When you look at the evolution of digital measurement in the enterprise and study organizations that have achieved a significant degree of maturity, you’ll notice that they come in two distinct flavors: the analytic and the informational. Analytic organizations have strong teams studying the data and driving testing, personalization and customer lifecycle strategies. Informational organizations have widespread, engaged usage of data across the organization with key stakeholders absorbing and using data intelligently to make decisions. It’s not impossible for an enterprise to be both analytic and informational, but the two aren’t necessarily related either. You might expect that organizations that have gotten to be good in measurement would be mature in both areas, but that’s not really the common case. Instead, it seems that most enterprises have either a culture or a problem set that drives them to excel in one direction or the other.
When people talk about data democratization, they are almost always focused on creating a strong informational organization. Analytics teams have found that as they grow and as the organization matures, the demands for data become impossibly widespread and frequent. For a centralized team, trying to answer every question around digital in a maturing organization is simply impossible. At a certain point, therefore, analytic teams begin to concentrate on data democratization – providing direct access to the data out to the organization and building the culture and foundation of digital measurement up to support effective use of that data. They do this because, for the most part, they have no choice.
This journey to data democratization follows a fairly well-worn path that includes KPI and metric standardization, centralized reporting in key problems areas like digital marketing and content consumption, training for targeted stakeholders and power-user development, and the gradual incorporation of more sophistication visualization and data exploration tools (like Tableau or Spotfire) in the technology mix.
This work is important and necessary and in organizations with a culture that’s highly tuned to data-driven decision-making, it can work. But if your stakeholders aren’t sophisticated information consumers, the journey to informational maturity via data democratization can seem – and be – neverending. When I look at many of our client’s struggling to create informational excellence, I’m reminded of those never-ending state-sponsored freeway projects. Here in San Francisco, they’ve been re-working the approaches to the Golden Gate Bridge for something like the last seven years. Whenever you drive by, there’s lots of work going on but nothing ever seems to change!
The reason behind this seemingly never-ending labor is that for most organizations, data democratization misses the point. The goal of an analytics team shouldn’t be to democratize data to an organization that doesn’t know how to use it effectively. The goal should be to democratize knowledge. That’s a very different task.
When you realize that democratizing knowledge is your goal, it re-shapes many of the best-practice strategies around data democratization. The best-practice methods of data democratization borrowed from organizations with the right kind of culture just don't work for others. I’ve seen two methods to effectively drive democratization of analytics in organizations that aren't naturally measurement driven.
The first method is top-down and involves the wholesale change of key decision-makers. This can work, though the cost is enormous and the risks are acute. The second method requires a fundamental change in the strategy of the analytics group – from creating reports to creating tools. Embedding analytics in the form of segmentation, predictive modeling, driver analysis and analysis of variation into tools for the business is the only effective path to bottom-up knowledge democratization that I’ve seen.
These alternative strategies around data democratization certainly got discussed at X Change, but there was another aspect to data democratization that got a surprising amount of discussion. Most of the folks attending X Change lead digital analytics practices. In large enterprises, those practices are often just a small piece of the analytic pie. A 30-40 person digital analytics team may seem substantial, but large enterprises usually have hundreds of folks in analytics centers of excellence or customer insight teams.
Increasingly, those larger teams want access to digital data. And where do they go to get it? The digital analytics team. For years now, digital analytics teams have been on the other end of this equation. We were the folks going out to customer and merchandising teams, hat-in-hand, asking for data. Now, our digital data is a prime asset.
In one respect, this really is data democratization. These folks in customer insight and merchandising are data professionals. They aren’t the data consuming masses. They’re the elite. They don’t want or need reporting aggregations of the data. What they are asking for – almost always – is the raw, detailed event data. And while there’s some question about whether that’s the right thing to give them, what they definitely don’t need is reporting level aggregations.
So what’s the problem? You just give these groups an Adobe data feed (or equivalent in whatever solution you’re using). Done.
Only it’s not that easy. And your own experiences getting data from other folks should help you understand why. Remember getting customer data files from the warehouse? Or merchandising data? Or even getting digital data sources like Double-Click data?
What you got was a complex set of files with hundreds or thousands of fields and a seemingly unending series of kinks and peculiarities in the data. Every time you tried to use it and found something interesting, there was the data owner ready to point how you’d misinterpreted some field and used the data incorrectly.
Digital data is exactly the same. Not only are digital data feeds poorly documented, but the data is far from intuitive and it’s positively LOADED with quirks and peculiarities. I don’t care how skilled an analyst is, if they’ve never dealt with digital data before and you give them a data feed, they are going to misinterpret and misuse all sorts of things. 1st Time visitors? Visitor counts? Average page times? Referring domain after an entry page? Stuff we take for granted is, when you stop to think about it, really weird, very obscure, and completely undocumented in the data feed. And this is before they get into all the stuff you’ve created. Which set of eVars should you use? Which set of props? What should you use if a variable is in both a prop and an eVar? What if it’s in multiple eVars?
This stuff is confusing.
So yes, if you want other groups to fail, then by all means just give them the data feed. You can rub your hands together, given your best villainous chortle, and be pretty darn confident that whatever use they manage to make of a raw digital data feed that it won’t be a good one.
If, on the other hand, you actually want to democratize data and…you know…help the organization succeed, then you need to do more than shove a data feed and an SDR their way.
I think there are three complementary strategies for democratizing data to these elites. All are important.
First, I’m a big believer in embedding team members in cross-discipline efforts. We’re working with lots of organizations creating Hadoop-based landing zones. These data lakes are both incredibly powerful and hugely frustrating. The systems are powerful for obvious reasons but frustrating because combining cross-discipline data doesn’t solve many problems unless you somehow combine the expertise necessary to use it. Enterprise data in any domain tends to be too complex to use without deep internal knowledge. The idea that a data scientist can somehow make sense of this stuff externally is just one of the many myths about data science. Dumping the data in one place doesn’t translate into real-world sharing of knowledge. So if you’re Customer Insights team wants digital data, my best suggestion is that a digital team member goes along with that data. You get a data feed. You get a data geek. It works much better.
Short of sharing team members, data democratization to the elites requires, at the very least, a real effort at education. Documentation is never enough. If you can’t embed a team member, I’d suggest both an upfront, detailed walk-through of the data feed with a discussion of every field, how it’s used and how it’s often misinterpreted. Following this, I’d suggest regular check-ins on the usage of the data (this was an idea that came up in our Huddle and was widely approved – you need to stay on top of how people are using your data). This will help them and, because it protects the usage of the digital data, it will help you too.
Finally, I’m a believer that the raw level of detail isn’t necessarily the right thing to share. Customer teams don’t want and can’t use reporting level aggregates. But hit-level detail data isn’t the only kind of detail data that exists. There are interesting aggregations of the data that are not cubes but are, themselves, types of detail files. Things like journey records, session aggregations, even page usage aggregations. These types of mid-level aggregations contain real join keys and lots of detail data. But they eliminate many of the useless and confusing fields that exist in a raw data feed and they contain processed data that is both easier to understand and more powerful to use.
For digital teams to drive truly effective data democratization to these elites, thinking about and building these intermediate files is huge.
To me, the core of the debate around data democratization is still around building a mature information enterprise. And there’s lots of interesting things to think about when it comes to the search for democratization of knowledge. But X Change made it clear to me that for the large enterprise digital analytics manager, data democratization to the elites is a new and important challenge. Other groups need and deserve digital data. How we provision it and how we support their use of digital data will make an enormous difference to the maturity of the analytics organization.
[And speaking of the maturity of the analytics organization, Phil Kemelor is currently running our state of the digital analytics enterprise survey. We’re exploring a host of issues around how analytics is organized and what issues are dominating the digital analytics landscape. Please help us out (and get back some fascinating data) by taking the survey!]
Great post! I've also witnessed large clients with "centers of excellence" struggling with a lot of these issues. Also, I like the idea of democratizing *Knowledge* as opposed to Data, which reminded me of the DIKW Hierarchy, which I explored in this post: http://brendan-regan.com/the-data-information-knowledge-wisdom-hierarchy/
Posted by: Brendan Regan | December 05, 2014 at 09:23 AM