Learning objectives
By the end you can:
- State the difference between a finding and an insight in one sentence and then in mechanism, and explain why a client pays only for the second, because an insight is what a finding means for the decision, and analysis alone never crosses that gap.
- Climb the So-What Ladder deliberately (data → finding → insight → implication → recommendation) driving any raw observation up the rungs by asking "so what?" at each step until it becomes an action, and diagnosing exactly which rung a weak deliverable stalls on.
- Apply the three tests of a real insight (is it actionable or belief-changing; would an informed client be moved or surprised; does it point at the decision) to separate genuine insight from the "interesting but inert" observation that only looks like one.
- Build a governing-thought pyramid (Minto): a single governing thought at the apex, supported by a MECE key line of two-to-four grouped arguments, with evidence beneath, enforcing the vertical Q&A relationship and choosing correctly between a deductive and an inductive (grouping) key line.
- Perform the inductive-summary act at the heart of synthesis, replacing a bland label over a group of findings ("there are three issues") with the higher-order insight the group actually implies ("the business is losing money on its fastest-growing product").
- Write and deliver an answer-first recommendation (leading with the governing thought, not building to it) and run the live-synthesis formats a consultant actually faces: the 30-second case-interview close, the steering-committee interim update, and the one-page recommendation, each stating answer · because · but · now.
- Map synthesis onto Rumelt's kernel (diagnosis → guiding policy → coherent action) to distinguish a real recommendation from a list of good ideas, and use synthesis power-moves (de-averaging (the blended number lies), "what would have to be true," and the falsifier) to find the insight the average is hiding.
- Practise intellectual honesty in synthesis: state calibrated confidence, name the one fact that would change your mind, and never let an adjective do the work that evidence should, the low-self-orientation of an advisor, not the bluster of a salesperson.
- (AI-Augmented Consultant: R10.) Use an AI assistant to pressure-test a synthesis (draft the "so what?", red-team the pyramid for non-MECE gaps), while owning the governing thought yourself and tracing every figure it touches to a primary source: AI proposes the phrasing; you dispose of the judgment.
- Produce a hand-off note to deck craft, converting a finished pyramid into a storyline (the horizontal sequence of governing claims) and handing slide production to the Presentation galaxy, so you know precisely where the consulting synthesis logic ends and deck making begins.
Prerequisites & connections
Builds on. This is the third and final node of the Structuring branch, and it is where the first two pay off. CN1.01 (Issue Trees & MECE) gave you the tool for taking a problem apart; synthesis is the reverse operation (putting the answer back together) and it reuses MECE directly, because the key line of a governing-thought pyramid must itself be MECE or the argument leaks. CN1.02 (Hypothesis-Driven Problem Solving) gave you the day-one hypothesis and the one-day answer; synthesis is what a hypothesis becomes once the analysis has confirmed, killed, or reshaped it, the "so what" is the matured hypothesis, stated as advice. From the wider program you need almost nothing technical, only fluency with the kinds of numbers a case throws at you (margins, growth, unit economics, the CN2 casing skills run alongside this, not before it). What you do need is the willingness to kill your own darlings: synthesis is subtractive, and the analyst who cannot bear to cut a hard-won analysis that does not change the answer will never synthesize well.