Excellent feedback so far and some answers to your questions!
Thanks to everyone for checking the new blog out and for the comments coming in so far. I added an XML button on the right so people can subscribe–I’m going to try FeedBurner again to see what kind of stats they’re able to generate (feed metrics are a hobby of mine.)
Neil Mason asked:
- Eric - are you going to focus on site centric KPIs or are you going to widen the field? For example, what about customer satisfaction, reach etc? Are you planning to cover these as well?
Absolutely! I hope to be able to cover a number of non-traffic and commerce related key performance indicators including customer satisfaction, site performance and response times (e.g., Average Time to Respond to Email Inquiries.) It is my firm belief that once companies get up-and-running with web KPIs the next place they should be looking is at the web-as-a-business.
Sam and Jerry had commented about the use of average as opposed to something that communicates more information (e.g., median). I had already written (but not blogged) a section header about averages that hopefully covers this. In general I agree but KPIs should be easily and quickly calculated. Does anyone have a simple strategy for calculating the median value for indicators like these?
Most of the other comments have been KPI-specific and are great! I’m going to use my Gmail account to keep careful track of everyone who contributes and will do my very best to acknowledge everyone in the final draft.
No, that does not mean I’ll be sending you a check ;-)
Phil Aaronson added the following ...
Median, for a lot of distributions won’t help much (… and all that sorting). Chances are pretty good that there’s a huge tail here and the median value will be 1.
A better indicator might be the % of 1 page visits?
Jerry Hosking added the following ...
Does anyone have a simple strategy for calculating the median value for indicators like these?
Median Pages Viewed per Visit = (n visits + 1)/2th ranked pages viewed values
Well, I guess it’s not presented sufficiently simply if you still have to explain it.
I can do it, I just don’t know how to write it down.
- Find number of visits for time period
- Add one
- Divide by 2
- Sort values for pages viewed in each visit (descending)
- Choose nth value from the sorted value list where n is the result of (number of visits +1)/2
Of course, this is just the simplest of median calculation methods. Others get more complex depending on the data.
Don’t know if this will help, but it might give you an idea to build on.
Jerry