Method
How the work is actually done.
Most analysis asks to be trusted. What follows is the decision engine underneath every engagement — five stages built so the work can be checked by the people who have to act on it, your own decision-makers and your own board, and challenged by a process that tries to break its own conclusions before you ever see them.
From the question to the next move.
Frame
Define the consequential decision precisely, in writing, before any research begins. Most bad analysis is the right answer to the wrong question.
Investigate
Build the evidence base across companies, markets, people, competitors, channels, and the external signals that move them.
Synthesize
Find the patterns, the contradictions, the risks, and the outcomes the evidence actually supports — with the probability attached.
Challenge
A separate adversarial pass tries to disprove the analysis. It runs its own searches and hunts for the source that changes the conclusion.
Decide
Translate the intelligence into an executive recommendation, the decision logic behind it, and the next actions in order.
PRISM gathers and synthesizes evidence in service of one decision. Information is the input; the decision is the product.
Evidence is classified, not just collected
Every claim that carries weight in a conclusion is labeled for what it is: verified against an independent source, inferred from evidence, assumed and said so, probable with the probability stated, or genuinely unresolved. The label travels with the claim into the deliverable — you can see it working in the sample analysis. An analysis that cannot tell you which of its statements are facts is an opinion with formatting.
The challenge is independent
Before anything is delivered, a separate adversarial pass — independent of the one that produced the draft, with its own live research — goes looking for the source that changes the conclusion. It is scored on what it finds wrong, not on agreement. Conclusions that survive are delivered with their challenge trail; conclusions that do not are revised, and the revision is recorded.
Confidence is a discipline, not a feeling
Every brief states how confident the conclusion is, which assumptions are most likely to fail, and — the part almost no deliverable includes — what specific evidence would change our mind. A recommendation that cannot name its own breaking point has not been thought through; it has been formatted.
A person signs it
Machines do the reading at a scale no team can match. The judgment — what the evidence means for your decision, and what to do first — is human, and it carries a name. The person who scoped the question is the person who will defend the recommendation to your board.
What this page deliberately does not explain: the specific tooling, models, and internal procedures behind the method. Those are the firm’s own. Publishing the engine’s shape rather than its parts is the point — it tells you exactly what standard to hold the work to, and what to ask for when a conclusion does not meet it.
The shape of the work
One complicated question enters. Structure comes out.
- Evidence
- Assumptions
- Trade-offs
- Probabilities
- Unresolved
A recommendation you can interrogate.
Follow the evidence.
See what is verified, what is assumed, and what remains unresolved. An independent model challenges the analysis before a human signs the recommendation.
- Question
- Evidence
- Analysis
- Challenge
- Decision
The sources. The uncertainty. The reasons to act.