Insight Prioritisation and Sizing

How do I decide which research insights to lead with?

Use eight criteria: decision relevance, evidence strength, population affected, magnitude of effect, actionability, urgency, novelty to this organisation, and the reversibility of the decision it informs. Apply the first two as gates rather than as scores, then tier what passes into lead, support and context, then order within tiers on magnitude and reach with actionability and urgency as tiebreakers. Write a sentence of reasoning for every placement. If a placement cannot be justified in a sentence, it is a preference rather than a judgement.

Why not build a scoring model?

Because the criteria are incommensurable and a total supplies an exchange rate that does not exist. In practice, averaging lets a weakly evidenced insight about a large population outrank a well-evidenced insight about a smaller one, which is exactly the trade-off a decision-maker needs to see and exactly what the total hides. It also produces a rank order that cannot be interrogated: once the number exists, the argument is about the weights rather than about the evidence. State the refusal in the output, so nobody reinstates the model later.

How do I size an opportunity from survey data?

As a chain, and only as far as it goes: a measured incidence with its base and coverage limitations, a defined population from a named and dated source, then an explicit assumption for every step in between. Vary the two most sensitive assumptions across defensible intervals and present the result as a range. Never combine a measured figure and an assumed multiplier into one number without labelling which is which, and if a required input does not exist, stop the chain there and say what is missing.

Why should a size always be a range?

Because a point estimate on a modelled quantity claims a precision the method cannot deliver, and it is the form most likely to be quoted back years later without its assumptions. A range with two named sensitivities is the honest shape of the quantity, and it survives the scrutiny a single figure does not. It is not a hedge; it is what you actually know.

What is the difference between the size of a population and the size of an opportunity?

Four gaps. What people report is not what they do, and the discount between them is an assumption. The organisation can reach only part of the affected population, and the addressable share is usually the binding constraint. Stated willingness converts at a rate the study did not measure. And some of the outcome would have happened anyway, so gross is not net. An opportunity stated as population multiplied by incidence has skipped all four.

What do I do with insights that matter but are not urgent?

Put them in a deferred register with three things: an owner, the evidence retained so the insight can be picked up without re-running the study, and a trigger, which is the date, event or decision that would make it live. A system replacement, a contract renewal, a planning round. A deferred insight without a trigger and an owner has been deleted with extra steps, and everyone in the room knows it.

Should I report something the client didn't ask about?

Assess it on the same criteria as everything else, and if it clears the gates, report it in a labelled section saying it was not commissioned and why you are raising it. The scope was written before anyone knew what the study would find. A study that answers only the questions it was given cannot tell the organisation anything it did not already suspect, which is most of what it was for.

Can a low-confidence insight lead the report if the effect is large?

No. Evidence strength is a gate rather than a contributor, precisely because magnitude is so persuasive. A large effect on a base of 28 is routed to validation and reported as a labelled hypothesis; it does not occupy the top of the list. This is the single most common way an under-evidenced claim reaches an executive summary, and a scoring model is usually how it gets there.

How many insights should a short list have?

As many as clear the gates and reach the lead tier, which is typically three to five and is sometimes one or two. The number is a property of the study, not of the slide layout. A short list of two, well-evidenced and connected to an open decision, does more for a decision-maker than five where they cannot tell which two to trust.

How do I stop insights disappearing between the analysis and the deck?

Fix the full register before you rank anything and never delete from it. Every insight keeps its criterion assessments, its tier or routing, and the reason it is where it is, and the register ships with the deliverable even when only the short list goes into the deck. It costs one page, and it is what lets someone eighteen months later see that a finding was weighed rather than missed.