Let's take a moment to interrogate a time-series plot in your next monthly review, shared PowerPoint slide deck, or performance boards around your facilities. As you are reviewing, recognize the aesthetical ease of interpreting time-series plots and the speed to plot them. Take the time to see what is being described on the x-axis and how it relates to what is shown on the y-axis. Look carefully because this time we are going to interrogate the chart.
In your example, the y-axis is likely something as simple as a number, dollar, or percentage that allows you to intimately connect to the information being shared. It can also be a calculated value that is then related over time with an x-axis that is calendar-based. This x-axis will show things like individual weeks, months, or years. When the y-axis and x-axis are mapped together, you will see the performance of the previous periods and begin to interpret the results as satisfying or disappointing.
In this interrogation, you will tend to transition your eyes to the left by evaluating the performance over time. This will allow you to immediately interpret if the performance is indicative of an improvement or a deterioration. You may shift your eyes now to the right to challenge your foreshadowing abilities based on the oscillating characteristics of the past. In a matter of seconds, you have interpreted the message of the chart presented. Congrats, you have completed phase one of this interrogation as the chart has organically created a form of motivation based on the interpretation. Now, on to phase two. But, we will have to wait until the next time this chart is presented.
Time series data in business, economics, environment, medicine, and other scientific fields tend to exhibit patterns such as trends, seasonal fluctuations, irregular cycles, and occasional shifts in level or variability. The objectives of analyzing such series are often to extrapolate the dynamic pattern in the data for forecasting future observations, to estimate the effect of known exogenous interventions, and to detect unsuspected interventions. - Guy M. Robinson, International Encyclopedia of Human Geography
The next performance review comes and the team is ready to present the same time-series plots showing the addition of the new period’s performance. Remember, we are interrogating this time, so let's look deeply at the data.
Wait! Last month, when that chart was presented, it showed 85% for October. Now it is showing 75% for October while we sit in the December performance review. What happened? How can history change? Someone must have been doing some shenanigans with the numbers because there was surely no management of the change conducted to change the number. So what happened?
If the intervals of the time series are regular but some values are simply not present. Sometimes data received through data ingestion may not have continuous data events as expected.- Divakar P M, Medium.com
The meeting just derailed and fell into a mess of cluttered attempts to try and explain the calculation. What a mess! How did we get here with something as simple as a time-series plot? If we interrogated this single chart, could there be others with the same mistakes?
We have all been in this meeting when a time-series plot has instantaneously exploded into a room of confusion and distrust. And when the meeting ends, regardless of the attempts to try and explain the calculation, the emotions of the meeting lead to frustrations and doubt. More meetings then get scheduled to evaluate the calculation and now we are committed to making it more mature. The time-series plot has created a dense death spiral of more meetings on what was assumed as a simple chart. Is this the fault in the calculation, the ease of presenting, or the inadequate amount of time training the definition? Yes; all of the above.
As data presenters, being more diligent at describing what the x-axis is presenting with the y-axis is necessary versus conforming to the ease of simply presenting. We may present charts where history cannot change, but in most cases, history can indeed change when presented again in the future. In these scenarios, we have delayed calculating additional values and that is okay if we describe clearly what the x-axis is showing.
You may elect to have a plot that shows when safety injuries occurred. However, months later it was determined that the incident was indeed not an injury. Or you may show when and how many quality defects are produced at a production unit historically, but the defect wasn’t found until months later at the customer. You may show manufacturing's Preventive Maintenance (PM) Compliance for a given month, but have team members that were delinquent in closing the work order out in time. In these examples, we have changed an input that influences the historical views of the chart but have appeased the audience by showing only its current snapshot in time. We presented a snapshot of history based on a moment's interpretation.
Rigidly locking in an interpretation of the time-series plot can be a dangerous slope if we rush towards the aesthetical ease of presenting versus spending time explaining the calculation. To emotionally counter this, we may strive to overcomplicate the plots as these prescribed flaws are revealed. But the trap remains that we are rushing to satisfy the quench of the audience versus educating the definition of what exactly is being shown. Something as simple as PM Compliance can be shown as green in a performance review meeting, and the team interprets the green as smelling like money. But a month later, we transact additional PMs that change our already presented historical performance to find out that the green smelt like puke. Be careful with time-series plots, their ease of development creates a dangerous scenario of misunderstanding what they are calculated to show.