Anomaly detection in Report Builder

Hi, this is Jen Lasser with Adobe Analytics product management. In this video, I’m gonna show you how to add Anomaly Detection metrics to your Request and Report Builder and then take that data and graph it in a nice Dashboard. So we’re gonna recreate this graph, you see here in the bottom left of the Dashboard. And just zoom in a little bit so you can see a little bit better.

So it’s a line graph, but it also has a bar that depicts the Expected Range based on the Anomaly Detection data.

So let’s go ahead and pull the data first and then I’ll show you how to create the graph.

So this is the data that’s actually powering that graph right now. I’ll go ahead and delete it so we can start fresh in and go through the steps together.

So you want to add a Data Request and the selections you want to make on the first screen are to go to Site Metrics. Select your Report Suite, apply a Segment if needed. Choose your Date Range.

Perhaps the most important part is selecting daily granularity. Anomaly Detection will only be available when you choose daily granularity for a trended a table of data. So go to the next screen here. First I’ll pick a place to input this Data Request top left there and I’ll choose my first metric so I’ll do Page Views. Drag that over. So Anomaly Detection can be added by clicking on None Anomaly Detection And then you’ll get a few options: Lower Bound , Expected and Upper Bound will replace the actual metric, which is represented by None. Insert will insert Anomaly Detection metrics in addition to the actual data point. And these are the options we want to work with. So Lower and Upper bound tell you kind of the range of expected data and then Expected values tell you exactly what was expected within that range. So to get the chart that I just showed, you want to add in Lower and Upper Bound. So notice that there’s three metrics in the requests now, None is the actual Page Views and then Lower and Upper Bound are the Anomaly Detection Page Views. So we can add in a couple more metrics if we like.

I’ll add in Visits and do the same thing Anomaly Detection, Insert and I’ll do Unique Visitors as well.

So now that we have our Data Requests we go ahead and click Finish and let this run really quickly .

OK .Great . So I’m actually going to build the chart off of this column of formulas, I’ll explain this in another video, but you can get pretty creative with dropdowns in excel so that you can allow your users to chart either Page Views Visits or Unique Visitors. For this we’ll just stick to the one that’s showing here so that’s Page Views, but be sure to check out that next video around using dropdown controls. So we have the data, we’ll go back to our chart here. It’s all pre-populated so I’ll walk you through how I created this.

So it’s actually three different Series, The actual data is pulling from that None column. So we’ll go ahead and go look at that really quickly. So that is the None data that came out of the Data Request or the actual Page Views .

What I’ve called Expected Range is really the Upper Bound. So if you go and take a look at what that is in the table it’s the Upper Bound that was returned from the Data Request. I’ll explain that naming difference in a minute. And then the third Series is the Lower Bound.

So what we’ve done to create this chart here is the actual is just a line chart and the Expected and Lower Bounds are Area Charts with different coloring. So Lower Bound is an Area Chart you can see it outlined here that just has a clear fill so that you don’t have any color for it. And then the Upper Bound is this darker color Area Chart. I’ve called it Expected Range because when you overlap the Lower Bound on top of the Upper Bound you actually get what the Expected range band is for Anomaly Detection.

So just to recap the order of the Series matters, it gets you this nice overlapped chart. So make sure to when you build this do Actual, Expected Range and then Lower Bound.

So hope you guys found this helpful Highly recommend adding an Anomaly Detection anytime you have a trended visualization it just adds a lot more depth to the data that you’re sharing because it really roots the actual data in historical information as well as expected kind of future behavior.