I’ve never thought deeply about performance measurement based on inputs vs. outputs. But then I read a blog post by Laura Albert, Industrial & Systems Engineering professor at the University of Wisconsin-Madison, in which she discusses the decision of changing running shoes based on inputs (how many miles running with the shoes) vs. outputs (the wear on the shoe soles). It got me thinking about times I’ve seen this practice in different areas…
Here are a few examples:
Calculating the production rate: Counting the finished products (outputs) at the end of the assembly line rather than measuring the time used on each machine (inputs)
Scheduling the preventive maintenance on the machineries: It’s easier, and perhaps wiser, to look at the number of operating hours or items produced (inputs) rather than waiting to see a sequence of defective products, or a higher frequency of defective products (outputs). In this case, unlike with running shoes, the stakes might be higher and measuring the inputs might be easier.
Determining truck driver hours by examining travel distances (inputs) or tracking hours directly (outputs).
In root cause analysis, there are a lot of examples that can be viewed as analyzing inputs vs outputs:
Looking at customer complaints and their feedback and frequency (outputs) might be easier in identifying the problems and fixing them sooner compared to looking at the sales numbers/number of customers (inputs).
Looking at the marginal revenue (outputs) in production planning of specific products, rather than looking solely at the sales number (inputs). One example of this is Volvo’s green cars story. The Volvo sales team was doing whatever they could to sell their green cars, without making much money. But due to lack of communication between sales and operations teams, the high number of green cars sold made the operations team think that there was a good demand for the green cars and that they needed to produce more of them.
In predicting the chances of truck accidents, it might be more useful to look at the driver’s hours of sleep, lifestyle quality, eating habits, etc. (inputs) rather than solely checking the history of accidents of a driver (outputs).
In finding the traffic to your website and understanding which pages are more interesting to users, looking at the devices and their lengths of stay on each page (inputs) might be easier than asking for their feedback (outputs). On the same note, it might be easier to look at the frequency of a user/IP address to the website and the length of stay as a measure of satisfaction (inputs) rather than relying on the user filling a survey or questionnaire to show their satisfaction (outputs)!
Although I don’t aim to suggest that measuring outputs should be preferred over measuring inputs, if easily collectible, evaluating the outputs should be preferred. After all, they provide the true feedback on how well the system/model works.
But approach this with care. Because, when making a decision, the relation between inputs and outputs, the complexity of collecting and evaluating them, and the usefulness and impact each one can have on the decision should be considered. Ask yourself:
Is it easier to evaluate inputs or outputs?
Which is more intuitive? Which is more useful?
Is it possible to quantify both?
If you provide answers to these questions first, you will then be better equipped to decide which approach can help you meet your needs better.