Dynamic Pricing Optimizer

Overview

A freelance engagement rebuilding a client’s dynamic pricing system, which recalculated prices across a large catalog on a schedule and had grown slow enough to lag behind real market movement.

Approach

The core loop ranks candidate price points per item against a cost/demand model, keeping only the top-scoring candidates at each step rather than exhaustively evaluating the full price range:

\begin{algorithm}
\caption{Candidate Price Ranking}
\begin{algorithmic}
\PROCEDURE{RankPrices}{$$item, candidates, model$$}
    \STATE $$scores \gets \emptyset$$
    \FOR{$$p \in candidates$$}
        \STATE $$scores[p] \gets$$ \CALL{ExpectedMargin}{$$item, p, model$$}
    \ENDFOR
    \RETURN \CALL{TopK}{$$scores, k$$}
\ENDPROCEDURE
\end{algorithmic}
\end{algorithm}

Result

Cutting the candidate set early, rather than scoring every possible price, was most of the runtime win — the remainder came from batching the model calls instead of issuing them one item at a time.