Five Pillars of Great Data for the Enterprise
Putting together an enterprise analytics implementation is no small task, especially when the project encompasses sites around the world. There are many considerations, beginning with questions like:
How is data collected now?
What does the site roadmap look like - will the sites basically stay as-is while the data improves, or are sites themselves being updated?
How responsive is the organization's IT?
How centralized are IT, analytics, management, etc.?
How digitally- and data-literate are site managers around the world?
Who's going to be working and/or consuming the data?
After years of doing these projects, with a wide variety of organizations that would answer those questions in completely different ways, I've settled on five "pillars of good data" that help focus an implementation. They're the goals I shoot for in any data implementation, and they're based on the simple premise that insights can only be as good as the data they're built on. These are the 5 pillars that support great data:
Efficiency
Accuracy
Consistency
Richness
Accessibility
Read on for more details...
1. Efficiency
Efficiency is the foundation the whole implementation should be built on, for one simple reason: the easier it is to make updates, the more likely it is that your data will actually BE accurate, consistent, rich, and accessible. Lower the number of touchpoints required to make changes, by using tag management as well as possible. Lower the number of places people have to go to get good data, by creating rollup views. All of this will lower the time spent on management, and as we all know, time is money - the longer it takes to manage the data, the more likely it is that corners will eventually be cut. Future-proof your implementation by being ruthlessly efficient on all fronts! (More to come in future posts about how to be most efficient with GA & GTM in a variety of scenarios...)
2. Accuracy
"Accurate" is a difficult word, because as all data nerds know, answers can only be "accurate" when questions are very, very, very specific. So, when I say "accuracy" here, I mean two things:
Of course, taking care of details that prevent data errors and "gotchas" like broken sessions, odd pageview counts, and/or events that don't mean what you think they mean.
More importantly, "Accuracy" here means having a setup that facilitates QA by people that know what to look for, and a streamlined way to globally deploy fixes. If you can't fix it, what's the point in testing? If you don't test it, how do you know if it means what you think it means? Test it! Fix it!
Tag management is (again) your best friend here. I love, love, love Google Tag Manager for a variety of reasons, but the possible workflows for debugging and deploying fixes are way up there on my list of favorite things.
3. Consistency
Collect the same data in the same way wherever possible. Obviously, different sites will have different functionality, or at least different architecture, appearance, and coding. It's counterproductive to shoehorn those kinds of differences into rigid specifications...BUT, there are things that all sites have in common, but aren't tracked by default in GA. For example:
Track all outbound link clicks (with a single tag!) in an event category called "outbound links" - it's easy to do with GTM.
Track all clicks on PDF links in an event category called "downloads" - again, easy to do with GTM.
Track all navigation clicks in some combination of categories that all include "navigation" - ie "top navigation", "side navigation", "featured stories navigation", etc. This will make it easy to search later.
Across all of these tags, use Event Action = {{Click URL}}, and Event Label = {{Click Text}}. Now, in GA, you can just switch to an Event Action report to easily understand all of the ways that clicks to a certain address could be done.
4. Richness
Go beyond standard pageviews – capture navigation clicks, video interactions, downloads, form submits, search terms, etc. With tag management, this is simple, so just do it! It's only a couple of tags. In addition, of course, think about custom metrics and dimensions. For example, I always include the following dimensions:
Site/Application name
Country (2-digit code)
Language (2-digit code)
Country::Language (the two above, joined by "::")
GTM Container ID
GTM Container Version
#s 1-4 above will be very useful when it's time to split data into different views within a single GA property, or for reporting on Market/Language combos within an international property. #s 5&6 are fantastic for debugging.
5. Accessibility
Provide a simple UI that’s understandable to site owners (GA for the win!), and find or create a support model for user administration. I absolutely believe in the democratization of data, rather than the "walled tower" approach where reports have to be requested from a central analytics organization. Data literacy continues to increase, and there's absolutely no reason to put roadblocks between people and their data (well - sometimes there are political reasons, but those should be navigated).Simultaneously, provide rollup views that easily bring data from different sites together for a global perspective - because someone should be paying attention on a global or regional level. Also, don't forget about training and education, particularly around communicating the non-out-of-the-box capabilities you've put in place.
Conclusion
Efficiency enables good data to exist now, and in the future. Insights are only as good as the data they're built on, so data stewards must be sure that the data is accurate, consistent between sites (to facilitate a broad view), and rich (to create new opportunities for analysis). Finally, making good data accessible will ensure that the interest in, reliance on, and demand for data - and data-driven insights - will grow.