An earnest and somewhat indulgent caricature of some principles that describe prudent quantitative research, borrowed from other fields:
“Measure twice, cut once”
In quant work, I take it to mean regularly making room to revisit the question/framing. Correctly answering the wrong question can be costly and damaging. If not discovered quickly time is wasted. If not discovered at all, it presents opportunity cost in forgoing a chance to clarify and improve understanding and worst of all risks a bad outcome by adding confidence to a poor decision.
“Do no harm”
The cost vs benefit of quant work is a function of how efficiently it sharpens an investor's command of the key tradeoffs. Math worth doing is used and organically improves as an investor applies it, finds holes and adapts their interpretation and/or the tool. This requires modeling that suits the problem and the decision makers.
The harm can creep in when there is too large a gap within the trio of model design, problem framing and investor understanding. A model that is not well integrated becomes the proverbial black box. You recognize this state when a conversation about model output becomes less nuanced and more binary. Discussions become about whether to trust or dismiss the model output rather than whether it adequately reflects the problem and how best to adapt to the insight.
I appreciate the spirit of "garbage in, garbage out" but it is more fun than helpful. In my view, models and assumptions are rarely inherently rigorous or flawed. I operate through the lens that rigor is entirely in the decision making, not in the analysis. The investor's capacity to interpret and apply insights derived from modeling (because it is well designed for the problem and presents relevant and material considerations), alongside other considerations, is the root of the rigor.
Modeling has a tendency to be distant and self-important, over-playing it’s certainty when it is really abstracted and ambiguous. Models become more useful and tangible as we remove the edge, an exaggerated fun house mirror version is something like this…
