September 13, 2008

Statistics Are Your Friend, Even When They're Bad

By Robert Seidman of TVbytheNumbers (Twitter / FriendFeed)

It should be no surprise that being part of a site called TVbytheNumbers that I’m obsessed with statistics and this obsession extends to all the web site analytics and statistics that are available to us.

While I hear and read things often about how Feedburner’s stats stink and Google Analytics stats stink and none of them ever sync up well, that really hasn’t been my personal experience. Using either Feedburner or Google Analytics as an intraday tool is certainly problematic, and I have had a day or two here and there where Feedburner did lose data for an hour of five that it never recovered, but mostly both are just slow and do recover. Google Analytics typically tracks visits and visitors correctly intraday within reasonable timeframes, but lags behind in counting total pages for hours. Usually, by 8am Pacific time (but not often before then) all the page views for yesterday show up. And once they do, on a page view basis, Google Analytics, Feedburner and Quantcast all seem like they wind up syncing up within 2%-3%.

Given everything involved, I find the 2% difference very reasonable and it doesn’t bother us any. We wind up triangulating between Feedburner, Google Analytics and Quantcast and it’s less of a hassle than managing our Web logs.

Because of the problem cited above with Google Analytics being slow to capture all the page views, it does make intraday monitoring fairly worthless, aside from tracking visits and visitors. All the other stats – time on page, bounce rate, pages per visit, etc. – are all wrong until all the page views are captured. But there’s little we’re doing that requires great analytics on an intraday basis. There are certainly times when it would come in handy, but even as it is, it works well enough intraday where we can at least figure out if we add something or move something around whether the desired result was achieved.

As a tool used after the fact, I find Google Analytics to be an extremely valuable tool, though I often don’t like what I see!

One thing we’ve thrown in the towel on is that referral traffic is almost always bad, no matter the source. There are some rare exceptions where linking produces good traffic (high time on site, number of pages per visit, etc), but that’s indeed rare. In fact, in almost every instance where a specific post is linked, the traffic is bad, with bounce rates often in excess of 80%. That’s whether Louis is linking to it, whether someone throws a link on Twitter, or even if Matt Drudge links to one of our stories. StumbleUpon and Digg show similar results.

Such traffic is great for jacking up visits and visitors, but bad for bounce rates, pages per visit and time on site. We’ve pretty much thrown our hands up in the air on that score and attributed it to web surfing behavior via links. As an aside, the stable link we have from Drudge to “TV Ratings” produces much better results, but if he links to specific story on our site and gives it any prominence on his site, the traffic has a very high bounce rate.

That seems largely out of our control, however there was still one stat that really bothered me. That was that if someone landed on our site via our home page, the bounce rates were still pretty high, approaching 50%. Better if someone came directly instead of via a referral, but still bothersome either way. Here's the landing page results for our site for August 1-31:

Recently, with that and a couple of other factors in mind – mainly wanting the ability to showcase more content on the home page – we redesigned the site. The bounce rate for traffic landing on our home page was around 47% for August. In the last week, post- redesign, that is now around 25%. The bounce rate for referral traffic to specific posts is still lousy, but again, we don’t feel like we can do much about that. Here are the landing page stats from September 6-12.

All of this has me wishing I’d gotten around to redesigning the site sooner. Who knows how much repeat traffic we may have lost as a result of design? I also feel silly because once upon a time I actually had responsibility for the web design/UI group at Charles Schwab. I recently had lunch with the VP who ran that group in my org and when I told her about the results she shook her head and laughed at me. My mentality had been this: our blog is a blog, pretty much like every other blog and designed pretty much like every other blog so spending a lot of energy on design tweaking didn’t seem like a worthwhile priority.

I definitely should’ve known better. I’m still not very happy about the bounce rates on referral traffic, but am quite happy about the reduction in bounce rates for people landing on our home page and would ascribe that improvement completely to redesigning.

By the way, for anyone interested, we went with the Live Wire theme from Woo Themes that we modified a little. So far I’d consider it the best $70 we ever spent. It’s not a perfect world, so the theme isn’t perfect, but setting the navigation structure (which we’ll certainly still need to tweak) and other modifications didn’t take much time. For $70 and time spent, cutting the bounce rate to our home page just about in half seems like time and money well spent.

Read more by Robert Seidman at