You go in to your shopfront and count how many people stand in the queue. Why do so many people ignore the same measurement for web sites?
Typical site measurements are of successful visits. That $250,000 analysis package some of you just purchased is based on software available free from a University in America. The $250,000 gives you a slightly better set of charts displaying flawed information. The critical bits are missing.
If your software looks at the "agent string", you are getting the least useful 50% of the picture. Forget the capabilities of the browser visiting your site. That is like predicting life expectancy in bus accidents where safety belt usage is not compulsory. You want to know what people are using.
One failed visit is worth more than 100 successful visits. Ask your marketing people about the cost of a new customer versus a repeat visit from an existing customer. The repeat visit might be prompted by a $5.00 discount voucher. The first visit might have cost $1.00 per visitor just to get people to look at the site.
The Delta text overlay problem, in the next slide, will kill 20% of visits. 30% or more will already be gone without seeing a screen.