蜜豆视频

Transcript

Hi, all. Thanks for joining. We will be getting started in the next couple of minutes. Today鈥檚 session, Mastering Sequential Logic in Analytic sense for a visual framework, will be led by the principal strategist, Andy Powers. We鈥檙e going to wait just another minute for attendees to filter in, and then we will get started.

While we are waiting. I would like to let you know that we do have several other, sessions coming up this month, and I will, paste them in the chat here for you.

So, again, a big, warm welcome. And thank you for joining today鈥檚 session. Mastery of sequential logic in AA and K, a vision framework.

My name is, Fred O鈥機onnor, and I work in 蜜豆视频鈥檚 ultimate Success team as, senior customer success manager and ultimate Success team. We focus on assisting 蜜豆视频 customers, to get as much value as possible from their 蜜豆视频 solutions. So I鈥檓 going to go ahead here and kick off our session today.

First and foremost, thank you for your time and attendance. Just to note that this session is being recorded and the link to the recording will be sent out to everyone who has registered. This slide webinar is a listen only format, but as we go through content in today鈥檚 session in free to share any questions into the chat or the Q&A pod and we will try to answer as possible.

And in addition, we have research some time a short time bit to discuss questions at the end of the session. And if there are any questions that don鈥檛 get that we don鈥檛 get to during this session, the team will take this and follow up.

And also, I will be sharing out the survey at the end of the presentation that we would love you to participate in to help us, shape the future sessions that we are planning.

I am joined today by our presenter and the powers. His title is principal strategist, and he has told me that this is his finest. Presentation ever. So I鈥檓 really excited. And Andy is a real, experienced digital strategist with a passion for data and analytics. Today we will talk about sequential logic with the then operator, that will enable high value analysis that is broader, under utilized. So with that short intro, just, please go ahead, Andy.

Thank you very much, Fredrik. And thank you everyone for joining. It is true that this is one of my, most excited topics and one of my most beautiful presentations. You鈥檒l see why. It鈥檚 just a sequential logic is very tricky to understand and communicate, and I think we found a way that will really help everyone who attends to understand more about how the logic in these work, and give you an example that you can take back in your environment today.

So I鈥檓 super excited to be here with you all. Let me.

Advance my slide if I can.

And try sharing this again.

Okay. It was frozen but we鈥檙e good now. So we鈥檒l do a quick overview of some of the material from the first session. The prior session to this was foundations on ten operators. We鈥檙e also going to look at visualization of report data and visualizing then operators.

And then we鈥檙e going to talk use cases. For the majority of the session. And Fredrik just a reminder I think you鈥檙e unmuted and you may want to mute.

So the expectations for today, the goal here is about this visual framework. And the idea is that I want to help, you know, how to take a scenario that requires sequence logic, make a plan, and then build it in CJA or AA. So we鈥檙e going to do a review of some operators, make sure we鈥檙e on the same page to start, and then look at the framework. And still like the prior session in this series on foundational elements. This is building up to even more important and valuable analysis. Things that we can do, that are going to come in another session to be had in in the future. I don鈥檛 expect time for Q&A because the topics are so complex and broad. But if you put questions in chat or follow up with your, CSM or Tam or Tad, we鈥檒l try to get questions answered afterward.

Lastly, reminder or information. Anyway, this is a series. So, that鈥檚 because there鈥檚 so much to cover here. The the way that we can define segments, especially sequential ones, is extremely valuable to all sorts of business questions, but it鈥檚 very tricky, and it requires a lot of guidance. So we鈥檝e split this up into a series. Today鈥檚 our second session. And you鈥檒l notice info to get to the prior one in the description of the meeting. And if you鈥檙e finding this valuable and want it to continue, please communicate. So in our poll at the end a couple quick more things on intro and then we鈥檒l move to the real thing. One, I鈥檓 going to assume everyone has some basic familiarity with analytics and or for some light familiarity. At a minimum. Kind of the idea of segmentation that we have controls and different levels and some operators, but we鈥檒l go through most of it from a fairly low level. Before we get up to the the more intermediate topics.

And then lastly, because segmentation exists and is powerful for both analytics and CJA, being 蜜豆视频, we have given new terms to things and it makes sense, but it still can be tricky. I鈥檓 going to be using terms that apply to both analytics and customer journey analytics. They pretty much represent the same thing in analytics. You have visitors visits and hits. In CJA, we have people sessions and events, and we鈥檒l use either the label of segments or filters. But we鈥檙e really talking about the way we define kind of a subset of data and analytics tool.

Okay. Click review. Some basics. Make sure we鈥檙e starting from the same foundation. So segments are kind of user friendly elements of workspace. They can also be controlled through the API of course that let you define a subset of data.

So things that are related to sequential segmentation are elements that are kind of highlighted here, as well as just segment controls in general.

Brief recap of them and then a couple other operators. So then just let鈥檚 say these conditions need to be met in a certain order. So this would tell us about sessions. We have session where our user saw both something that was tracked as web, something that was tracked as email and web happened in time before email.

There are modifiers that let us say these checkpoints need to happen within or after a certain period of time. So here we鈥檙e looking at a person level of data with conditions that say give me people data who saw web and mobile app, events or hits. But the web activity was followed by app within the second day after where it was after one day, but within, two days. So in that second day afterward.

And then we also have controls that let us look at things before or after a sequence. The prior elements I showed are giving us a group of data for a person or for a session. Here we鈥檙e going to see a subset of a session or subset of a person. So if I have this definition, I鈥檓 saying give me the portion of someone鈥檚 session or visit.

Before up to when they saw web followed by mobile app. We鈥檙e going to go through examples to solidify this. But it鈥檚 covered in more detail in the prior session as well. And the materials, they鈥檙e helpful. So the idea here that鈥檚 important to remember is we can use these controls to get a portion of sessions or have a visitor, data set back and we can do it and the other direction as well. So give us the portion of a session where we saw web followed by mobile app, and then just everything after that point. So it鈥檚 useful if we want to study what someone did after they went to app following web.

And one new one I want to cover today. It鈥檚 going to be, easy to understand for anyone who鈥檚 worked with just normal and or and exclude conditions and segmentation, but the idea is that we can exclude a condition here, and whether it鈥檚 at the where it is in these set of checkpoints with then operators will determine kind of what its function is. If we put it at the end here, you鈥檒l see if I click this gear icon I say exclude and it changes this header bar into red. So this bar here is going to be the indicator that it鈥檚 an exclude. This is going to tell us, tell us about people who saw web with no email afterward.

It鈥檚 going to give us everything for that person if the conditions match. And really, you may as well think of it as if the last time that they saw web within their set of person activities, there was no email after it. Then we鈥檒l get back all the person鈥檚 data. We鈥檒l have examples further on this too. I just want to introduce it. That鈥檚 what happens if we see excludes at the end.

If we put it at the beginning, we鈥檙e going to say give me the people who interacted with email, but not with web before email.

And again, this is going to give us full data for a person so we can study that person鈥檚 entire behaviors. And it鈥檒l also come down to really being the first time they saw email. Was that or was that not preceded by some engagement with web data? They definitely need to see an email, but not necessarily have seen what they could see web after email. But, if it was before the first email, they won鈥檛 qualify here. And then lastly, if we put this in the middle, then we can say, give me the people who saw web and then they saw mobile apps, but not if they had email in between. So I need to have at least some web followed by app with no email breaking that sequence.

Okay, so just trying to set the playing field. Now we鈥檙e going to really get into the the content that鈥檚 new. And that I want to introduce today. Visualizing or reporting data. Whatever data collection mechanism you use here sending data and analytics or CJ or app and it鈥檚 coming in. It鈥檚 coming in with different sources, the timestamps. So really we鈥檙e focused on not like metadata. That鈥檚 like a classification to give kind of enrichment out of a field. But here we鈥檙e focused more on talking about the fields that had the time stamps. These are things people customers did on a digital property in store. However, we collect, these are things with time associated.

So what happens in analytics and CJA can be thought of like this. Going to take all the activities for that are identified to to me, to Andy. And we鈥檙e going to separate those and we鈥檙e going to organize them by their timestamps. So here I have some actions that I tracked that were tracked with my identifier and the 9 to 930 window, some later, some later again. And then we have other visitors or people who are having their sets of data ordered by time. All of these boxes here just represent either a hit or event, depending on the terminology which tool you鈥檙e in. It just means some some set of data sent in.

And then we grouped things into visits or sessions based on there being some period of inactivity. You can customize that, but the default is 30 minutes of inactivity between a set of, activity that was not broken by 30 minutes. And then lastly, we can look at data as an aggregate for a whole person or whole visitor. And that鈥檚 going to be everything within the box for that person.

So what I want you to get from this is think of the data that we work with, in this manner, because this is the framework we鈥檙e going to build on. So all the activities ordered by time, we have events or hits that are the smallest units, we have sessions or visits that are going to be groups that are ended by some set of inactivity, and then it鈥檚 all contained as a person or unique visitor.

Even within those boxes, there鈥檚 still lots of data points. Whatever you send in, in a given hit, take analytics or you send in through web SDK that goes to platform or that comes into ADP and then CJA from some customer data set, you have, there鈥檚 going to be lots of fields in there. But the point is that if they鈥檙e in the same hit, the same event record, these are going to be related, which lets us do things like, say, logic. Give me all the events that had both platform equals A and page name equals B on the same event. Or give me all the sessions of people who had an event that had campaign C or product products queued.

So we can think of things like that. Then we can move on to visualizing some more of how sequential segmentation works. Let鈥檚 start really simple. Here. We can look at sessions or visits where email and web both occurred and web preceded email.

So here let鈥檚 say we鈥檙e going to do this a lot. Make a hypothetical set of data, for the kind of enablement portion. And then we鈥檒l have some live demos, at the end of the session. But here, say I have some data, some is from web, some is from email. I have three sessions here that I鈥檓 kind of representing with these gaps.

And the way this is evaluated is we鈥檙e going to look at every session. We鈥檙e going to look at every hit from the web. And each session we鈥檙e going to say, all right, we see this activity that鈥檚 tracking to web. And here web just means there鈥檚 some you know I鈥檓 not revealing what the condition is, but there鈥檚 some field on this event that I鈥檓 matching for in this definition I just consolidated the definition just to make it less less detailed here. So I see a web hit. Then we鈥檙e going to look forward within the session. Do I see an email hit I do. So this has now matched this session against this logic I鈥檓 going to look at the next session or not I but you know the the the systems that evaluate the analytics queries web followed by email. Good. And then this last one has web. But there is no email following it. So the result that鈥檚 going to come back for this is that we鈥檙e going to get these two sessions in full because we said, give me session full data and we won鈥檛 get this one. So that鈥檚 the easiest kind of starter example. And if we change this to person or visitor we鈥檙e going to have the same evaluations here. But the second and third ones don鈥檛 even matter because here we鈥檙e saying if we ever see web and then email anywhere in the whole person lifetime, then bring back the whole person data set.

Then let鈥檚 think about how this after and within works where we鈥檙e trying to say, give me people data where there is web an app, but there鈥檚 some kind of rules on when it occurred. So for this example, let鈥檚 think of it over time. I have some web and some app activities. This web activity started late on day zero, and I鈥檓 going to be approximate by kind of what period in the day these happened. As I talk through it. So I see web hit on, let鈥檚 say this is 11 p.m. is the timestamp on day zero. Well, I need to wait until after one day, which is really 24 hours here before I qualify to kind of match the next checkpoint. So here we鈥檙e looking at 11 p.m. on day one, and now I鈥檓 looking within that next period until two days. Happens to see if I see any app activity. And here 11 p.m. on day two I do not. So this web data point is not followed by mobile app. That kind of fits in the rules. So that鈥檚 not going to be a match. However, if I look at this web data point go out 24 hours, look within the next 48 we do find an app hit and that matches. And because we鈥檙e looking at whole person data, we鈥檒l get that whole set of persons data back to us.

Now let鈥檚 look at an example for how this kind of only before sequence works. This one鈥檚 more complex at first glance. So here I have app I have web data. Well like always we鈥檙e looking for checkpoints and the sequence. So I see web. Then I see mobile app okay I met that and then I have this. That鈥檚 going to say don鈥檛 give us the whole person data. Just give us everything before the sequence. So this is the result I get back here. It鈥檚 everything starting right at and before the sequence. It may help to think a bit more of this way when I鈥檓 doing this sort of only before sequence. So you can kind of think of it like it鈥檚 evaluating backwards, and that may be a little more intuitive. I see an app activity. I鈥檓 looking did I see web before that? I did and now I鈥檓 going to match that and everything up and everything before it.

Now I also need to call out that segment. Logic is what I鈥檓 calling greedy. Late greedy is in regex. So if there鈥檚 a way it can match and return more data, it will. So actually the most data that can be returned here is based on this app. And this well, any of these apps here. But this is the most data that we can get where there鈥檚 web followed by app. So this is actually going to be the full return here for this data.

And I promise we鈥檙e almost at the end of these kind of just informative, hypotheticals. And then we鈥檙e going to talk the real use case and, two use cases after this for most of the rest of the session. So if I do this exclude at the end, we鈥檙e looking for people that they touched web. But there was no email after that. So here I was only use blank because I also want to remind you that all these data points, they鈥檙e going to come from different sources, have different pieces of information on them. The only thing we鈥檙e doing is we鈥檙e trying to think about how these evaluate is we鈥檙e focused on the web and email elements, but there鈥檚 going to be other activities tied to all your visitors across all their data sets. So these are just here to remind you that it鈥檚 it鈥檚 everything. Not just the ones here. These are just what we鈥檙e focused on because they鈥檙e part of, the logic we鈥檙e using. So here. I鈥檓 looking for web, and I don鈥檛 want to see email after it, but I did see email after it. So this sequence of web two emails not going to match our condition. However this web activity is not followed by email. And so it does match. And even though we saw before there was no match since the it does match here, the idea is that again, it鈥檚 greedy. It鈥檚 giving us as much data as possible. If at least one thing matches, we get the result back. So we get the whole data set for this person. Okay, so if you鈥檙e following at least with that kind of foundational understanding, now let鈥檚 talk about this with some scenarios that will solidify it more and explain why we鈥檙e we鈥檙e going through this, why I think this framework matters. And for just kind of baseline assumptions like this works on the idea that we have data coming in. It may come from different sources, and we鈥檙e able to group it to a given visitor to a given person. Ideally, if it鈥檚 coming from lots of sources, not just digital, but also, other external non web, non app sources, there鈥檚 going to need to be some way that we鈥檙e tying them together by shared customer IDs.

We know that analytics it can unite and align some outside event data sets. But there鈥檚 a lot of limitations there. Whereas in KE and AP there there are new applications built to do this. So it tends to fit better for views like this. But regardless of that, the logic here and how these are evaluated, the kinds of things you can do with it, they still apply in both toolsets. It鈥檚 just that the scope of your analysis between maybe digital or web and customer level analysis will differ.

So let鈥檚 say that we have an app, you have a website, and we want to increase our customer retention, by making sure that our mobile app covers the core needs of our most of our customer base. So we want to study the data and see, are there any customer actions that are maybe signaling. Here鈥檚 something the customers wish was in the app, but is not currently there. So to think through this, we鈥檙e going to study the customers who view the app and then view web because our hypothesis here is that, hey, if we see someone going from app to web, maybe that鈥檚 a subset that might be more likely to be kind of indicating that they鈥檙e looking for something that鈥檚 not there.

And actually that鈥檚 pretty broad. So let鈥檚 focus on maybe single sessions where someone鈥檚 moving from app to web and really there鈥檚 going to be plenty of reasons that that鈥檚 a legitimate action. So for instance, if someone is on web first and they may go back and forth between app and web for some reason, and our fictional kind of digital presence here, the idea is let鈥檚 exclude things that are irrelevant and one way we鈥檒l do that is let鈥檚 excluded. If they saw web first and then actually we don鈥檛 really care about what they did for this purposes of this study, this hypothetical use case, we don鈥檛 care what they did until they got to the web and they went to our areas where they can really start demonstrating something that they鈥檙e seeking. They go to our fact pages, they go to our search and start looking for other data. So this is just one way that we can think through and kind of build a segment step by step, which is what I want to to convey to everyone today. We鈥檒l do this this hypothetical use case. And then the next one is going to be more complex but also more tangible and something that you can recreate that will apply, to analysis that you鈥檇 like to conduct and that you can replicate. So I still need to start here and keep building it up. But the next one, the last one we go through, I鈥檒l have demo, and it鈥檒l be something I promise that you can actually apply today and get significant value from.

But for this customers viewing the app and web, this is just the first of our points that we kind of broke down as we were thinking through how to address our scenario here. Start very simple. You know, there鈥檚 not even a lot that we need to kind of plan here. But what we want to do is we want to think whenever we鈥檙e doing sequential segments, what are the kinds of ways that the checkpoints we鈥檙e looking for could occur? Because we want to make sure you鈥檙e covering all the, the expected and unexpected things like, say, here, that we have six different sessions that could be from different people. It doesn鈥檛 matter. The point is, some might be app only, some web only, some app followed by web, web, an app, and so forth. If we consider all of these, then we can make sure that we鈥檙e considering the very, the full variation of what customers may create in our data set. So here I鈥檓 looking at person. All I care about is app then web. And that鈥檚 that鈥檚 it. I鈥檓 for this step. It鈥檚 not doing a lot. That鈥檚 very interesting. We know someone saw app and they saw web afterward. But then we decided actually we should scale that down. We were and our hypothesis that customers doing this in the same session might send some signals about things that could be added to the app. Let鈥檚 look at the session. So here. This has app followed by web. This has Apple by Web and this has app followed by web. So these three sessions are going to be returned instead of the whole person lifetime data.

It doesn鈥檛 matter here that we had web before app. We just need to have based on our current definition app followed by web. And here too, it doesn鈥檛 matter that we went back to app afterward. But we know there are going to be reasons that this doesn鈥檛. Demonstrate that there鈥檚 a gap. So let鈥檚 try to remove some of those. And one that we can do is let鈥檚 just say that we鈥檙e focused on someone who is really app first and then web. So if they were web first, we don鈥檛 want to consider them. So we鈥檒l add this as an exclude condition at the start. Here. And the way this will evaluate we鈥檙e still going to get apps and web and app and web app. Those sessions will come back. But here we start looking at the session. We see web and then that鈥檚 it. This session does not meet our conditions anymore. So we鈥檝e turned it down further. Just two sessions app followed by web but not web before app.

And then we wanted to layer on to that that we know in our company that our website is not very, large. There鈥檚 one key section where we send our customers to, to type in search words, look for frequently asked questions. So that鈥檚 what we want to see when someone goes there. Then what did they do? What do they search? What are the actions they take from? Then on. So we can add this as a final step here. But because we also want to get rid of the noise that happened leading up to that point, and again in real life, that may or may not be be applicable. But for this use case, we鈥檙e going we want to ignore all that stuff that happened before. We鈥檙e going to use only after. That鈥檚 going to say someone met these conditions. And then we鈥檙e going to look at what they did on search traffic and everything afterward. So the way that鈥檚 going to play out is that here, let鈥檚 say now, now we鈥檙e just looking at two of our sessions that would match the criteria so far and sort of giving more detail to them so we can really think about how this might play out. So this has app and web and then this is actually a web page as well. But it鈥檚 that PCU page. And that is part of our logic. So I鈥檓 highlighting it here app that in web then fake that matches our conditions. And because we set only after sequence we鈥檙e going to get that a fake page and everything afterward within that session.

In a session like this here app web again making sure there鈥檚 no web that happened before apps and then a back and we get everything from fake onward.

So this is the process that I want to recommend for building thinking through sequential logic. That it matters a lot, especially when you鈥檙e getting started. But more more than anything else when you鈥檙e trying to do something complicated.

We want to make it clear. Business question. We want to translate it into kind of modular ideas or pieces that we can handle separately. We want to make a plan, think visually because it鈥檚 visually and in the tool with validation. But part of this is that I鈥檓 introducing is this visual idea of how to think through it. And then we want to draft logic and work on validation for this scenario we just went through. We didn鈥檛 do the validation step, but we鈥檙e going to go through that and in this next use case, and we鈥檙e going to show how we build this out and analytics or CJA and the segment builder. And we validate the pieces. And then we can save these pieces that we validated that weren鈥檛 complex, but add them all together to make something that does a complex use case that we know works because we went through it piece by piece and validated it along the way. And then we can also try to validate at the top. But when you have something very complex, it can be hard to validate it. You always have to break it down to do the validation.

So for this business case, this is the the one we鈥檙e going to focus on for the rest of this, let鈥檚 say we want to understand the journeys of our customers. After submitting an application, we want to explore their interactions grouped by kind of what visit number a session number. It was from the application submission. And the reason I鈥檓 saying that this is going to be valuable no matter what your data set looks like, is you can substitute the idea of submitting an application to any point of interest. It can be very valuable to study what happens after that, not in terms of like the next day necessarily, but if some customers don鈥檛 engage with us for three days or others within an hour, they have another contact with us. With nothing in between for that hour. So it鈥檚 a new session. Just knowing what was the next kind of brand exposure to this customer. That is a very valuable thing that you can unlock with this approach, that we鈥檙e going to walk through.

So we鈥檙e going to want a set of segments really. We鈥檙e going to want to see the first visit after the application submission, the second one after, and so on. So that鈥檚 why I鈥檓 calling this to the nth session after something, some application submit here. And it needs to work for all customers. We can鈥檛 do this one off for all the customers who had their first contact. Exactly one day afterward, or whatever they are going to have. Application submits at different times, different follow up timing. So this needs to be robust and you can apply this approach immediately.

So high level you know we want to study all the customers who did the thing we care about. They submitted an application. But since every segment can only identify one result, the way we鈥檙e going to do this is we鈥檙e going to probably make separate segments for the session, the first session after the second session after. And really, if we can do this for one scenario, we can probably duplicate that approach and extend it. So if we can make the first visit after, application, we can probably replicate and tweak to make the second and third and so on.

So here鈥檚 some theoretical data where there鈥檚 application for visits. Here. We want to get at this visit the kind of next engagement after they did the application. We know that we can get something like this because we did that with only after. We also know we can get something like this because we did this with only after. So if we can take those two and sort of subtract them, what we鈥檙e left with is just the single visit that came after the point of interest.

So let鈥檚 study all the customers that had a visitor session with application.

Let鈥檚 look at all the activity that came after that session. Then let鈥檚 also identify just the activity that came after the session, after that application, and then do that subtraction. That should just give us the piece we wanted. So let鈥檚 step through it. Start super simple. All customers with an application session. We鈥檙e going to put it at the person level because we鈥檙e going to look across sessions or visits. So it needs to be ultimately starting from a person level. We鈥檙e going to look at a session or visit that had application, so within this I鈥檝e just collapsed something like the application event exists. So we make up some fictional ways that this data could look, I think through. All right.

Application existed basically. And we get back the whole person data start really, really simple. We need to narrow it down. Of course we can do some quick validation too. So here we鈥檙e looking at two segments all our data. And then the people in that segment we just made, the ones that had an application visit or session.

Well we鈥檇 expect that all the applications happened in happening in our whole data set match the ones we captured. If we made our segment properly. So that鈥檚 good. And then also, we鈥檙e expecting that this group of people is going to be some subset of the whole group of people. And we see that鈥檚 validated because we only want those who had an application.

So let鈥檚 move on to that second step where we鈥檙e trying to make all the data, starting the visit after application.

So we鈥檙e going to start at the application, submit the session with that. And then we鈥檙e going to look at the next session. And we鈥檙e going to match everything after beginning with and after that point here where I have a exists, the idea is this is just saying something happen.

All our activity has time stamps. So by saying day exists or timestamp exists, it鈥檚 a way to just say true, something happened. So I鈥檒l often use to exist when I just need a generic. There was data. So in this session we see application.

And then because this is a session and then we鈥檙e having then it means we鈥檙e not going to consider anything until we鈥檙e out of that session and reaching a new session. That鈥檚 when we can apply the second checkpoint we see day exists. Some data came in from a subsequent session. We match it and then that鈥檚 how we鈥檙e getting everything. Starting one session now.

To validate this, let鈥檚 compare it to the segment we had before people that had an application. And then this one is all the activity in the session after and subsequent to the application. So we expect there to be less people. These are all the people that had an application. If we鈥檙e saying look at the next visit after that, well, some of them didn鈥檛 have a visit after that. So this number should be lower than the total number of people who applied. We see that our counts of sessions or visits, decrease. It鈥檚 about the number of people where we鈥檙e seeing about that many sessions decreased. And here if we鈥檙e actually just looking at the raw count of, sessions where there was an application, this is exactly the difference of the nine, 57,000 people, because we said skip ahead by one session for each of them where they applied.

If we order things by time, it can be really useful for the validation step when we鈥檙e using sequence details. So here I鈥檝e got a person that I filtered for. I鈥檓 looking at time. And I chose this person because I was looking at the data to find someone who had some set of visits here over time and they had one application visit or application session.

Because then when I apply my segment here, say, this is the one where I鈥檓 trying to only get everything after the application, it can be really easy for me to see visually that I only get the data coming back after that.

So let鈥檚 look in the tool at this.

Here I鈥檒l zoom in a bit.

And let me collapse this.

All right. So this is the example I just showed. Here鈥檚 another person for this person in the middle we see they had some activity. They had an application on February 22nd. Here. We only expect to get everything following that February 2nd session. And this looks good. I鈥檝e got another person here where I鈥檓 applying the same logic. We want everything after February 1st, so we get the second. And they had two sessions on February 8th. So this plus going back to our presentation deck, This plus looking at it from this point of view, like these are strong validators that this is doing what we wanted.

It can be very helpful to think what kind of persons, what kind of example scenario do I need to test and prove that this little bit of my logic is working properly. So that鈥檚 why we took that step.

And then let鈥檚 go further. Let鈥檚 say all right, one session after the application. Well that鈥檚 not that hard. If we just add in this after one session bit, then we can say, all right this is the session with an application. We鈥檙e going to jump ahead a whole session. And now we鈥檙e going to see is there any data that exists after that.

And there is. So this is going to give us that two sessions out and everything afterward. The second piece of our puzzle for the kind of subtraction to get our result.

And here again let鈥檚 compared what we had before. Does it make sense? Usually people count visit counts, sometimes page view, event counts. Those are the kinds of things that are kind of a good gut check to start.

Yes, there鈥檚 fewer people here because some of them don鈥檛 have a second visit after having their application submitted. And we see exactly one session decrease for everyone from our first sample. And that is what we expect. As we said, move one session ahead. So that looks good. Here if we add this next to what we had before, it can also be nice and visually easy to confirm that we鈥檙e getting the kind of effect we expected. So here there were two sessions on the 12th.

The fifth is when they applied. Our first segment got us everything after that. This one is getting us everything. Two sessions after. And I鈥檒l just show that also and two other examples here, three other samples.

So in this person to be here鈥檚 their application, here鈥檚 everything after it. And here鈥檚 everything after the second session from the application or point of interest. And same thing here. Applied everything after application everything two sessions out and after application.

So now all we need to do to finish this as we build it piece by piece is just take what we had before. We already built these two parts. The second part we built was all the activity. I mean, we just saw it here. We built this one activity after the application visit. We built this one activity after two visits after application. So here鈥檚 where we鈥檙e kind of applying that that plan that we made visually this piece represents this where we have an application in this first visit.

This is our most recent segment where we鈥檙e saying here鈥檚 everything that was out a visit from application. And the way that we kind of execute this like subtract is by excluding the second element here. So we鈥檙e going to say give us everything that matches this, which is this group. And look at what match this, this group. But get rid of it.

So the result of this should be just that session right after application. Look at our numbers again. Our numbers of people involved should match our first step, because we鈥檙e starting from the same group.

The difference in those two segments is exactly the number of sessions in our final segment.

And if we look here at this example, we鈥檙e seeing again exactly one session out. Here鈥檚 the application. And we are now getting the first of these two visits that are exactly one session or visit out.

So here鈥檚 the same kind of view from the tool set one session after application, one session out from application, one session out from application. So these are three different people. This person had their next engagement a day later, this one two days later, this one a week later or so. This works for all visitors. It doesn鈥檛 matter when they did the application, it doesn鈥檛 matter how far out their next contact was. And that鈥檚 a very tricky and very valuable thing to be able to study with one segment. So we can say the next activity session for any particular visitor, and we can look at this for anything we care about. It could be a purchase, it could be a cancellation. It could be a sign up, an application could be the first time that they walk in store and we scan their loyalty card, good things and bad things. And we want to know what鈥檚 the next kind of engagement they have. And it doesn鈥檛 matter when they had that activity, it doesn鈥檛 matter how far out the next one was. And we can just take that segment we made, just throw it on the whole data set and everything subject to that segment condition is going to just be the next activity, the next session they had after the point we care about.

This is what our final result looked like. Building this from scratch is hard.

Even if you鈥檙e comfortable with it and are experienced with it, it鈥檚 still tricky to do it right. So the way that we did it piece by piece and then we took this is here and we combined it with this one but excluded the second one. This is why I recommended that process I think had like seven steps. And that鈥檚 what we鈥檝e walked through today.

But then we were talking about can we maybe see their progress over the end session out. And yeah, we can do that because all we need to toggle here, this was originally working one session out. Well if we up at the after here by one and after here by one, we鈥檙e going to be able to get two sessions out and three sessions out and so forth. So this is what I mean. Like this is how it鈥檚 immediately applicable to you. This logic. Take your point of interest and it doesn鈥檛 have to be visitor session out. There鈥檚 lots of ways we can define this after here and ways that you can define, the point of interest and what happens next.

But a simple way to start if this is a new capability is with visits. So let鈥檚 look here, I think I have the bottom here. Let鈥檚 expand all these.

And I鈥檓 going to zoom to normal.

So here person A one session after the application. This is the application on the fifth they had two sessions. Then on the 12th all I鈥檝e done is here. Just kind of cut out the in between days where they had no activity. So first session got it. Second session got it. This person applied on the second. Here鈥檚 the first session out. They had two on the 10th. So here鈥檚 the first on the 10th. And here鈥檚 the second on the 10th. And it actually shows up on two days because this second session of theirs straddled midnight. So it even it doesn鈥檛 matter. All that matters is that it鈥檚 within the scope of a session or visit. So we will get the right result even if it spans two days.

Same thing here for second, third and then here.

I鈥檓 not applied it to any person. If I was to study this, let鈥檚 say I looked at one session after application here. I put it against days because that鈥檚 what I had in my example. But more I think I have this on the slide. More interesting than just counts. Might be things like what did they do in the first session out? What did we do? What communications happened? What methods did they use? What content did they explore? So we wanted to do this with our use case, understand journeys after point of interest so we can look at what what trends, what shared experiences as each of them engages over time. Can we learn from and optimize on so what campaigns and channels are popular or effective or not working? What content, what tasks are they pursuing next? Do we see? Can we take those insights and turn them into opportunities, maybe to delight them with some sort of personalized journey that is going to address something that we didn鈥檛 realize until now that they needed or wanted? So you can take these principles and build it today. And of course, if you need reference to the deck, you鈥檒l have it.

Quick wrap up. I know we covered a lot. I hope that it came through to you. And, I look forward to watching the recording and looking over the materials again too, as reference points. So we looked at some reflection and then we talked all about visualizing data. We talked about the process and the plan, help you really understand how it works, and build it up to something that鈥檚 more complex and unique and valuable. So the things I like to go into more, there鈥檚 some other kinds of things called logic groups. This one is the best that I just couldn鈥檛 get to till we went to. These are topics. So this only before and after adding with an exclude is something really cool that I think we can finally cover in the next session. And then we can also think about in next session or two, have we continue this series. How do you think about agile data? How do you think about what you can do with cohorts and follow and flow and journey canvas, and how do you use those along with these to kind of do the most valuable analysis possible? I have links in the deck to the documentation. I link to both, tools sets of them because they explain things in different ways. That means same concepts, but they they both have way pieces that are useful and helpful. And I will wrap it up and headed to Frederick. But thank you for joining. And please, I think a poll will come up momentarily. Please fill it out and give us your feedback back to you, Frederick.

And you may be on mute.

Thank you so much. Can you hear me now? Yes we can. Yes, yes, yes. Thank you so much. I had some problems here. Thank you so much. It was very interesting. I have posted in the chat here, the previous sessions, for anybody who wants to have a little recap on that and we look forward to the future sessions. Of course.

My understanding is that, the question that I鈥檝e been asked has been answered as well.

So thank you for answering, to everybody who has done that. And, thank you for a great presentation. And as I said, we will send out, the recording to everybody who is registered. Wonderful. Thank you. Frederick. Thanks, everyone. Thank you. Thank you for attending, everyone. Bye bye.