Analytics Pipeline

Challenge

For over a decade, my company was excellent stewards of data. We believed in, and evangelized its value long before the big data hype train had arrived. This put us far ahead of our competition and was viewed by our clients as a differentiator. As the industry and our clients evolved, I identified that our lead in the industry was in danger of eroding unless we were able to make meaningful use of all these data we had meticulously curated. We've all heard the mantra, "Data, data everywhere, and not a thought to think." Well, that was us.

If we have data, let's look at data. If all we have are opinions, let’s go with mine.” Jim Barksdale

Our strategists' thoughtful recommendations were losing ground to misinformed client opinions, all because the evidence we needed was hard to expose. Ad-hoc queries were inefficient and required on-demand technical assistance that was in short supply. We needed to create a reporting and analytics infrastructure, and we needed it fast.

Solution

In order to make progress and move forward, you have to do something. If you focus too much on creating the perfect solution, you're likely to end up with no solution. This is commonly referred to as "analysis paralysis", and it's an easy trapping to fall prey.

Incremental improvement was our path forward and I tackled it in four steps:

Transform the relational data into a more analytics-friendly format
Identify and implement a flexible, best-in-class reporting solution
Gather business requirements from key users on the metrics they use most often
Build, deliver, refine, repeat

Version 1 was initially well received and I considered it successful because it sparked a dialogue about data that was previously non-existent. The infrastructure was sound, easy to use, and automated. After a few months of refining the reporting end of the solution, I saw something curious in the usage metrics: no one was using it. Well, not "no one", but usage had precipitously dropped. I had exposed every key metric the users wanted, and had built feature that allowed for an incredible amount of "slicing and dicing". So why was it being abandoned?

After interviewing some of the users who had abandoned the reporting solution, I quickly determined that, although the solution was flexible, it was too complex. With this new information I did the opposite of what most product developers do: I removed features. This approach immediately brought back users.

Errors using inadequate data are much less than those using no data at all.” Charles Babbage

Is it the perfect solution? Not by a long shot, but it was progress and it created the foundation for all of the company's future analytics offerings. It set us up to become a more scientific organization and helped promote a test and learn mentality.