How we work.
Four moves, in order. Slow at the start, fast where it counts, and we don't leave when the deck is done.
Diagnose, frame, model, follow through.
- 01
Diagnose.
We start with the decision itself, not the data. What's actually being decided, by whom, by when, and what's the cost of being wrong? Most engagements spend the first week here, and most of them discover that the real question is one rung up or one rung down from the one on the agenda.
Example: A $4B industrial we worked with last year arrived asking whether to acquire a competitor. Two days in, the real question was whether they should be in the segment at all. The acquisition stopped being the decision; the strategic exit became one. - 02
Frame.
We translate the decision into priors, evidence, and outcomes. This is where most engagements unstick — the question was never asked sharply enough, the assumptions were never written down, and the team was disagreeing about beliefs they'd never made explicit.
Example: A PE fund choosing between two add-ons spent six weeks arguing about EBITDA multiples. Once we wrote down each partner's prior on customer concentration risk, the disagreement resolved in an afternoon. The two priors were three standard deviations apart and nobody had noticed. - 03
Model.
We build only as much model as the decision warrants. Sometimes that's a one-page back-of-envelope. Sometimes it's a full posterior with sensitivities. Always: transparent, inspectable, and small enough that the people making the decision actually understand it.
Example: A consumer business deciding on a $90M capex commitment got a half-page model. Three inputs, two outputs, one explicit assumption that drove 70% of the variance. The model was less interesting than what it revealed: the decision was effectively a bet on one number nobody had been willing to argue about. - 04
Decide & follow through.
We don't hand over a deck and leave. We sit with the decision-makers when they make the call, and we come back six to twelve months later to score how it went. Calibration only improves if outcomes are scored honestly against the confidence going in.
Example: Of the engagements we revisited at twelve months, the teams scored their own confidence within ten percentage points of the calibration we'd assigned at decision time. The ones who'd been overconfident knew it before we showed them the numbers. That's what we're optimizing for.
Three things we do that most don't.
We never disappear after the deck.
Most consulting ends at the recommendation. Ours starts there. We sit in the room when the call is made, we read the memo before it goes to the board, and we come back to score the decision after the fact. The work is the follow-through, not the slides.
We tell you when we don't know.
Calibrated honesty is the product. We'd rather say 60% with an interval than 90% with conviction. Clients sometimes find this jarring at first; without exception, they come to prefer it. It's the difference between being told something true and being told something flattering.
We measure ourselves on calibration, not confidence.
We track our own Brier scores across engagements — how well our stated probabilities match observed outcomes. We'll show you ours. Most advisors won't, because most advisors don't keep score. We do, because the point of the work is to be right more often than we sound.
Practical matters.
How long is a typical engagement?
Most projects run four to twelve weeks. A decision quality audit is often shorter — three to four. Retainers run quarter-to-quarter; advisory relationships are open-ended. The shape follows the decision, not the calendar.
How do you charge?
Fixed-fee for projects, scoped after a discovery call. Monthly for retainers. We don't bill hourly and we don't take success fees — both create the wrong incentives. If a decision turns out to need less work than we quoted, we tell you, and we charge less.
Do we need to have our data in good shape before we call?
No. Half the time the engagement starts with the data being a mess, and the diagnosis surfaces that the data was never the bottleneck. We work with what exists. Where the analysis genuinely needs better inputs, we'll say so and we'll be specific about which ones.
Do you work onsite?
Sometimes. For projects involving sensitive material or where the team works best together, yes. For most engagements, a mix of in-person workshops and remote work between them. We travel for the calls that matter — the kickoff, the framing session, and the day the decision gets made.
What industries do you work in?
We're industry-agnostic by design. The method travels: industrials, financial services, healthcare, consumer, technology, energy. The constant is the shape of the decision — significant, hard to reverse, and made under uncertainty. We will tell you if a sector is genuinely outside our reach.
What's the smallest engagement you'll take?
A two-week decision diagnostic on a single call. Below that, the overhead of getting us up to speed swamps the value of what we can contribute. If you have something smaller, we'll often point you to a question to think through on your own and ask you to call back if you still want help.
What happens if we hire you and you tell us we don't need help?
It happens. We refund the unbilled portion of the engagement, we write you a short memo on what we saw, and we tell you the truth: this one you can run yourselves. The reputation we want is the one where executives call us for the next decision, not where we manufactured work on this one.