Not every analysis needs to become software. A clear answer to a one-off question is often more valuable than a polished tool nobody needs twice.
The case changes when the same question returns with new inputs, different scenarios or another person waiting for the result. Repetition creates a product-shaped problem: the method needs to be reliable, the assumptions visible and the output easy to revisit.
The LGA Language Explorer crossed that line because the underlying Census analysis had to answer the same question for different suburbs and LGAs. Turning it into a tool meant preserving the harmonisation method, exposing the geography choice and letting someone explore the result without asking me to rerun the work.
An interface can preserve choices that disappear in a spreadsheet handover. It can show history, validate inputs and let the user test a scenario without rebuilding the analysis.
The decision-first approach helps determine what the interface needs to show, and the broader dashboards and decision systems guide explains how I connect that decision to the design.
The signal is not simply that the analysis is complicated. It is that the decision is recurring and the reasoning is worth making reusable.