Most investigations into a number start the same way: someone notices NGR fell, opens a dashboard, and begins clicking. The result depends entirely on who is clicking. A good analyst finds the cause; a tired one finds a plausible story. The difference is not talent, it is method, and method is the thing that can be made repeatable.
Executive summary
A guided investigation turns "why did this move" into a structured path rather than a hunt. You ask the why question, confirm the period, and the workspace opens the branches worth checking for that domain, each with the supporting chart and a short narrative. The value is not that it finds the answer for you. It is that it checks the same drivers in the same order every time, shows its evidence, and produces something a colleague can reopen and continue.
This guide sets out the method: how to frame the question so the investigation is answerable, how to rank the drivers so you spend time where it matters, and how to keep the period, the definition and the evidence attached to the conclusion so it survives a second look.
1. Start from the question, not the dashboard
Separate the what from the why
A dashboard answers "what happened": NGR fell six percent. It cannot answer "why", because why is a different kind of question. It needs a hypothesis, a set of candidate drivers, and evidence for each. Opening more dashboards does not close that gap; it just produces more whats. The first discipline is to state the why question explicitly and let the investigation, not the chart, carry it.
A well-formed question already contains its scope. "Why did NGR fall last month" names the measure and the period. If the period is ambiguous, settle it before anything else, because an investigation run over the wrong window produces a confident, useless answer.
- Name the measure and the period before you look at a single driver.
- Treat the first chart as the start of the question, not the answer to it.
- Expect candidate drivers, not a single cause: a real move is usually two or three things, ranked.