Decisions, decisions, decisions

Source: https://commons.wikimedia.org/wiki/File:Decision_Tree_Elements.png “A simple decision tree showing the major elements: Squares, circles and triangles (decisions, chances, end nodes)” This image is in the public domain.

What’s new?

In May 2020, Lori Beckman wrote at Production Machining about the company MachineMetrics. She quotes Eric Fogg, co-founder and COO of the company: “MachineMetrics can collect any data that is useful to a customer,” he says. “That data can be whether the machine is running, how many parts it has made, and its alarm history, for example. We take all the data, encrypt it and forward it to our cloud service where we perform analytics and create dashboards for our customers to use.”

What does it mean?

Most machines on manufacturing floors are ready to be part of the IoT, Internet of Things, that is, a network of devices that share information; they include sensors and data collection. MachineMetrics can easily connect its device (MachineMetrics Edge) to almost any machine (through the machine’s ethernet port). The device reads the data from the machine and sends it wirelessly to software that will store, analyze, and display the data, combined with data from other machines. Data as simple as status of the machine can be used for asset management, machine utilization monitoring, and real time alerts when repairs or adjustments are needed.

What does it mean for you?

Collecting and analyzing data is done for two major reasons. The first is to make better decisions. One of my areas of expertise, decision analysis, stresses this purpose for data. Decision analysis even has methods for determining the expected value of information, and for determining if collecting data is likely to be worth the cost of doing so.

The second major reason for collecting data, somewhat in contradiction with the points I just made, is to achieve understanding, often for long run purposes and not for any immediate improvement in decision making. One of my first professional jobs, while I was in graduate school, was for the corporate planner for a large medical group. One of his overall goals was to use data to reach a solid understanding of the organization as a system and of the organization as part of the larger medical system. For example, in one study I analyzed employment records to seek to understand the plausibility of some commonly held beliefs about physicians using the organization to establish themselves in a new geographical region before setting up their own practice. We found only weak evidence to support the belief, not enough to worry about the impact of this behavior on the organization. This study was just one small piece adding to the understanding of the system.

Collecting data to make better decisions gives a focus and structure to thinking about data. What decisions do you or others in your organization make? They range of course from strategic (should we acquire that company?) through major (should we purchase a new piece of equipment?) to daily (which job should be done next on which machine?). Decision analysis focuses your thinking by requiring you to explicitly state the alternatives, the uncertainties concerning what occurrences will follow, and then ensuing decisions and occurrences. A decision tree (see the small one at the top of this article) is a visual representation of that sequence of decisions and occurrences. With such a visual representation, you can then ask what type of data would reduce uncertainty about future events and enable better decisions to be made now.

Collecting data to achieve understanding can also be valuable, but it is sometimes a way to bury oneself in data and confusion. In my job with a medical group, I learned a lot about data analysis, but I also learned a lot about system modeling. The tools for data analysis 50 years ago did not allow the powerful search for patterns in large databases, so we rarely (probably never) simply looked for patterns, nor did we often collect new data. Instead I learned a lot about how to formulate a good question about how a system works and then to use existing data bases to explore answers to that question.

Data collection tools like those offered by MachineMetrics can help your decision making and can help your understanding of your organization as a system, but data can also bury and confuse you. For example, what is measured and monitored may become a goal that distracts from what cannot be measured or monitored. If you are measuring and focusing on productivity but not on job satisfaction, you may have good short term and poor long term results. Monitoring of machine utilization might lead to an incorrect focus on increasing machine utilization to the detriment of other organizational goals.

The first example in the MachineMetrics article at Production Machining can, in fact, be read as a cautionary tale in which the newly collected data made the shop’s managers finally listen to the machine operators’ insistence that machine break downs were slowing productivity. Isn’t this story really about communication between people, not between machines?

Always ask yourself about any data collection: will it help me make better decisions? If not, will it help me understand my organization as a system? Don’t just collect data because you can. Don’t be misled into thinking that gathering more data is always good.

Where can you learn more?

An Internet search on “decision tree” or “decision tree analysis” will lead you to many useful pages with an introduction to this technique, such as here (a 1964 Harvard Business Review classic), here, and here. Decision tree analysis relies on the question “What happens next?” Many business degree programs and probably all MBA programs include decision analysis. I still think the best introductory book is the 1968 Decision Analysis: Introductory Lectures on Choices under Uncertainty, by Howard Raiffa. When I wrote this blog posting, Better World Books had used copies for $5. Software exists to help you create and analyze decision trees. My favorite is TreeAge.

The IoT, Internet of Things, doesn’t necessarily mean connecting everything to the Internet. The “Things” in the Internet of Things are devices that have built-in sensors, communication capabilities, and controls. These are connected with each other to collect data and to enable control of this network of devices.  A simple example is when your mobile phone connects to your car so you can take a call hands free. The IoT relies heavily on shared protocols for data format, so interconnectivity is a crucial capability of any device your organization acquires; interconnectivity’s evil twin is computer security, also a crucial capability.

This work is licensed under a Creative Commons Attribution 4.0 International License.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.