Callaghan's five-step methodology (abbreviated SOS) developed at BiggerPockets for adding observability incrementally in response to concrete production problems: (1) Question — articulate the specific question you want answered (refine vague 'why are emails failing' to answerable 'what jobs take time in this queue'); (2) Data — decide what event to capture, filter by, group by, and plot (mentally expressible as SQL); (3) Instrumentation — build the logging/tracing, confronting convention, missing-attribute, and API-coverage gaps; (4) Graphs — query and visualize; (5) Improve — critique your work, show colleagues to uncover what they care about, add attributes (job latency, IPs, request IDs) and loop. Explicitly not a 'spend three months adding observability first' approach — always start from a live problem.