Posts Tagged: Segmentation

Feb 10

Segmenting Data with a KPI Overlay

This week in the Web Analytics Master Certification course, we are still talking about Segmentation, but we are looking to drill down further with the data to provide more meaning, and get to actionable insights.

One way to provide a huge boost of meaning to a chart is to combine two sets of data to see if there is a relationship. With Excel 2007, it is fairly easy to combine two sets of data and make a nice little chart. If you have never done it before, Mathew McDonald, the author of Excel 2007: The Missing Manual has a video that describes how to do it here.  Combining data sources can really bring data to life, which is good good good.

For part of my assignment this week, I decided that the most important KPI for my subject website (in addition to overall number of conversions) is: Profit Margin / Sale.  Meaning, it is great if we made a ton of sales, but how much are we actually earning from those sales?

At a first glance, it is important to know where our visitors are coming from.  If they are coming from organic search, links from forums, links from social media, youtube, etc., of course it is cheaper than if they are coming from PPC (Google Adwords in this case).   Perhaps there is a relationship with our Profit Margin / Sale and our traffic sources.  Let’s take a look.

This chart displays the number of visitors from each source and compares to profit margin/sale.

KPI Profit Margin per Sale with Visitor Sources

Looking at the chart, there appears to be a relationship of some sort.  Overall traffic is heading up with Organic Search leading the way, and profit margin is heading slightly up (in general) over time.  There are some dips that don’t seem to make sense though…

We can go deeper with this by picking a few additional segments to look at.

In this case I chose:

  • Number of Conversions that were for “High Margin” products (those that are >65% margin).
  • Number of Conversions that included two orders in the same session.  (thereby reducing shipping costs — in this case the retailer is offering free shipping, so he can usually have a better margin if he can ship more than one item in a single box, unless that item is a fully assembled motorcycle or something).

This chart shows “high margin conversions” and “multiple order conversions” overlaid with profit margin.

profit by visitor type

Looking closely here, it seems the relationship for profit margin to conversion type is much stronger.  Of course we can throw the data into excel or minitab to evaluate the “correlation coefficient”, but first glance, the relationship seems evident by the chart.

Some may argue to do the statistical correlation first, to which I wouldn’t disagree… but the point here is that you can get some pretty powerful information with just some raw data and excel.  HiPPOs (as Avinash likes to call the Highest Paid Person’s Opinion)  care about statistics, but they don’t like to see them.  They would rather see a chart that tells a story.

Please feel free to share your thoughts on this and any examples of using data overlays in your analytics.  Would be great to see some other examples.

Jan 10

Web Analytics and Segmentation

This week with John Marshall and Avinash we are talking about Segmentation of data.  When looking at some of the default reports from Google Analytics and other tools, it is easy to get caught up in the numbers that are shown:

  • How many Visitors?
  • What is the Bounce Rate?
  • What is the Average Time on Site?
  • What is our Total Sales for the month?
  • etc. etc….

Here is the standard “Dashboard” from Google Analytics.  It is nice… but without any segmentation, we are looking at “faceless” numbers.  They tell us some basic things, but we are missing a big chunk of the story.

Let’s say we are looking at the common KPI of Visitors.  How many visitors came to my site?  Looking at the dashboard, we can see something that is like a bar graph going up and down.  We can see how many visitors came each day, each week, each month.    This simple graph below shows visits per month.

Somewhat useful, but kind of boring.  It is just aggregated data that doesn’t tell us much of anything except that we had a bunch of visitors month after month, and it went up and down.

This is where is gets really interesting.  By segmenting this, we can see how many visitors came each month by what source. Very cool!   Were they from organic search? PPC? Affiliates? YouTube? Facebook? Stumbleupon?  Where did they come from?  By segmenting this simple graph, it now becomes alive and tells a story.  We go from the above boring graph to this more interesting one.

Now we are getting somewhere.  Here I can see that my “Search” segment was growing at a decent pace (nice job SEO guys) until October, when it took a precipitous drop. What the heck?!!  Oh, wait, that is when Google changed their search algorithm in a big way and my SEO guys dropped the ball.

What else?  Affiliates traffic rose at a steady pace through the summer, then started to drop a bit in the fall.   “I’d better talk to the Director of Affliate Marketing to see what changed towards the end of the year… ”

The value of segmenting your data can not be underestimated.  By segmenting, you bring your data to life with actionable insights.

In Google Analytics, you can create custom segments, or use some of the default ones.  To set them up, you go to your main dashboard, and down on the left bottom under “My Customizations” you click on “Advanced Segments.”  Then to see the default ones the GA team created, you click on “Default Segments.”  Some that I’ve set up are shown below.

The great thing with the segmentation tool in GA is that you can set up these custom segments that query for specific things much in the same way you would set up an SQL query in a database, except much easier. Query by “contains” “exactly” “greater than” “equals” etc.

One that I’m excited about is iPhone visitors… it will be interesting to watch that metric as mobile traffic continues to grow.  You can then dig deeper into your site and see what pages bounced your iPhone users and perhaps make adjustments to those pages.  Maybe the graphics were too large, maybe there was some flash content, etc.