Numbers as a Thought Exercise

Could you use metrics as an aid to thinking?


As a researcher, software engineer and former physicist, I’m fascinated by numbers. Software works by numbers; even every data item is expressed as a numbered location in memory. As a researcher, I collect numbers even for qualitative data, and express them using infographics—an approach that works very well for my software engineer readership!

Mostly, those numbers are themselves the point; the numbers themselves tell the story we need to know. But sometimes, it is not the numbers themselves that matter, but thinking about the measurement.


Consider a couple of examples.


First, we recently carried out a literature review: searching a wide range of academic and commercial literature for work related to a particular academic question. The ‘gold standard’ approach to such reviews is PRISMA; a PRISMA diagram shows how many documents were found using each approach, how many documents were considered, how many were rejected for each reason, and so on.


The important thing about such a diagram, we found, is not the actual numbers. It is that filling them in honestly forces the researchers to be systematic; and the report of the numbers in the paper therefore demonstrates to a reader that the research was systematic. The insistence on numbers forces a systematic process.


Second, numbers are important when we express the requirements for a system.  ‘Digitisation’ of requirements was first championed in the 1980s by former IBM expert Tom Gilb. He promoted the concepts of metrics for software design—not just for features like performance and memory size, but for such concepts as reliability and usability. My experience from using such metrics in software design was that defining them (such as “mean time between failures: 1 year”) allowed us to justify and focus our software design decisions around aspects that were vital to our stakeholders. We may never have actually measured the mean time between failures of the resulting product, but simply considering such metrics made for a better design.


So think metrics, not as an exercise in accountancy on the final product, but as an exercise in design up front. It’s worth it!


-              Charles