I’ve been working with (Google/WebTrends/Omniture) Analytics data for 4 years now and the requests for Analytics data usually come in 2 styles.
1. The super basic: just tell me my site is still running, whatever everyone else looks at.
2. The super detailed auditors: tell me each of 180 customer segment’s data sliced and diced 10 ways and the month to month change, YOY change and a dozen other things the software doesn’t calculate for you. This could take months to implement and most of the time they have lost interest in it by the time you get it working properly.
I get frustrated with both. The super simple manager needs to look at more than just visits from month to month. The uber detailed guy needs to hire a developer to implement all that and not make changes each month to how they want data tracked and processed because all the time is spent on implementation and none on analysis and most of the time nobody even looks at all the 180 segments of reports.
They also need to realize that all the systems take the data and summarize it or cut off tracking at a max number of log files, web pages or analysis processes to maintain the integrity and size of the database tables. Try and do a full audit of every page view and click and you will crash WebTrends and re-processing it can take months. Google Analytics doesn’t even give you options to do more than what they summarize. Omniture really tries, but its a slow slow process.
Instead I am supporting the idea that web analytics data is really about trends and not audits. These numbers will never match your server logs perfectly nor your clicks from campaigns and that is OK. I also have listed here 5 metrics to look at and why they are important for your online site. One caveat is that I do not work for an e-commerce website so that has not been our focus. The focus is on conversion to application for recruitment purposes for companies.
1. Visits - yes month over month traffic is important. What is more important is to look at the difference in traffic and drill down into what gained or lost traffic in the way of pages/content on the site and what sources changed in their contribution of the total traffic. This is actionable where as just visits aren’t. Also check back with the costs for each of these budget areas and compare the cost per visit provided by each.
2. Referrers - in a nutshell you should know how much traffic is coming from search, direct and your advertising/marketing plans online and offline. Within those groups you can drill down further but the direct category is always problematic because many analytics packages track page pop-up forms as new visits as well as returning to the site after a conversion process. Also remember that a session is usually 30 min, after that its reset as new.
3. Implementation - no this isn’t a metric but it is a focus you should have on a monthly basis to make sure new sites, pages get tracking added, new campaigns get tracked and that you keep researching new technology developments with your analytics package that may change everything. Having a good web developer along that has access to the servers and can make these changes is key if you’re not a developer (and no developers don’t make good analysts, a best case scenario is a dynamic duo where they are paired up and both work on projects together and learn from eachother) and the helpdesk type services available through Google are non-existent so good luck there interpreting the overly simplified online tutorials that don’t match what your clients want or answer your client’s specific needs/questions. WebTrends and Omniture are slightly better with web support but they expect you to pay a lot for it. A good independent consultant may be the fastest most reliable way to go here.
4. What people search for on your site. This can be tricky to implement but if you get this data it can be very telling. if people can’t find something on your site and search for it, you get a window into what they were thinking. this may tell you that the content you have isn’t what they want or that it isn’t as navigable as you thought. New product ideas also come from this data.
5. Where people exit from your site. This is classic application drop off analysis within any online linear process. But guess what? People don’t always think linear-ly. Expect some of this data to drop off in chunks but a small amount to drop off at all points for unknown reasons. Its more actionable to focus on the large chunks and look at each page and the click maps for them but sometimes only so much optimization is possible here without doing real life usability testing with 5-10 people.
I’m sure there are more things that people can look into with geographic data and time on site but sometimes I think those are less actionable because you have little control over where your ads run because geo-targeting doesn’t always work well (excluding more than it includes) and time on site can be good or bad at short and long times. The content/pages that are popular on your site are also important but this is one of those custom setups that each division will need tracking by their geo-location and they never admit that so much traffic cross pollinated from each other’s campaigns. You have to read into the specific needs of your client to see if these apply and how to evaluate them without over complicating the reports. I really believe you should look at 5 key metrics or less in a report, more than that is not actionable and is distracting from your purpose/process of improvement.
There is also a difference between researching a question/metric once, and doing it monthly when it never changes. I don’t believe its a good use of time to report on time on site if its been consistent month over month for the last 2 years. Check in once a year and leave the other data to be reported monthly, save the analyst’s energy for the new questions that need answering and trust your site.
What else do you think is applicable? Any feedback?