon, which is going to be basically, like, which events are verified, what do the events actually refer to. All of those are are designed to make the give the AI engine a little bit more to work from and know kinda generally what to focus on. But there's even more you can do, especially with root cause analysis itself. There's I don't know if you guys have seen the AI settings, within your project settings. You can generally talk about, what your company does, what it is focused on, and that helps orient the AI. And then, additionally, you're gonna be able to specify in root cause analysis specifically, what do you want it to look at first? When we were kind of originally rolling out this feature, we got a lot of questions about, like, how does it know what properties, what cohorts, what types of segments to break down by, in order to give you some of these actual results? And, like, yeah, essentially, how does it know what to prioritize? We used to have kind of a more deterministic way of answering that question, which is, like, the top most queried events and properties are the ones that get, broken down by. But since kind of rolling this out, you really get to help the agent be opinionated and kind of derive some of these shortcuts, in order to really spit out meaningful answers. Now as as you can kinda see here, we were in the report, and we saw a dip. And I think that's that's still a lot of what generally happens is that folks like Danny and Sierra are in Mixpanel every day. They are sort of monitoring these metrics, as they evolve, and acting on them when you see them go up. This need not be the only way, to understand and be privy to your metrics live. Right? We also have recently built AI automations, which are essentially what they sound like. You can basically if you are interested, for example, in how your checkout funnels conversion rate is doing and you want to be alerted to spikes and dips in that metric, you can set up an automation. You can give it kind of specific instructions. Maybe you care about week over week. Maybe you want to be alerted when the actual results are at least the standard deviation above or below, what was expected, etcetera. So you can give kinda like AI style prompting instructions here, and the automations will respond accordingly. They feed directly into inbox and or Slack. And I'll give you a couple examples of what some of those outputs might look like. This is from that specific one of, okay. So here is your checkout conversion rate. It looks to be relatively steady. Not a whole lot of changes. Okay. Sure. View the report. View root cause on that. View the automation on that. It is not simply a conversion rate metric that is all you can do. Maybe you're interested in video watch events. This is a different example, of an automation. Here, we also have some of those, what happened, why it happened, and what you might want to do about that, kind of built into the automation itself so that you can be privy to changes in metrics, steadinesses of metrics without even necessarily having to be in the product itself all the time in order to act on these. So this really turns Mixpanel from a reactive analysis to, oh, look. This happened. What should we do? To a proactive one where you can sort of predetermine all the metrics that you care about, and simply get well informed substantiated digests on how all of those metrics are doing. The point about the, the sticky path, that that Sierra was mentioning, I think is a particularly good use case for something like root cause analysis because the reasoning itself we saw that the metric moved. Right? But the reasoning itself was somewhat unexpected. So this is the type of thing that an an RCA like this would have caught. So those are two of the in product, in product AI offerings, that we have been working on that I think are particularly exciting and relevant, to what the ecommerce space and really spaces in general are focused on these days. We heard, talk, from Danny of the Claude MCP connector. We also find that stuff, very exciting. There's a bunch of things that you can do with Claude. Here is an example of a dashboard, from a couple months back that, was simply we just asked Claude to build this. Right? And it was able to, without, overwhelming context, come up with some relevant reporting types. But I think perhaps more importantly and and helpfully, there are some other opportunities here. We talked a little bit about Claude's skills. Right? But if, for example, you want to be able to as I just, as I just mentioned here. So, like, if, for example, you want to be able to run an experiment and then keep folks in the loop on why this experiment is being run, what's gonna determine the success, and that type of literature lives in a tool like Notion. That's a perfect perfect use case for something like Claude to kinda automate looking at all of the data coming up with, reasons justifying the experiment itself and what you're ultimately going to judge this experiment by. This is a fairly substantial skill, but it's the type of thing that is really a great use case for, for that cloud MCP connection specifically. There was another one that's analyze experiment. And this was another thing that I think Sierra ultimately alluded to, in her example, where even if you were looking at the results of an AB test and you saw Lyft, that in and of itself may not tell you exactly why the Lyft was what it was. And in Sierra's example, it was actually it was not necessarily it was it was more due to the sticky nav than the actual thing that they were testing for. So this is the type of thing that not only looks at the results on the experiment itself, but also is looking at session replays, looking at a bunch of different components, to ultimately give you a more robust analysis of this. I think going back to and I believe it was Danny who had the, significance calculator integration with Cloud. That's another thing that you could do, if you want to, juxtapose the lift analysis that that, Mixpanel gives you in the experiments with a proprietary tool or something that you're more used to to to using in your other experimentation tool, that would be all fair game for a skill and a workflow like this as well. We touched a little bit, on this just in terms of experiments for those of you who are not, super familiar with this relatively recent change. I guess, new new offering. Mixpanel now supports experimentation, and we support experimentation, not just with our own native feature flags where you can use Mixpanel to flip coins, roll dice, and determine which variation of a certain flow, a user a user should see, but also tie in directly to any, like, third party tools that you already have. So you can create a new experiment here, and use any type of third party. You can set it up with exposure events such that maybe you already have a good, robust targeting, AB test releasing mechanism, but you really want that advantage of using Mixpanel for the analysis itself of that experiment, and that is fully supported as well. Looking at this example, I guess, what do I mean by that? The coolest thing about experimentation analysis in Mixpanel is the flexibility with which you get to judge the the success of this experiment. And it also does give you that statistical significance lift analysis, which you can, customize as well via your your configuration. You can see here, it's around 18% that, variant a is increasing the shop funnel conversion. But you could imagine that maybe you're interested as as, as Danny was in, like, those high value repeat buyers. Right? And you're running a test that is designed to improve the shop funnel itself, but you might wanna see what the byproducts are on retention, for example. Or you may even wanna see what the byproducts are, not just on the overall number of checkouts, but actually, like, the average, order value for those checkouts. It's not so much what the numbers say. It's the the flexibility with which you get to kinda ad hoc decide, throw in another metric to judge this experiment by, which is really gonna give you a much more robust sense of how your experiments are doing. So those are a few of the different, offerings that we have recently been working on, been iterating on. That's gonna include kind of directly from within the product, that root cause analysis, those automations. We haven't really even talked about mixed panel agent, itself. But everything that mix panel agent can do basically also is included in those MCP, including those integrations with other tools like Notion, like significance significance calculators, like Jira, etcetera, that may not directly exist in this tool. As well as that experimentation analysis, we know that folks in the ecommerce space are interested in iterating. They are interested in testing, shipping new functionalities. So all of this is designed to really support that full product improvement iteration loop. That is the extent of, what I wanted to show. Let's get into the next phase. Thank you. Yeah, I think, from there maybe we jump, before we get to some questions to your trivia questions. Oh, man. These trivia questions. Alright. So we're looking for, accuracy and quickness, in these responses. So let's find out. First question is, what are the two in product AI functionalities that we discussed today? That may or may not be one. Awesome. Yeah. So, Megan got got, got one of them. So I think that that is our winner for this question. The other one being automations, kind of the the the reactive versus proactive, kind of both sides of of that coin. Alright. Next question is, what what is an example of an integration or a like a tool destination effectively supported by MCP that is not directly supported by the Mixpanel UI itself. And as as Christie said, $50.50 dollar gift card. So so exciting exciting prize. Are there integrations where you might use MCP to support a whole workflow, keeping other stakeholders in the loop that aren't necessarily in the product every day. An example of an integration or a tool that MCP supports that the mixed panel UI itself does not. There. we go. Look at that. I was gonna I was gonna answer and take the take the gift card. So so thank you, Yana. Oh, man. We'd have to get lawyers involved. That's okay. Alright. Here is perhaps the hardest, for those of you who, were were paying attention. When I was going through that, shopping experiment example, to the closest percentage point, what was the lift associated with variant a, in the shop funnel metric? It was the it was the the green one, the green metric. Look at that. 18%. Unbelievable. Alright. Correct. Great job, Madison. Let's go, Madison. Awesome job. Great. So, yeah, I think, with that, we can get into, some of our questions. Some of you shared them beforehand when you registered, so, feel free to jump into the q and a section. I see Madison has one there, so we can we can start with that, and then then we can, go through the the preceded ones as well. So Madison's question is, going back to scheduled alerts, are there any templates or best practices on some key metrics that many clients use to get a pulse on their overall site health? So, yeah, that's that's a great question. I think, and, Ruben, feel free to chime in as well, but I think that was gonna defer. to you. So what I see a lot of customers, it may be a wishy washy answer, but it really depends on what your key metrics are. One thing that can be helpful if if you utilize Mixpanel's, metrics tree functionality. We we do AI generated metrics trees where you can give your website as well as, a prompt about kind of what what some of the things you care about are, and it will actually suggest, like, okay. Here's the logical business representation in a metrics tree of of what your inputs, the things you can change, and what your outputs might be of, like, your lagging indicators. So, you know, for for ecommerce companies, it's a lot of, what content is driving repeat purchases, what you know, how is how is our subscription or membership changing. Like, I I work with a customer that has kind of automated purchases when they're on a membership plan, and so that's that's what I would say that alerts often get set up on. And in terms of best practices or templates, our we have two kinds of alerts in Mixpanel. You can set it up to alert you if, some specific threshold is crossed. So if your metric increases or decreases by a certain number, by a certain percentage, and you can send those to Slack as well as, as well as email and, webhooks as well. And then the second type of alert is anomaly detection. Detection. So say you don't really know what what threshold to set, mixed panels anomaly detection uses an open source, framework to basically find, like, what's that, like, expected band of a certain metric, and we'll alert you when metrics go outside of that. And you can configure it to, you know, only alert me, if it's a surge or a drop. And so and and that's really great as well because when that alert triggers and sends to your Slack channel, it'll have a button that says root cause analysis, and you can actually trigger root cause analysis directly from that. So, yeah, a few different ways to handle scheduled alerts there. There was also a question, posted in the chat, from Dhruv, which I read and kinda had an idea for. So, basically, yeah, looking to optimize mobile listing pages to boost click through rate, conversion rate. When if and when there's, some type of product badge, high discount percentage, top star rating, I don't know. How do we know sort of which of those drive better conversion? If you have a demo environment open, look at that. I do. Show how that breakdown report would look like. Okay. So I don't have specifically the badge property. But we can certainly talk about what this would look like if you'll sort of bear bear with me in terms of the extrapolation. I would recommend a funnel for this. So something like, okay. You're viewing a product page. Right? And then we are trying to boost the conversion rate of it being added to cart and or checked out upon. Give it, like, a day. The way I would generally recommend this is you're gonna need a property on this view product page event that indicates if it is special in some way. Is it featured? Is it on sale? Does it have some top rated badge, etcetera? That property should be subject to the breakdown. And for this, we're just gonna say product category, but in your case, it should be badge status or, you you know, specialness one way or the other. From there, it's very simple to compare the relative conversion rates of these items to each other, because we are looking at basically on what whatever the product category was on step one of this funnel to what extent did these different components convert. So that's my thought on that one. Hope that helps. Great. So feel free to share any other questions you have, but, another one that was asked in the registration was, how does CART analysis work? So, Reuben, maybe you could again, pull up your your dev environment if you have a a way to show cart analysis. Yes. I do have a way to show card analysis, and the deck is actually in a slightly different demo environment, but who cares? Okay. So here we are. So we got, like, card analysis. So the way I I like to think about this is that if all you are doing are tracking something like a checkout event and you have maybe, like, a total cart value for that checkout, which is basically the situations we've been, discussing, earlier in this conversation, It's fine. It will certainly tell you how your numbers are doing, but in theory, it's a little bit of an oversimplification, because it isn't necessarily telling you item by item what the like, what types of categories of items, get check out check check out the most, what is a subtotal of particular items, and how does that type of granular information trend. So the example I'm giving you here is basically, like, if you want to know what are the most common items that get paired with a burger. Right? This is one possible example. It's like that this has implications for, like, upsell offers if you're running a I don't know, let's pretend you're a bedding company, right, and you wanna do sheets as well as pillowcases or something like that. What tends to get paired? So then can we surface that in our recommendations? So cart analysis is a way to basically set up your cart events such that as an object, as like a list of objects, such that each of the individual items has its own name, category, price, etcetera. So, like, the way this report works is, like, show me all of the checkouts that have a burger in them in any of the items have a burger, but then show me but then filter it by all of the items that aren't burger and then show me which ones are most popular. So the way to read this chart would be like, wings is the most common thing ordered alongside a burger as well as a cheeseburger. People like to go burger and cheeseburger. Go figure. Awesome. Yeah. Yeah. And to add on to that, you might have price in that cart and, that's where you can you can actually go a level deeper to, use aggregate properties and say what's, you know, the average price of your items in your cart or the sum of of the the price across the items in your cart. How did it exactly. How did the doll yeah. Sorry for not showing that. But, like, how how do all of the dollars that have been checked checked out shake out by individual item categories, which you would not be able to do if you were just tracking a checkout and tracking a a cart value. You wouldn't have that that granular flexibility. So we recommend cart analysis for some of those more advanced, use cases. Yeah. Alright. I don't see another question in the chat, so I'll continue with the, ones we got prior to the webinar. One that I think is really great for, you know, so many econ businesses are very seasonal. So my business is very seasonal. How would root cause analysis understand spikes versus seasonality? And, yeah, I think I think it's somewhat straightforward. You you can very easily adjust the time period of the report to look at, you know, the whole year, and over year. over year. Look at the, you know, by month or by quarter on a line chart and then, as Reuben showed, it's as easy as clicking into a spike or drop and surfacing, you know, those three hypotheses that our root cause analysis can show of what might be driving that behavior. And then like Reuben also showed in the settings, would probably be really good to fill out, as he's showing in the AI settings for business context, you could add, more details about your seasonality, what what kind of things you expect to see, what you've seen in previous years. And same thing with the root cause analysis AI setting. You you could, you know, if you're really more interested in looking at does is the spike driven by membership status or by, product type or or what have you? You can you can have the root cause analysis a little more specifically focus on a set of properties that you're that you're most interested in. I will also add to that that I have found, mixed panel agent has some basic knowledge that you don't even need to give it with context. For example, I set up a, I set up a demo project one time that, like, had weaker Friday conversions, and it knew. So I ran a root cause analysis, and it was like, yes. It's slower, but it's a Friday. This seems typical for a Friday. And you didn't even need to, like, specify that type of context. So I think it know it knows about that type of variation. I it knows about general holidays, including Black Friday, Christmas, etcetera. So the more context you can give it specific to your own business, the better, but it also does have some generalized knowledge, Yeah. we can say. another thing about seasonality beyond root cause analysis, I've I've seen a customer, you know, look at the previous year's cyber week and get really detailed in their their cohort building. So basically build cohorts for any number of, behavior types during cyber week. Like someone that, viewed a viewed a product page on Monday of cyber week and actually returned to purchase on Friday of Cyber Week or, active days during Cyber Week. So someone that was active for two days or three days and and basically build out a bunch of cohorts in Mixpanel, which a cohort is essentially a list of users that you can define by what behaviors they've performed or properties they have and then you can use those cohorts as filters and breakdowns on reports to see, you know, is some specific metric impacted by users being in or out of certain cohorts. So, great. I think with that we can kind of wrap things up here and, yeah. Thank you for your time, everyone. Great attendance and great speakers. A big thank you again to Sierra and Danny. And yeah, you can you can find us after the the webinar. Thanks so much, everybody.