£100m+ delivered through smarter S&OP.
Forecast accuracy up significantly.

Reported sales weren’t measuring demand. They were measuring supply.

The manufacturer’s Sales & Operations Planning process was running on distorted signals. External events – launches, promotions, holidays – caused fluctuations in sales patterns that were hard to isolate from underlying demand. Worse, supply chain and production constraints meant reported sales often reflected what the factory could ship, not what customers actually wanted.

Demand signals came from everywhere: retail deliveries, wholesale shipments, order books, Google Trends, independent industry forecasts – each in its own format, on its own cadence, with none of it aligned to a common data model. Forecasts built at vehicle-line and market level still had to reconcile with the global figures used for planning. And newly launched vehicle lines added another problem: no sales history to forecast from at all.

Three constraints were non-negotiable:

• Sales data reflected supply, not real demand
• Demand signals scattered across formats with no common model
• New vehicle lines had zero historical data to forecast from

Five data sources, one demand model.

The team, led by our Fractional Head of AI, built a data engineering pipeline that ingested retail point-of-sale, wholesale shipments, order-book data, Google Trends consumer interest signals and independent forward-looking auto industry data. Each source was normalised, timestamp-aligned, and mapped into a single unified demand data model – a reusable data product the business could build forecasts on.

An event intelligence layer, fed by a global event calendar of launches, model refreshes, promotions and holidays, detected and quantified shocks to typical sales patterns, separating event-driven volume from baseline demand. For newly launched vehicle lines with no sales history, the team used analog forecasting: matching each new model to comparable past launches in the same segment to build credible demand curves from day one.

Hierarchical forecasting solutions reconciled the numbers top to bottom, combining bottom-up signals where the data was strong with top-down constraints where supply or strategic targets set the limits, so vehicle-line, market and global forecasts stayed self-consistent. The outputs fed directly into the S&OP cadence, giving planners transparent forecast drivers, event attribution and scenario comparisons.

Forecasts planners could finally trust.

The result was a Sales & Operations Planning process built on real demand rather than distorted supply signals. Planners could separate event-driven spikes from baseline demand, forecast new launches with confidence, and trust that line-level numbers added up to the global plan. That translated into more effective inventory allocation, reduced mismatch between supply and demand, and over £100m in quantified impact, plus a business better able to absorb shocks and reallocate stock with confidence.