We document every assumption so it can be challenged.
Every model output we deliver carries a written register of the assumptions behind it — economic, demographic, behavioural, methodological. Versioned, time-stamped, reviewable. If you cannot challenge an assumption because it is not written down, we have not finished the work.
We finish the job, including the boring last 10%.
Most consulting work delivers 90% of the substance and leaves the documentation, lineage, handover and sign-off thin. We finish those. The last 10% is where the work survives our involvement — or does not.
We name the limits of what our models can tell you.
A model has a domain of applicability. Inside the domain it is useful; outside it is dangerous. We document the boundaries explicitly. We do not let confidence intervals stand in for limits the model cannot describe.
We tell clients hard truths early.
If a number does not reconcile, if a method is breaking down under a new scenario, if a deadline is at risk — we say so the day we see it, not in the steerco the week before sign-off. The cost of an early hard conversation is always less than the cost of a late one.
We credit and protect the people who built the work.
On every engagement we name the consultants who did the work, junior and senior, internal and on the client side. Knowledge stays hoarded by individuals when nobody can see who contributed. It gets shared when credit is visible. The same applies to responsibility: clear review paths and escalation routes mean problems are handled as process issues before they become personal blame.
We tell you when you do not need more of us.
Engagement length is a deliverable, not a target. When the work is done, we say so. When another firm is the right fit for the next phase, we say so. Long client relationships come from being right about scope, not from extending it.
We hand over work you can run without us.
At handover you get the working models, the documentation and the source — not a black box that only we can open. No proprietary runtime, no annual licence to keep your own model alive, no dependency on us to make a change. Our independence cuts the same way: nothing to sell you, no software book, no vendor or hyperscaler kickback. The only thing tying you to us is whether the work is good.
These are testable claims. If you have engaged us and we have not met any of them on the work in front of you, we want to hear about it.