Trending up and to the right! But why?

Interpret your KPI graphs correctly!

In order to get the most out of Big Data and AI, corporations must implement and nurture a data-driven corporate culture. This takes time and abundant leadership to establish, nurture, embrace and grow. Becoming data driven is not just an issue of spending money in analytics software and graphical displays; it is a thoughtful, immersive and iterative process whose success largely depends on understanding what the data is telling you. As part of my blog series on how to “Transform Your Corporate Culture to Best Use Big Data”, below are three suggestions on how to best interpret KPI graphs and get most value of your Big Data and AI investments.

Aside: Remember when you asked your younger self “when am I ever going to use this?” in geometry or differential equations class back in college?

Yep; that day is today.

Let’s assume that you have managed to normalize data across your corporation, and have chosen to monitor weekly shipments of your flagship product. On your dashboard you see a graph with a line that is “up and to the right”, which makes you happy to no end. A few days later you check the dashboard again and the line has a downward inflection point, followed by an upward inflection point and a line that is also “up and to the right”, but at a much shallower slope. Is the corporation’s business still ok? In order to answer that question factually, we need to understand not only “what” and “how” KPIs are being graphed, but also what the different elements of the graph represent. Let’s dig in.


Narrow fields of view take away contextual references, and losing them limits your ability to correctly interpret KPIs. In the example above, looking at a short span of time could result in either alarms or complacency depending on what section of the graph is being analyzed. A wider field of view allows graphical variations to be studied in a greater contextual reference that help you ask the “better questions”. Why did we have a downturn in shipments? What “X” did we do to recover? Should we do more or less of “X” to improve the recovery?

While in this example an “up and to the right” trend is a good thing, a better thing is to understand the levers (see my blog on KPIs). Resist the urge to blow past graphs that at first look seem to be ok. Dig deeper and work to develop a sense for the levers underneath the KPIs and to validate their impact. There is such a thing as “over corrections” that can be as bad if not worst than doing nothing. You got to find that goldilocks zone of best changes to make, and you can only do that if you understand the factors that went into selecting the KPIs and the context in which they live.


Staying with the weekly shipments example, and asking the “better questions” – don’t stop at the first answer that satisfies your initial observations; dig deeper. Was the downturn in the graph a result of delays in raw materials to manufacture products? Was it a result of limited man power because of an unexpected quarantine? Or was it perhaps because of a faulty QA procedure that was much more restrictive than it had to be? Was the upward trend a result of paying some handsome overtime? Relaxing the QA procedures? Offering cash type MBOs to the workforce for increase productivity?

Most business “systems” should be modeled as second or third order systems. This means that there are one or two deeper levels of factors influencing the variations that you see. Therefore, selecting a linear trend – even when it fits the graph – may be an oversimplification that will keep you from really understanding what where the underlying factors that caused the positive change, their magnitude of relevance and also their timing for optimal influence. Here is the beauty – when you model the systems carefully (and OBTW – Big Data tools and AI DO THIS FOR YOU), you have a better chance at validating the correct levers and learning how to use them. They are “hidden” in plain sight; all you have to do is look for them, perhaps with the help of your favorite BI tool.

You get the idea; the changes we observe in our KPI graphs are usually the product of, and reflect the interaction of multiple variables, and we all know that “you need two equations to solve for two variables” – yeah, you are using those “useless tidbits” now, aren’t you? There are plenty of second layer KPIs, that can be analyzed from graphs when we carefully study them in our dashboards. The slope of your graph may represent your capacity rate (units per day or week). The area under the curve is your total production over time. Don’t get lost in the “ooh shiny”, and look past the cool colors and fonts of the graphs to focus on their underlying factors.


Keep in mind that not everyone in the team is a regular John Nash, and can abstract complex trends from written equations; that is one of the reasons we use graphical representations. One of the neat capabilities of Big Data, AI and graphical representation tools is that you can present the same data in a myriad of form factors with little to no effort. Present KPIs to your management teams in different form factors, as this may eliminate blind spots that are likely keeping you from achieving optimal performance.

Give it a go! Change your SalesForce, Tableau, Birst or INFOR analytics graphs from linear to logarithmic scales. Change the bar graph to a donut graph. Ask your team what they surmise. You will get a better understanding of how your team ingests the information, and will likely be surprised of what you can learn through their eyes.

In the end, Big Data and AI tools help us to dig deeper; to justify the perhaps intuitive but not inherently obvious. A corporate culture that is Big Data driven and empowered to act because everyone UNDERSTANDS what the underlying data and KPIS are screaming has a solid chance at success. Watching the charts show positive trends is not enough – you have to understand why, before you can achieve a state of ever-improving, sustainable and profitable business growth.

3 views0 comments

Recent Posts

See All