Hacker News Reader

Comment by kayo_20211030 | Hacker News Reader

kayo_20211030

Tech debt is categorically unquantifiable. Most of the time it's more of a feeling than a number, and it's not an accessible number ever. What's the ROI on paying down debt? Hold as is, renegotiate, or extinguish? It's the same calculation that goes into deciding which of the five thousand value-add proposals to prioritize. The piece is unconvincing for that reason. There's an assumption that TD is known, and an implicit assumption that the ROI on its remediation is known. Neither of those is true. Systems evolve to where they are, with all their TD warts, because value-add was prioritized.... and then we have TD. I reckon we should just live with that uncertainty, move the product forward, calculate ROI using the same bogus productivity metrics we always have, and stop inventing "better" systems which are just another form of magic, but manage to suck up time and resources not required by accepting on faith the old bogus metrics.

View on HN

It’s really not unquantifiable. I read “How to Measure Anything in Cybersecurity Risk” and it was an eye opener. Using a table of risks and outcomes with associated probabilities and 90% confidence intervals of dollar impacts we can quantify categories of technical debt.

If "Cybersecurity Risk" were the only form of technical debt, we'd be just fine(?). Or, at least, we'd have some sort of metric. It wouldn't be a good one, but it'd be there. Chance of a breach: 1%. Existential or not? Probably not. Cost of mitigation? Probably small. Worth addressing? Mostly no, unless you're a regulated entity; then it's mandatory. Quantifiable, for this narrow case, but what of the rest?

Apply the same mentality to other things. If the cybersecurity folks can quantify risk so can you. Are you keeping track of your supply chain? How modular is your code? How easy to refactor is your code? You could think of reasonable metrics to measure various aspects of technical debt. It won't be perfect but it's better than nothing.

I think a bad metric is very much worse than nothing. It sucks away time to record, debate, report, and discuss. It encourages bad decision making. If you throw up a number people will give it weight, even if it's stupid. Multiplying 6 gut checks and trying to make a decision about engineering direction is like tracking someone's mood by the metric of whether they ate an odd or even number of calories yesterday. There's theoretically a signal under all that noise, but the direct gut-check or any number of qualitative clues are so much better than the distracting number.

I agree whole-heartedly. A bad metric is a curse. It's misleading, resulting in waste, and falsely reassuring simply because it exists as a number. +100 on the gut-check qualitative approach

You can say whatever you want, there are studies backing up the beneficial impact of quantitative over qualitative.

They'd be quantitive studies of quality, I hope.

Reply on news.ycombinator.com