During the Emetrics Summit, we also had interesting discussions about the top 10 KPIs and the inevitable discussion about what could be considered a KPI and not. I'm not as purist and tend not to bother too much when clients consider some figures to be KPIs while they actually aren't as I believe that they have to figure it out, with our help, slowly but surely. The Web Analytics learning curve is long so it's often easier to go along and let the clients find out by themselves after a while. I often hear "but that's not really important" after a couple of months in a project and that shows me I'm doing a good job.
So, Eric Peterson defined a KPI as a percentage or a ratio. I can understand that for some business models, pure figures can be considered KPIs as well. A website based upon the advertising model will consider unique visitors as a KPI as that is the information that will make their clients decide to choose their website in place of the competitors'.
Of course, the important thing will be to see whether the number of page views per visitor have actually increased or not (if we are still taking about such a business model) and the frequency of visits and we could indeed put this into a percentage. Note that seasonality will also be an issue so an increase in terms of percentage on a monthly basis might not be good news for July and August.
Anyhow, getting back to my KPI list.
- surprise, surprise: the ability of defining visits and unique visitors as accuratly as possible. I can't say it enough: tagging + first party cookies. Do me a favour, implement that, stop talking to me about accuracy and let's dig into the figures;
- search results & abandonment is also becoming a quite easy one as Jason Burby from ZAAZ wrote a couple of interesting articles about that and most sites do have a search box so use this information to help your clients find the information they're telling you they're looking for. Today we set this up quite easily and streamine it towards the people responsible for content. We also often plug scenarios on top of search results pages in line with the sites objectives in order to send the visitors towards the paths we'd like them to take. After all, this search box has a cost. The best it could do is serve the website's objective in order to become cost effective;
- Once you've got that rolling, plug your referrals also to that in order to determine how people are finding your site BUT don't engage into traffic acquisition before you've fixed all the bucket holes on your websit through;
- conversion! My favorite topic😉 We discussed this during the dedicated workshop on the Web Analytics Day held November last, where I quoted Bryan Eisenberg in his definition of conversion: "a conversion rate is the measure of your site's effectiveness in persuading your visitors to take a desired action." This of course means that you inevitably know your websites' goals and objectives. During the workshop, we had listed a couple of conversion events such as (non exhaustive list):
- make a purchase
- opts into a Newsletter
- submits some type of personal information
- subscribes to a RSS feed
- prints a page
- uses "email this to a friend" functionality
- spends more than 10 minutes browsing the site
- downloads a document or an application
- looks at a set of important pages (for WebTrends, use the key pages parameter)
- views a set number of of pages during a visit
- clicks on a particularlink to leave the site
- search for a specific product or piece of information
- Cost per lead, costs per savings: provided you get this information, that's for me the killer figures showing exactly where you are, how you're doing with your online communication, compared to other acitvities or channels. I love that one but find it quite hard to define as information retention is sometimes @ hand but for those who can face the facts, this is the way to go as it's only through these kinds of methods that you'll be making the right choices and realyl learning by doing!
- Reach is a metric I've always found hard to work with as I know how many people are viewing my website but I can't determine the total population. Natalie from Business Objects cleverly remarked that reach could be measured for an Intranet as there, you know your entire population. So true, didn't think of that one, thanks, Natalie! Note that marketshare was also mentioned but at the time of writing, I can't really see how this can be solved using Web Analytics. Unless as Siegert wispers to me, you have a centralising association such as for example the CIM here in Belgium that gives figures for your traffic and your competitors.
- Stickiness for those websites who need people to hang on for some time and thus tend to focuss on increasing the number of pages viewed and the time spent on the website. Good if you're selling advertising space. No if you're a self-service site, off course as it could mean that people are not finding what they're looking for.
- Measuring offline activities using online support. That's getting hot for us lately as more and more of our clients are using unique URLs or landing pages for their offline communications. A definite way to go for the year to come if the online channel is streamlining responses coming from off line channels. Another one used is the unique telephone number of the call center that can only be found on the website, all other channels using another phone number. Always interesting if you want to improve website use and decrease call-center costs as it's a good place to start and easy to set-up.
- Content ratings by visitors suddenly popped-up in the top 10 list as Web 2.0 is heading more and more our way. It will be interesting to see the evolution of measure of visitor contributions.
- Last but not least, profit, net present value and lifetime value, the holy grail of metrics that directly impact your companies' bottom line. Note that these figures will not directly be rendered by your Web Analytics tool and efforts but can give you a nice advantage if your WA is used with intelligence and intellectual curiosity (yes, I'm preaching for A/B testing again). Lifetime value for me is not something I would suck out of my thumb or my WA tool like that but an interesting goal for the months or maybe years to come. It always makes me wonder about predictive modeling and actors such as SPSS…