1. Which pages are most worth promoting.
2. How you can improve on your future content.
But how can you identify your best content? With this article, I'll avoid
weighing in on which metric in particular is most important (be it sales,
pages per visit, etc). Instead, the focus will be on how to interpret your
analytics without relying on "most" as an indication of "best."
For example, a page that results in more sales may simply be doing so,
because it has more visits. That says nothing about whether it's the piece
of content most worth promoting.
This is going to be an advanced article, involving spreadsheets and standard
deviations, so it might not be for everybody. I'll try to keep it
step-by-step and fairly easy to comprehend but, even then, it's worth asking
how much time you want to invest in analytics, as opposed to outreach and
other activities.
Take what I'm saying with a grain of salt, and choose your time wisely. This
is going to be most useful for sites with a decent amount of resources and a
lot of analytics data to work with. With that in mind, let's get started.
Analytics
1. Start by getting as much data as possible. Head up to the top right
portion of analytics and expand your date range. I would advise expanding it
to include everything from the day you first set up analytics on your site
up to the present day.
2. In the left sidebar, click through content, site content and landing
pages.
3. Stick to comparing apples with apples. At the top left corning of
analytics, click advanced segments and select search traffic, then click
apply.
4. Make sure the data you're going to export includes more than just the
traffic. Above your graph you will see a visits vs. select a metric. Click
on select a metric and choose your metric of choice, such as pages/visit.
Building Your Spreadsheets
There's no denying it, this step's a pain. If you can build an application
to pull this off for you, I'd advise doing it. The steps below assume you
selected pages/visit, but it could be a metric.
1. Make sure you are only viewing search traffic (or a different source if
you prefer, just make sure all the data is from the same source). Click on
the page at the top of the list, then go to the top of the page and click
export. For a spreadsheet, you will typically want to select CSV.
2. Open up your CSV and scroll way down the page to the bottom of your day,
visits, and pages/visit stats. In the cell below your pages/visit data, type
"=stdev(" and highlight the data from this column, then type ")" and hit
enter. Make sure that you only highlight the data that comes after analytics
started recording data from the page. This will give you the standard
deviation of the sample, which is basically a measure of how much the
pages/visit fluctuates.
3. Repeat this process for all the landing pages that you want to consider.
I know, it's a pain and not always worth it.
4. Go back to analytics, and export a list of all the pages you are
considering.
5. Create a "standard deviation" column. Copy the standard deviation of each
page and paste it into this column.
6. Create a "confidence interval" column. A confidence interval tells you
how reliable your data is so that you can avoid favoring statistical flukes.
Excel has a function for this. At the top of this column, type
"=confidence("
7. Excel's "confidence" function requires three values. The first one is the
"alpha," which determines how accurate you want the results to be. To
understand what this means, if you type "0.01" you can expect one out of
every 100 of your results to actually fall somewhere outside of your
confidence interval. There's a good chance you don't want more than one
fluke in your data. If you were comparing 50 pages, then, you would want
your alpha to be 1/50, or 0.02, or smaller. Type ";" after you enter your
alpha.
8. The next thing Excel needs is your standard deviation. Click on the cell
from your standard deviation column, and type ";"
9. The last thing Excel needs is the sample size. In this case, it should be
from your visits column. After you click on the cell from this column, type
")" and hit enter.
10. Click on the square at the bottom of your "confidence interval" cell,
and drag it down to the bottom of your data.
11. Now create one more column, called "minimum pages/visit." Subtract your
confidence interval column from your pages/visit column to get this value.
12. Select the full table and sort your spreadsheet in descending order by
"minimum pages/visit."
That was a chore, wasn't it?
Why do all this? The end result of your efforts is that you will know which
landing pages on your site produce the most pages per visit (or whichever
metric you decided on).
Why can't you just sort it this way in analytics? Well, you can, but the
problem is that analytics (and this infuriates me) doesn't offer any data on
statistical significance. When you sort the pages by pages/visit, most of
the pages you see have just one or two visits.
If you have limited resources and don't have time to use the method
discussed above, it is possible to filter the results by a higher number of
visits. For example: Above the results, click on advanced, and change
landing page to visits. Adjust the command to say include visits greater
than 50, or whichever number you feel works best.
The problem with this is you are forced to "feel" your way through the data,
and work off your hunches. Is 2.6 pages per visit with 56 visits really
better than 2.2 pages per visit with 1,036 visits, or is there a good chance
it's a statistical fluke? You have no way of knowing without using the
procedure discussed above.
You will have to weigh your options to decide where your resources and time
are most valuable. Sometimes it's best to simply identify what appear to be
your 10 best pages and focus on them, knowing that some of them are probably
flukes. As you promote those pages you will collect more data so that you
can adjust your strategy accordingly.
Don't forget to pass this along if you found it useful.
my motto is "Keep it simple" and "don't leave anything for tomorrow that can
be done today."
Regards Gerald Crawford
Stellenbosch South Africa
Cell: +27-0720390184 (mobile)
E-mail: gerald@webcraft.ws
Comments