What is the purpose of an analytics culture? (It's not the dashboard.)

To arrive at the answer to this fundamental question, I believe it is helpful to explore the following:

• What is the ultimate purpose of an analytics culture?
• What are the foundational beliefs and values that drive successful behaviors?
• How can we get started?
• How do we sustain the culture we built?

Let’s first examine the purpose of an analytics culture. What is the purpose or ideal outcome of any scientific work? After all, for most of us, the point of analytics is to spur evidence-based scientific discovery and problem solving. I have asked this same question to many audiences in my past roles as an engineering leader at US Synthetic, a Shingo Prize winning company, and as a Systems Engineer at JMP. I get a variety of answers. Sometimes the purpose is intellectual property, other times sellable products. Technical papers come up as well. These are good answers, but each is a consequence of the useful knowledge derived from the work we do. And there it is, our purpose – usable knowledge. Plain and simple. Analytics is simply an engine to create useable knowledge.

The four beliefs that set learning organizations apart

With knowledge as our guiding light, how we act to gain, spread, use, and repurpose knowledge is crucial to a successful analytics culture. The most effective organizations I have worked with share some basic core beliefs that lead to effective behaviors. Important to this conversation are:

• Learning is more important than being right.
• Using sound repeatable methods is just as important as success, maybe even more important.
• Knowledge only has value if it can be used in the present and reused in the future.
• We must fight like hell not to waste the knowledge we work so hard to attain.

At a high level, these beliefs free people’s minds to experiment and innovate freely without fear of failure and with confidence that whatever the outcome, there will be value in it. This is crucial. It’s also where I have seen organizations fail, so it’s important to get it right. Ideally, leaders should be held accountable for all aspects of a company’s culture, but, in my opinion, this component should be at the top of the list.

How to launch an analytics culture without waiting for top-down buy-in

The crux of getting an analytics culture up and running is aligning the organization around the basic principles I laid out. I can tell you from firsthand experience that this does not happen overnight and often begins with resistance.

Ultimately, alignment can only happen with sufficient evidence that the hard work required will add value. How then can a case be built in the absence of the very culture that you need to generate knowledge? I have seen some creative approaches to kickstarting an analytics culture. Here’s one I have used personally:

• Begin with a pilot group that includes a sponsor from senior leadership.
• Tackle a problem that has value to the organization but is also compact enough that results can be shared in a matter of weeks or months.
• Use a problem-solving framework to guide activity (PDCA, DMAIC, A3, QBD, Agile, etc.).
• Have the team and its sponsor report the results to the executive team and any other key stakeholders. Be sure to highlight any knowledge attained and the analytical discovery process used.
• Make sure company leadership recognizes the team’s success publicly.
• Review the activity at its conclusion to highlight what worked and what needs improvement.
• Celebrate success.
• Repeat until more people and teams want to join the effort.

Using this approach, buy-in grows organically as knowledge is attained and shared. One important caveat is that leadership must buy into the the idea that the knowledge gained by the team is the critical outcome, in case the first-round results are not home a run. The other caveat is that there must be sufficient means to gather the data that’s needed to solve the analytical problem. If that’s not the case, the first item on the to-do list needs to be developing a measurement system. In my experience, top-down or leadership driven initiatives that have not evolved with the contributions of people who make them work long-term, either fail or require a lot of firefighting to fix the things that were missed.

Why no-code analytics tools are the foundation of a democratic data culture

The foundational data science and statistical tools needed to launch an analytics culture depend largely on the problems that an organization needs to solve. At a minimum, there should be a working understanding of:

• Data preparation
• Data visualization
• Exploratory data analysis
• Hypothesis testing
• Basic statistical modeling
• Design of experiments
• Quality methods

To help companies assess and bolster their staff’s statistical literacy, JMP has built an online training series, Statistical Thinking for Industrial Problem Solving (STIPS) that teaches everything (with the exception of data prep) beginners need to know. I can say from experience, this course is an effective way to get started without requiring anything other than time. It even includes a virtual instance of JMP available to do the exercises. There are extensive additional resources that can help deepen and broaden expertise beyond the basics.

When thinking about culture from a long-term perspective, there are two basic building blocks that I find to be key.

• Software that is easy to use and comprehensive – and that requires no or low coding.
• A hub for capturing and sharing knowledge.

I started my career scripting or writing computer code for data analysis workflows. It is what I learned in school, and it offered built-in automation capability and flexibility in my analysis efforts. It was fun but involved skill and time. As I found myself leading a project to modernize a test lab and build the foundations of an analytics culture, I attempted to train my team to script. As you may guess, it did not help to democratize the work we needed to do, in spite of my best hopes. Scripting is not for everyone, but analysis is and must be for a successful culture.

JMP offered a solution to the problem. Between its comprehensive capabilities and approachable interface, everyone on my team, including technicians, could contribute to analysis and discovery. The people with scripting skills in the group were able to use the JMP scripting language (JSL) to handle the automation tasks we needed for reporting. Automation has become even easier in JMP recently with the Workflow Builder, allowing everyone to automate analytical workflows. JMP has been a springboard on the path to analytics excellence for my former company and many companies I have worked with. Software should never be the limiting factor for building an analytics culture.

Knowledge waste: The most expensive problem in analytics that no one is measuring

Analytical problem solving often ends with some type of summary report. The report may be in form of a Word document, a PowerPoint deck, or an email. Often, it gets filed away and rarely resurfaces when the results are needed in the future. This form of knowledge waste often leads to someone repeating the work. Sometimes more than once. Of all the forms of knowledge waste, such as using poor methods that are not adequately accounting for variation or solving irrelevant problems, I find this one to be the most frustrating and costly.

One of my hobbies is building bicycle wheels, which is a challenging, costly, and time-consuming hobby. To use my hobby as an analogy, the type of knowledge waste I’ve just explained is the equivalent of throwing away a nice wheel-build after only one ride. So how do we avoid it?

Certainly, we must have a culture that demands that we look to see if someone else has already solved the problem we are about to tackle. But this practice has its own risk, namely repeatedly having to ask the people who generated the knowledge for help. Referred to as hero dependence, it is another source of waste. To avoid wasting that time (and the good will of the original problem solvers), creating a hub or platform to curate and share that knowledge is a structural solution to the problem.

Ideally – and at a minimum – such a platform, should be:

• Interactive.
• Easily updated.
• Standardized to help with interpretation and management.

Ease of use ensures knowledge can be digested, used, repurposed, and expanded with minimal time investment. I have seen several homegrown software systems that work well enough. They were often very similar to the report binders I used early on in my career. The software system I used in my last job had its pluses and minuses, but worked – mostly. Pharmaceutical company Regeneron has a nice article, in which the manager of continuous improvement statistics, Diana Nadler, describes a more modern approach to tackling this problem. It’s definitely worth reading. In principle, a systematic approach to knowledge sharing will help prevent knowledge waste and support the behaviors required for an effective analytics culture.

How a knowledge-sharing culture becomes self-sustaining (and why AI doesn't change that)

At the end of the day, if we are constantly learning and applying our knowledge without discarding and repeating our work, we can expect the value to be visible in how consistently we outlearn and outearn our competition. Incentivized by consistent demonstration of value, our analytics culture will be self-sustaining. It is also comforting in our age of AI to know that even as the way we interact with data evolves, knowing which questions to ask and how to apply the knowledge gleaned from data analysis still depends on all of us humans.

The Lean Enterprise Institute has many resources for those who are looking for more information on knowledge-focused research, product, and process development.