The reading level for this article is All Levels
How Web Analytics Helped Find A Million Dollar Hole
I’m often asked how we go about using web analytics to really pinpoint problems that make the tools worth the investment. Many people are dubious when asked to fork out 50,000 a year to have reports about how people visit their website. What I’m about to describe is a situation where one of our clients could potentially earn $1 Million per year because of the analytics tool they have installed. This article will describe how we used a key performance indicator to raise the problem and then go onto describe how we then found out what the issue was on the clients website.
The Key Performance Indicator or KPI
Ahh the KPI! It’s the latest buzzword flying around in the industry. What key measurements to use is an important point, but much more important is how you use them. One KPI I have written about before is page views per visit. Page views per visit is a KPI I always use regardless of the site goal. The reason being it’s what I refer to as a tripwire metric. Like a tripwire it gives you a warning when something is not right. The way you should use this KPI is to first find out how many pages it takes to complete the desired action. In this case the desired action (a purchase) took 7 pages. Then consider what a good browsing experience might be from your point of view. In this case we figured if the visitor viewed 5-7 pages and then completed a purchase (another 7 pages) it would be a good visit from the businesses point of view. It means that the visitor finds out that more is on offer than simply the product they were looking for. Then we add another 7 pages on top of this to flag a too many pages warning. So we have a bottom limit of 7 pages, a happy medium of 14 pages and a top limit of 21 pages.
Why is too high a warning?
The next job was to figure out whether this was a good thing or a bad thing. If an average visit took 22 pages it meant that either the visitor was happily browsing around and our client should be very happy, or it meant that there was a problem and if so we needed to find out where.
Good or Bad? Happy or Sad?
that follow the same behavior patterns. We wanted to know if the visitors were flicking through pages very quickly (a sign that they were unhappy) or if indeed they were traversing a great many pages each and spending a normal amount of time on the site (a sign that they were happy).
Therefore we segmented the visitors into only those that spent less than 2 minutes on the site and those that visited the shopping cart. This would enable us to see if the page views per visit of those visitors only on the site for a short period of time were racking up lots of page views or whether it was those that hit the cart that had trouble finding what they wanted.
Less than 2 minutes showed normal behavior. The people that spent less than 2 minutes on the website generally browsed 2 or 3 pages per visit. The people that hit the shopping cart again went off the scale but this time it was even more problematic. The average page views per session was 58 pages. We’d found the people who were having problems.
58 page views per visit?
Since we’d found the visitors who were having the problems we now needed to know what they were doing. How on earth could people be going through 58 page views on average each? It seemed unlikely – we even asked the developer to check that the tracking code was correctly installed. However when we checked the path analysis the problem on the website became crystal clear.
One visitor had traversed 97 pages. We looked through his visit path and noticed that the path kept referring to one page – a search results error page. We checked other individual visits and noticed the same key trend – the search results error page.
This lead us to check the failed searches on the website. When we totaled them up there were over 2000 failed keyword searches and the big majority were product codes. The sites internal search engine simply couldn’t read a letter and number combination and most of the product codes consist of numbers and letters. We’d found the problem.
The solution therefore is to fix the search engine. This one fix is a potentially huge find. There were over 1000 people that keyed in those failed keywords and didn’t complete the purchase. Our customer brings in over $160,000 per month in online revenue from a little over 1700 people that did complete a purchase. That means by doing a little mathematics it’s easy to work out the potential. It’s well over $1 million a year in lost revenue.
It’s easy to worry about the cost of a web analytics system. They are expensive and with everything that most businesses have going on they aren’t easy to get the most from. What you really need is an in house expert looking at the systems to pin point the problems or outsourcing to a consultancy to get the most from the systems. However to not use web analytics is like throwing away money – a frustrating and expensive waste of time.