Friday, October 24, 2008

How Many Readers do You Really Have? Find Out.

Google Analytics has a lot of useful pieces that many people don't understand. A lot of this has to do with hiring too many tech writers and not enough "explainers," leaving a lot of their guidance murky.
One section of GA bears discussion: Loyalty. It's a sub-report under Visitors. Click VISITORS>LOYALTY>VISITOR LOYALTY. You might really benefit from this post if you open your GA and follow along. It's not going to give you a headache, I promise.
Loyalty measures how often people have visited the site. It's just a bar graph showing that some have visited three times, some 50 times, and so on.
You can select for loyalty during a given period. Want loyalty for July? Set your date period to July and get loyalty for July visitors.
What is notable about Loyalty is not what it is, but what it could be. Within the loyalty metric is the foundation for measuring readership with more accuracy than we now have.
Sites that install GA are usually either merchant sites, selling something, or content sites, like comics. GA is highly sophisticated when it comes to tracking a customer's moves from landing page to store pages to shopping cart to payment. It's quite primitive, by comparison, when it comes to measuring readership of a comic.
One way people measure readership is by looking at their visitor graph, noting whether it has risen or fallen, and closing the window. This is, of course, wrong. They are confusing visitors for readers, and readers are much more valuable.
Another way of measuring readership is to click VISITORS>NEW VS. RETURNING and up comes data and a nice pie chart. Many people use the RETURNING figure to calculate their readership, on the assumption that people coming back are there to read the comic.
But it's not that simple. People coming back might be ripping off your art for a homework assignment, sizing you up for an ad placement, looking to see if you're still linking back to them, taking a screen shot of your masthead so they can identify the font or could arrive for many other reasons. Some may have clicked the same ad several times over a month, forgetting each time that they don't actually like where it leads. Some may have clicked a link by accident.
So, we discount the REPEAT VISITORS number. Or do we?
In order to be counted as a repeat visitor, the visitor must be carrying a cookie issued on a previous visit. If the computer user has deleted that cookie, they will go into the new visitors column. People who mostly leave cookies intact are estimated to be around 81% of the population in the conservative Redeye Study.
We might find ourselves knocking off 20% of repeat visitors for the reasons I described, then putting a similar number back to compensate for cookie deletion. (I think pulling out 20% is way too high, but it keeps my example simple.)
Another problem with repeat visitors is that we might have 80 people who visited twice or forty people who visited four times, or almost any reasonable mix of visits and visitors.
Here is where Loyalty comes in. The most important point: you must measure visitors from the same time period as you measure Loyalty. If you measure Loyalty from a different time period, your data will mean nothing.
Let's move toward figuring out how many readers you have. I think we're ready.
We'll compare repeat visitors with loyalty and see what we learn.
We'll use the standard thirty day measure for our example. First, a decision. How many visits per month count as a reader? I asked this question of users of The Web Comic List,  and they were pretty split between 1/week (4/month) and 2/week (8/month).  We'll use the 1/week figure, mindful that doing both might be wise, and that update frequency differs between comics, making 2/week high for some titles. Open your calculator widget if you're participating, and add up all the visits from 4/month and up. If you find your top numbers are ridiculously high, there could be these problems: you might have set the time period incorrectly, and you might have forgotten to tell Google not to count your own visits to your site.
For the site I'm using, I got 4215. That's the number of visits from readers, we hope. The number for returning, meanwhile, is 5106.
I don't like that difference, so I'm going to note the number of repeats who visited less than four times: 891. The actual difference between the two number above is SURPRISE 891. The numbers are in agreement, only our definition of who is a reader makes a difference.
This particular site has a subdomain, so I'll check the process with those numbers. Returning: 1786. Returning weekly or more: 1113. Difference: 673. Uncounted returning visitors: 673. Again, they match up.
Here is how I suggest you use this data. Being scientific and thirsting only for truth, you want to know how your comic is doing. You have no interest in self-deception. Anyone still reading?
Calculate your reader number each month. Keep track of it on your computer in a file, and in a notebook, because you don't want to do it all twice even though it is worth it.
Each month, measure the change. Hopefully, it will rise each month, and your data will look something like this:
July +45%
August +39%
September +16% and so on.
Don't be alarmed if you go through a period of doubling and then growth slows down. It's much harder to double 200 readers than to double 2 readers. Do take note of trends, however, which can alert you to problems with your comic or your site or your work ethic.
Remember earlier I mentioned all those people who screw everything up by deleting cookies, even though cookies barely require any disk space? Intuitively, they seem to outnumber the false positives from people visiting your site multiple times for reasons other than reading your comic. This suggests that readership may be higher -- 10%, 15% perhaps -- than we have counted.
Before you "round up" to accommodate that possible margin of error, consider a few things. One is that I used the Redeye Report to get a low number, since they want to sell you an alternative type of analytics that only works for outfits like Wal-Mart. Still, their report seems solid, so we won't blame them.
But where are the other 19% of readers? I don't know, but I have a theory.
Google has access to one of the largest databases in the world. I can assure you that they know pretty well when cookies have been deleted, what types of users (males, tin foil hat types, maintenance freaks) delete cookies more, and many other metrics. My guess is they can weight your analytics to account for the cookie deleters.
News of this massaging might not go over well with the public, which is filled with people who make it their job to misunderstand things.
Comparison with other analytics doesn't work, because they're all shaky to some degree, especially on smaller sites, and you have to watch constantly for false readings.
Soon, Yahoo Analytics will be available, a result of their purchase of Index. People are going to rush to try it out. If the analytics are good, and Google massages and Yahoo doesn't, Yahoo will look bad. Same thing if you do it the other way.
Right about here is where it gets easier to wait and see what happens than to try to out-think two Fortune 500 companies. I do hope one of them is smart enough to add a good reader measure to their analytics, and soon, to save having to watch your eyes glaze.
As if this post wasn't long enough, I'll mention Comic Rank, a site from the UK designed to count readers. Anyone participating in CR and GA who isn't already working with Steve (the ComicRank guy) and I to explain why the CR count is so different from Google's, please get in touch with me and add your data to our analysis. Do keep in mind: You must multiply your CR number by 3.3 to equal a Google 30 day period. If you're like me, you'll find CR is way lower than Google. If you have a new comic, the difference is significant but less. (So it's not just a multiple.) Steve and I have been going nuts trying to figure it out, so this is like giving blood. You could save lives. Or, figure it out for us, and we'll do a little oil painting of you for the cover of next year's fashionable desk calendars.