Tuesday, November 21, 2006

My rant on web stats

Since I sell advertising on the Crooked Cog sites I’m inevitably going to be asked for my web stats. Here’s my problem with that… they are all crap!

Just so you know the background I’m coming from... I have a degree in computer science (programming) and have been doing nothing but building/maintaining websites and web applications since I graduated. I have a good bit of experience in measuring sites and have dealt with dozens and dozens of sites in many different capacities.

Here’s the dirty secret very few people will tell you that are “in the know”… nobody really has a clue how many people are coming to their site. Sure you can get software that tracks that sort of thing and there are literally hundreds of different products out there that can be used, and they all GIVE DIFFERENT NUMBERS. And we aren’t talking small differences, we are talking differences of hundreds of percentage.

Two popular statistics programs that I use are Google Analytics (used to be called Urchin Stats) and Webalizer. The difference in their October numbers for www.TwentyNineInches.com is exactly 525.69% (Webalizer being the higher one).



If the math is a little to much for you, it’s like the difference between getting paid $30,000 a year and $157,707 a year. I don’t know about you, but that would probably change my standard of living a bit.

So a 525.69% difference between two very popular web stats software packages. You’re probably wondering how this can be and there’s a whole lot of technical stuff that goes into it. Tracking the differences between google bots hitting your site for their search engine and a real person visiting. And the differences between one person only getting counted as one visit even if he visits every day, or getting counted as a new visit every 30 mins. Those are just a couple of examples. My gut feeling is that Google Analytics (the lower numbers) is closer to the truth.

But here’s my moral conundrum… when Marketing Guy from Company A asks for my web stats, I know he’s comparing it to the stats of other websites to see if my prices are to high. Marketing Guy obviously can’t tell me what numbers are coming from what sites, and my guess is that he doesn’t even know to ask what web stats software he is using. So what numbers do I give? The bloated numbers from Webalizer or the closer-to-the-truth lower numbers? I feel like I should give him the number I believe to be more on target, however he will, through no fault of his own, ignorantly compare those numbers to the stats from another website, stats that could very well be pulled from something like Webalizer.

So what do I do at this point? Seriously… I’m open for suggestions. Recently I’ve been trying to focus on the impact of my sites, but if I dodge the web stats question it’s gonna sound shady. Then at that point what do I tell him? Both numbers and try to give him an explanation? Not exactly something I want to be explaining on the phone when I’m trying to sell advertising.

I’m at a loss at this point, but I can tell you that web stats are crap and nobody should be divvying up their marketing budget based solely on these numbers.

reposted at www.crookedcog.com

Posted by Tim Grahl at 7:04 AM


  1. Blogger James T posted at 10:03 AM  
    I use statcounter and performancing metrics to track web stats. Though the numbers between the two vary a little, the percentage increase of visits over time is roughly the same. I can get a rough idea of how much the blog is growing if I rely on both. I may not know exactly how many people visit, but the weekly hits seem to climb at a somewhat predictable rate.

    Of course, my goal with my blog is different than yours. I don’t need exact numbers to satisfy my own curiosity about who is reading. Still, it seems like potential advertisers would be interested in both sets of numbers as long they paint a consistent picture of growth.
  2. Blogger BClifton posted at 1:23 AM  
    Hi Tim,

    Just a quick comment on your thoughts about web stats. As I'm sure you are aware, Google Analytics is a JavaScript page tag based web analytics program and Webalizer is a log-file based analytics program. The two will never produce the same results - as you point out, Webalizer will track everything that hits your site, SE bots, page scrapers, translation bots, and much more leading to vastly inflated numbers of irrelevant visitors. GA will give you much more accurate figures.

    Also, when tracking visitors for marketing purposes, absolute numbers are not a good indicator because no web analytics system is perfect. Web analysts instead use trends to measure the success of their site, and I can assure you that millions of marketing dollars are spent every month "based solely on these numbers" :)

    Google Analytics Blog
  3. Anonymous Anonymous posted at 7:30 AM  
    Jim, I agree with you to a certain point. I believe there are some issues with Javascript based web analytics as well, however the main point I want to get across is discrepancies that exist. My guess is the majority of marketing people see the web stats numbers as hard and fast. I know that I've never been asked how I was coming to the numbers... they just want the numbers. So that is the frustrating part when I give them out... I realize they may be compared to incredibly inflated numbers.

    But as you mentioned, log-file based analytics sift through a lot of junk, however if done right, those can be accurate as well. There is just no real number that can be had. I want the impact and, as you said, trends measured... not straight numbers. Unfortunately that doesnt seem to be where we are currently.
  4. Anonymous Anonymous posted at 3:21 PM  
    I am novice at this issue, but why can't you just show them both numbers along with where they come from? Granted you will need extra time to educate them as you are doing here, but you also will come across as an expert on web marketing, which I think they will appreciate.
  5. Anonymous Anonymous posted at 10:36 AM  
    Fascinating - I'm no techie and I've been wondering about this myself (a bit of ego involved, of course - but not too much....). This post is a great primer for at least what questions to ask about stats and also a great fallback when talking about them. My take on stats: I had to take 15 hours of economics back in undergrad. The thing I took away (I never did get the bell curve thingy) was that stats can be made to support any argument. Looks like this might be the same thing - depends on where you look, what numbers you use and who you're talking to. Great post!

Post a Comment

« Home