Optimising Global Media Budgets
How to allocate marketing spend across markets when each one has different maturity, different competition, and different data quality.
Global media planning is the hardest problem in marketing measurement. Local MMM in the UK is a solved problem. Optimising £500m across 40 countries with wildly different brand positions, data quality, and competitive dynamics is not.
Why the standard approach fails
The default global allocation method is some flavour of "spend in proportion to revenue, weighted by growth ambition". This produces budgets that look defensible in a board pack and bear no relationship to where the marginal pound returns most.
It fails for three reasons:
- Diminishing returns are non-linear. A mature market like the UK may be on the flat part of its response curve while a smaller market with the same brand strength is still on the steep part. Proportional allocation overspends in the saturated market and underinvests in the growth one.
- Local channel mix matters more than total spend. £10m in India spent on the wrong channel mix delivers less than £6m spent on the right one. A purely top-down optimisation misses this.
- Brand effects cross borders. A US Super Bowl spot lifts brand metrics in markets where it never aired. Pure within-market optimisation undervalues anchor markets.
A workable framework
Three layers, run in sequence:
Layer 1 - Cluster markets. Group countries by maturity (brand awareness, category development, media cost), not geography. A good cluster might be "high-awareness, high-CPM, saturated TV" containing the UK, Germany, Australia, and Canada. Treat the cluster as a single optimisation unit at the top level.
Layer 2 - Allocate between clusters. Use a global MMM at cluster level to set the split between growth markets, mature markets, and emerging markets. This is the lever the CFO cares about.
Layer 3 - Optimise within clusters. Run local MMMs (or pooled hierarchical models for small markets that lack data) to set the channel mix within each cluster. This is where local market knowledge comes in.
The data problem
Most global advertisers cannot do this properly because their media data is fragmented across 15 agencies in 40 currencies with no consistent taxonomy. The single highest-ROI investment for a global effectiveness function is usually not a fancier model - it is six months of data engineering to produce a clean, harmonised global media spend table.
Build that, and the modelling becomes straightforward. Skip it, and no algorithm will save you.