Website Optimization and Data Analysis
Well, we are winding down on our “lay of the land” topics and then we can start getting into the real fun stuff. Today’s topic is Data Analysis. This can be a really nerdy (not to be confused with geeky) topic. Let me just touch lightly on a few of the high points that you should be aware of, whether you are doing your own website optimization or have a consultant doing it for you.
Perhaps the most important point here is that if you are going to analyze data, you need sufficient amounts of data. What is sufficient? Well, it depends on your tolerance for error. More data will generally lead to more accurate decisions. In your college statistics classes the minimum sample size is 30, or something like that. In the real world 30 is very deficient. 100’s are good, 1000’s are better, and 10’s of 1000’s are great. Now that I have scared small businesses into thinking their web data is useless let me share a brief story.
I have a good friend that is a data analysis expert. He did some work for a local company regarding their media placement. He ran an online survey and within a few days had 200 results (not optimal). He did his regression analysis, and massaged the numbers in SPSS and concluded that the primary market for the client’s service was working moms and that they should advertise on three specific radio stations.
The advertising/marketing agency working for his client was quite upset because they had already determined that stay-at-home moms were the primary market. When a “discussion” ensued the ad agency argued that 200 responses was not a sufficient sample size and that because the survey was web based it was likely biased. My friend simply responded that a base of 200 customers is certainly more reliable than 4 “marketing guys” sitting around the table pulling ideas out of thin air.
The client opted to use the data my friend had analyzed. They ran ads on the specific radio stations (instead of taking a “throw it at the wall and hope it sticks” approach) and adjusted the message for working moms. They sold out their season tickets before the season even started!
So, the point of all that is: Making data-driven decisions is always better than guessing or thinking for the customer. Sometimes people stall on making decisions because they feel the data is insufficient, has a bias, or (in the case of web analytics) is not “perfect” data. Those are all excuses for not taking action, and if you are serious about increasing your bottomline you should be looking for reasons to act. This is especially true on the web for two reasons: 1) You can test it and get results quickly, and 2) it is less expensive to adjust a website than most other media (like a billboard, tv commercial, etc.).
This blog post is not encouraging shoddy data gathering or acceptance of amateur data analysis. In fact, following Avinash Kaushik’s idea of the 10/90 rule – 10% of your web analytics budget should be spent on the tool and 90% should be spent on the analysis. Also from Kaushik, we know the data isn’t perfect; get over it.
As a closing thought to this really long post let me just tie this back to website optimization. Your web analytics solution will be one of the primary places that you can detect opportunities for improving your website, but alone it cannot provide the analysis you need to extract actionable data. Your website optimization efforts will be more effective, as a whole, if you make your decisions based on the best data you have instead of just guessing. Save yourself some time and money before you overhaul your website or dramatically increase your online adspend and get some good data to improve your decisions.
