Start learning with our library of video tutorials taught by experts. Get started
Viewed by members. in countries. members currently watching.
Get a new Google Analytics tip every other week from online marketing expert Corey Koberg. Most users unlock just a fraction of the power that Google Analytics offers, so in this course Corey exposes tips and tricks to unlock insights into one of the most sophisticated tools in the marketer or site owner's arsenal. He offers peeks into the latest power features, advice for deeply mining your digital data, and actions you can take to optimize your site for both traffic and conversions. Corey answers common questions about online marketing and web analytics, including installation, tag management, reporting, custom variables/dimensions, attribution modeling, segmentation, multichannel funnels, data accuracy, visualizations, Universal Analytics, and more. What's more, Corey welcomes your questions and will shape future videos based on member requests, so send them to us at email@example.com.
In all the training and writing I've done on Google Analytics, I've never done anything on cohort analysis. It's a useful technique, but it just wasn't something that was easy to do in Google Analytics and you had to jump through a lot of hoops to try to get it done. I'm going to back up a step, and set the contacts. So, for those who are unfamiliar with this analysis. A cohort is a group of people who are tied together by some common characteristic. They're generally grouped around a certain defined time period. For example, in the last video we talked about an airline. We talked about how they might treat customers who are buying during the holidays differently.
We implied that they might be buying just for that trip home, but that they're unlikely to buy during other times of the year, and therefore maybe they're not as valuable to the airline. Now of course, we never want to make that assumption. As analysts we want to look at the data and we want to make sure that what we're saying is actually true. And that type of analysis based on people who do certain things around a certain time period is a great example of a cohort. Another example might be at a university, maybe you want to look at graduation rates or the average GPAs or something that takes that particular group of people and pits it against other segments.
So, in that case you're looking at students who enrolled at a particular time and we do that analysis on them, even as they move beyond that initial stage, we're not just continuing to look at freshmen, freshmen, freshmen. Now that's an important point. So, it's not just looking at time periods such as people who buy on their first visit, or if they bounce. We're actually identifying a group of people and then we're following them as time passes. But we're not updating that definition, we stick with that original group. Now, most people are familiar with cohort analysis from studies in the medical world.
So perhaps, it's a study on children who are exposed to a drug or lead paint. Then the analysis that follows them and their cohort is going to pretty much go to rest of their lives, may be you want to see if some long-term ill effects result it as cancer. And the idea here is to have a characteristic we're interested in such as exposure to the drug and are bounded by that time period for that attribute. Now, in our world we might think about an e-commerce site so, if it's users who have not previously seen this site, so they're not aware of it. But then they went and signed up for a trial within 30 days of a clever new Super Bowl ad that we ran.
The question is, did those users turn into useful, long-term clients or did they simply bail after that trial was over? Now, this analysis has always been useful, it's just that Google Analytics didn't make it particularly easy to determine that precise time when something happened, and then do the analysis and the broader context of the overall visitor. Remember before, our segments and everything around that were session-based. Now that we have new visitor-based segments, things get a bit more interesting. In fact, before most of the solutions involved injecting the date as some kind of additional custom variable, and then setting up some parsing schemes to create the segments.
But storing dates as text not only wastes your variable, it's really unnecessary. Google already knows things like when you first visited, and they certainly know the date of those visits. So now it's just a matter of having the interface to be able to query that information and get the reports back. As you saw in the last ones, we now have that built-in to GA, particularly in the visitor segments, and we have some specific things around cohorts as well. Pull up that interface again. We come down here. This is our new advanced segments interface. We're going to create a new segment. And you'll see down here, one of these on the left hand side is Date of First Visit.
So, this is our cohort creation tool here, the first stop. Let's take an example. We said before, the clever Super Bowl ad, and we want to see what the visitors who saw that and how they compare to say, the week before. So let's first go in here and say, Super Bowl will be February 2nd, and let's look at the week after there. And I want these to be people who signed up, so let's go ahead and say Behavior >Transactions >Per user, greater than or equal to one. We'll name this one post Super Bowl and go ahead and Save that.
And now we want to have another segment to prepare that against. So what we're going to do is we're going to create another segment, the exact same conditions in this case transactions, but for the time period prior. So we see here these are all the folks that started here the cohort and then as they trailed off, let's go ahead and create another one. Come here, Date of First Visit, and we'll do the week before. I'm not going to worry too much about making sure I have the same amount of days in each one, because I'm really looking more for quality here instead of quantity. You can look at rates and engagement, and things like that.
Come on up here to behavior, greater than or equal to one. I'm going to call this pre-Super Bowl. So, we can see our two cohorts here. In orange, we've got the pre-Super Bowl ones here. These data first visit, and then the next week as the Super Bowl comes on, we start to see the trail off of the pre-Super Bowl folks here, and the visits that were from the cohort. Now remember, it's, that they first visited here. It doesn't mean that I'm only looking at visits here. It's just that they first visited here, and then what did they do? And the same thing here. We see people who first visited after the Super Bowl, and then I'm comparing what happened after that.
And again I can come down, I can look at things like bounce rate, before and after, engagement metrics, pages per visit, duration. We could look at all kinds of stuff, average order values. These really get interesting as time goes on. In our case here, in the data set I'm looking at, we didn't actually run a clever Super Bowl ad, so there's not a whole lot of, you know, really useful conclusions to draw here. But you can see how you would set this up and most importantly, how you would define these cohorts. And there's lots of different ways to look at this. Awhile back I had a client who had to invest a fairly significant amount of resources for each and every lead that came in so, they were actually quite concerned about diluting that pool.
So unlike the rest of us who are actually trying to make our form submission, you know, that process as simple as possible and as frictionless as possible. They were actually trying to make sure that you really wanted it. They actually wanted to make that form a bit more painful. Like any good analyst, we wanted to figure out if that worked or not, and make sure that they were actually doing the optimal thing in that case. So, as they put in efforts to make the form more painful, putting in captchas, asking for more and more information. You'd want to take a look at those cohorts in each of those cases.
So the example how we might do that. I can come here and just create another advanced segment. We would put in our date range around when those particular forms are going live there. Maybe we fired an event when that can come in there. So we can actually set that here in Condition, and we could say maybe, our event action was a captcha. Or maybe there was a value that we set in custom variable. We can look at the ways that we fired that through there, and make sure that we were looking then. Now, if we only had one form running at the time.
We could just look at the date range but again, just trying to show here how we can put some additional conditions on there especially if you were running AB tests or anything else on there. And if we have a way of identifying which of those folks saw the new form, we could do it in that way. Now, there are some limitations here. We have a date range cohort that is limited to 31 days. And we also want to remember that these are user segments that we are using here, and so we know those can only go 90 days back. There are other ways to do it that are a bit more involved if you need to get around those so, ping me over Twitter if you're interested in that and if there's enough interest I'll create a video on that.
But for the interface here, just remember it's a 31 day cohort max and that it can go 90 days back for these user segments. Now, the last limitation is a significant one. They have not yet created a way to do cohorts around the date of conversion. Right now, we're just looking at the date of the first visit, which it pretty clearly says here. So, if you wanted a cohort of people who signed up during the Super Bowl, but you didn't care at all if they had been there, the first time then, if they had been there a week before, a month before, this isn't going to work.
This is for folks who, their first visit is around this day. So, it's kind of setting the cohort for you, which is still very useful, but isn't going to take into account all the different ways that people might want to do cohort analysis. In general, this can be really powerful. Quarter Analysis is something that should certainly be in your analyst arsenal. It's not perfect here yet in Google Analytics, but it's much better than it used to be and I, encourage you to give it a try.
Find answers to the most frequently asked questions about Google Analytics Tips.
Here are the FAQs that matched your search "":
Sorry, there are no matches for your search ""—to search again, type in another word or phrase and click search.
Access exercise files from a button right under the course name.
Search within course videos and transcripts, and jump right to the results.
Remove icons showing you already watched videos if you want to start over.
Make the video wide, narrow, full-screen, or pop the player out of the page into its own window.
Click on text in the transcript to jump to that spot in the video. As the video plays, the relevant spot in the transcript will be highlighted.