Authoring Good Analysis Questions

The quality of an analysis can often be traced back to the quality of the initial question or topic. Once you've established your research question, all the work that follows - research, decomposition, scenario development, forecasting, writing - inherits the shape of the initial question. A high-quality question focuses all of that effort onto a useful, tractable target. A vague one can doom an analysis to irrelevance.

Time spent sharpening the question up front pays compounding dividends, while no amount of downstream sophistication can rescue a question that was wrong from the start. The good news is that framing strong questions is a learnable skill with well-understood principles, drawn from intelligence tradecraft, scenario planning, forecasting science, and research methodology. This guide distills those principles for the way analysis works in Hinsley.

What a good analysis question is for

Three first-principles purposes should drive how you frame any topic.

• It serves a decision or a consumer. Analysis exists to improve decision making. Before anything else, know who will read the output and what they will do differently depending on what you find. This is the "So what?" test: if getting the answer wouldn't change anyone's understanding or action, the question needs rework.
• It explores genuine uncertainty, but remains tractable. Good questions live between the obvious and the unknowable. If the answer is already settled, you have an information-lookup task, not an analysis. If it's too broad or too complex, it becomes unmanageable. Aim for the productive middle: questions where careful work over evidence and competing drivers actually moves your understanding.
• It can be decomposed. A strong topic breaks cleanly into parts - the drivers and signals beneath it, the alternative outcomes it could resolve into, and the resolvable sub-questions that give you evidence. If a topic resists decomposition, it's usually too vague, or it's secretly several questions bundled together.

The qualities of a strong analysis topic in Hinsley

Synthesizing across the disciplines - the FINER criteria from research methodology, the key-intelligence-topic and key-intelligence-question (KIT/KIQ) framework from intelligence practice, and scenario planning's focal-question rules - a well-framed analysis topic tends to share these qualities:

• Decision-relevant. It connects to a real choice someone is facing. You can articulate what becomes possible or avoidable once it's answered.
• Right-sized. It strikes the central tension of all good framing: specific enough to provide direction, broad enough to allow exploration. Too broad and it becomes unanswerable; too narrow and it collapses into a single forecasting question.
• Forward-leaning and uncertainty-facing. It points at something not yet settled - a trajectory, a risk, a possible shift - rather than a fact already in the record.
• Bounded. It names its scope along the dimensions that matter: who (actors, organizations, markets), where (geography, domain), and when (a time horizon).
• Decomposable. It can be broken into drivers, signals, scenarios, and resolvable sub-questions. You should be able to imagine the tree beneath it.
• Neutrally framed. It doesn't smuggle its own conclusion into the wording. A leading question primes every downstream step toward confirming it - exactly the bias Hinsley's red-teaming features exist to catch.
• Tractable with available sources. If answering would require information that cannot be obtained or observed, the question - however interesting - won't produce useful analysis. Check that the evidence plausibly exists before committing.

A key distinction: an analysis topic is not a forecasting question

Hinsley supports forecasting questions, but they sit inside an analysis as components - they are not the topic of the analysis itself. The two operate at different levels, and confusing them is the most common framing mistake.

A forecasting question is narrow, resolvable, and time-bound. It has a definite answer that the future will reveal. The gold standard here is the clairvoyance test - a clarity standard introduced by decision theorist Ronald Howard and later made central to forecasting by Philip Tetlock's prediction tournaments: if you handed the question to an all-knowing clairvoyant, could they answer it without asking you what you meant? "Will Country X hold national elections before July 1, 2026?" passes - there is no ambiguity about what counts as a yes.

An analysis topic or research question operates a level up. It is strategic, exploratory, and often deliberately a little fuzzy. It is not asking what will happen on a single resolvable point - it is asking you to understand a situation, map its drivers, surface alternative outcomes, and inform a decision. "How might the political trajectory of Country X over the next 18 months affect our regional supply chain?" does not have a single clairvoyant answer. It is meant to open a space for structured exploration.

In scenario-planning terms, the analysis question is your focal question - it acts simultaneously as an anchor and a fence. As an anchor, it pulls relevant evidence, drivers, and arguments into orbit. As a fence, it keeps the work from sprawling into everything tangentially interesting. Forecasting questions are then nested underneath it as the resolvable instruments that feed evidence back up into the larger picture.

A practical authoring pattern

If you're unsure of where to start, you can follow the scaffold below to produce a workable topic. Fill in the slots, then smooth it into a natural question:

[Situation / trigger] + [actors or system in scope] + [the uncertainty or decision] + [time horizon]

Worked example:

• Situation: New export controls on advanced semiconductors
• Actors: Our top three Asian suppliers and their downstream customers
• Uncertainty / decision: Whether and how to diversify our sourcing
• Time horizon: Next 12-18 months

"How are recent semiconductor export controls likely to reshape the viability and pricing of our top three Asian suppliers over the next 12-18 months, and what sourcing options should we be preparing?"

That question is decision-anchored, bounded, forward-leaning, neutral, and decomposable - and it naturally suggests the drivers (regulatory enforcement, supplier exposure, demand shifts), the scenarios (gradual adaptation vs. sharp disruption), and the resolvable forecasting questions (e.g., "Will Supplier A announce a capacity reduction before Q4?") you'd nest beneath it.

Common failure modes (and the fix)

From topic to analysis: how the question flows through Hinsley

A well-framed topic makes each Hinsley capability work more effectively:

• Decomposition maps cleanly only when the topic has identifiable drivers and signals beneath it. A bounded, decomposable question gives the decomposition something real to break apart.
• Scenario Builder needs a focal question that admits multiple plausible futures. A topic framed around genuine uncertainty produces meaningfully different scenarios; a settled or leading topic produces one obvious scenario and some strawmen.
• AI + Human Forecasting plugs in at the level below your topic, as the resolvable instruments that feed evidence upward. A good topic naturally suggests which forecasting questions are worth attaching.
• Red Teaming / Bias Detection has the most to catch - and the most to offer - when you've been honest about a topic's framing. Neutral framing up front means red-teaming refines rather than rescues.
• Outputs & Publishing produces a focused deliverable when the topic gave the work a single clear target. The "fence" you set at the start is what keeps the final product from sprawling.

A 60-second self-check before you commit

Run any candidate topic through these:

• Decision - Who acts on the answer, and what changes for them? (If nothing: rework.)
• "So what?" - Does answering it usefully advance someone's understanding?
• Uncertainty - Is this genuinely contested or evolving, not just lookup-able?
• Scope - Have I bounded the who, where, and when?
• Size - Broad enough to explore, narrow enough to direct? (Not a single forecast; not "the future of everything.")
• Neutrality - Have I avoided baking in the answer I expect?
• Decomposability - Can I picture the drivers, scenarios, and sub-questions beneath it?
• Feasibility - Does the evidence to address this plausibly exist?

If a topic passes these, you've done the highest-leverage work in the entire analysis - and Hinsley has a sharp target to aim everything else at.