GM: 48-Model Delivery

The Problem

General Motors has 48 nameplates, each requiring its own marketing mix model built and maintained separately. Results had to be delivered on four different cadences simultaneously: weekly, monthly, quarterly, and annually. The scale was unlike any other engagement — and there was no existing process that could handle it.

The delivery structure added another layer of complexity. Engineering was done by a team offshore. Client-facing work sat with a small US team. I was part of that US team — which meant sitting in the middle of both relationships at once, responsible for what came in from the offshore team and what went out to GM.

What I Did

As part of the US delivery team, I sat between the offshore engineering team and the client — responsible for what came in and what went out. I worked directly with engineering on data collection and model validation: making sure inputs were clean, outputs were correct, and nothing slipped across any of the 48 models on any given cycle.

At that volume, improvising wasn’t an option. I helped build the QA processes that gave the US team real visibility into what engineering was producing — structured validation checks, clear handoff points, and consistent standards across models. With four timeframes running simultaneously, the question was never just whether a model was correct, but whether the right model was ready at the right time. The processes made that trackable instead of reactive.

The Impact

48 models. Four delivery timeframes. One small US team. We delivered on all of them — at a scale the firm hadn’t handled before. Issues surfaced in validation, not in client meetings.