As I’ve said before, I really enjoy my job. Perhaps the most enjoyable part of it is when I discover a sudden and unexpected emergency and it is entirely up to me to fix it.
For example, I came into work yesterday feeling great. There is little doubt that my life is awesome. The site is kicking it in overdrive. Plenty of real work exists for me to do. And, assuming everything works out the way I’m planning, we’re going to give away an Ipad on the Lubbers Facebook Fanpage, based on the high score for one of three games. BTW, if you haven’t, I highly recommend going to the fanpage and voting in the comments for Bejeweled, Pac Man, or Tetris. Once the choice is made, it’s not changeable.
Also, please don’t vote for Bejeweled. I’ve never played that game and have no idea how to get a high score in it.
Anyway, things were looking up. Then I looked at last week’s web numbers. And the world went to crap.
I have no idea why, but the numbers had dropped by approximately 2000 page views. That’s a BIG drop! Naturally, I freaked out. Outwardly, of course, I just sat and stared at the screen for a while, but mentally I was definitely hyperventilating. (I’m not one for visual histrionics.)
Immediately, I threw my brain into hyper-drive (or whatever it is that we humans do), and tackled the problem. First, what’s changed recently that could have that kind of huge effect?
Oh, of course! Two weeks ago I put into place a new way of listing pre-owned vehicles.
Awesome, is that the cause?
I have no idea.
Well…, figure it out!
And so, for the next 3 to 4 hours, I drew up elaborate excel spreadsheet charts using approximately four weeks of data, trying to ferret out evidence that the listing change was the cause of the problem.
In the end, it seemed clear that our users were not actually down by very much (though they were down), and the degree by which they were down did not account for the full decrease in page views.
This suggested that, unless potential customers had way less time to screw around at work while looking at our vehicles, a web-based factor was the root cause.
Still, I had no way of knowing whether the vehicle listing change WAS that cause. Logic seemed to dictate that it should be, but experimentation and testing are the true key to website success. I considered using the Google Web Optimizer to run a series of either/or tests on whether linking to one Vehicle Listing Page or the other would result in significantly more page views, but realized that the back-end tool provided by Reynolds and Reynolds (our delightful web provider who is REALLY dragging their feet on AJAX list searches, even after I demonstrated how awesome AJAX Vehicle Listings could be*) would not allow for multiple home pages, a necessity of the web optimizer.
So I went a little more old school. Two primary links (plus like 6 less primary links) point toward the Listing pages. Yesterday, I just switched one of the primary links to the old Listing Page. At midnight tonight, I’ll reverse the two. After another 24 hours, I’ll add up and compare the results. In theory, page views should be fairly similar, accounting for weekly variance, which is usually pretty small from Tuesday to Wednesday. If they AREN’T, however, then I’ll have a culprit!
Reason #72: I use the scientific method and understand why it is useful!
Edit: Quick update. I’ve just discovered the miracle of the Exit Rate statistic on Google Analytics. This stat tells me if a page that directs users to other pages fails in its job by instead confusing or pissing off users and causing them to leave the site.
So far, New Inventory Page has a 19% exit rate. Old Inventory Page only has a 13% exit rate. 6% COULD account for the difference.