A pricing recommendation no one can interrogate is a recommendation no one will approve. So we built the engine inside-out: explanation first, score second.
Each recommendation decomposes into weighted drivers — demand headroom, competitive position, cost movement, and the guardrail that clipped the final number. The weights are the model; there is no hidden layer doing the real work.
That constraint shapes everything. We favor interpretable estimators over opaque ones, and we attach the plain-language reason to the response itself, so the price and its justification can never drift apart.