Rebuilt marketing financial forecasting from 10% variance against actuals down to 1%, embedding the model as the standard for budget planning across the brand portfolio.
→ Forecast variance reduced from 10% to 1%.
Problem
The Marketing Operations team was forecasting with variance of up to 10% against actuals, creating commercial uncertainty and undermining finance’s confidence in the team’s numbers across multiple brands.
Context
Forecasting accuracy sat within my wider remit for the team’s delivery discipline, not as a standalone finance exercise separate from campaign work.
Business objective
Redesign the forecasting process so variance came down significantly, without adding so much overhead that it slowed campaign delivery.
Customer/user objective
Finance partners needed forecasts they could actually rely on for budget governance across multiple brands, rather than numbers that needed independent verification.
Constraints
Any new process had to work without adding so much overhead that it slowed campaign delivery, since the team’s core job was still delivering campaigns, not producing forecasts.
Stakeholders
Finance, who relied on the team’s forecasts for budget governance, and the Marketing Operations team who would need to work to the new model day to day.
Research and discovery
I audited the existing forecasting methodology first, to understand exactly where the variance was coming from, finding poorly documented assumptions, inconsistent data sources, and no structured review cadence.
Options considered
I considered a lighter documentation fix for existing assumptions versus rebuilding the model itself with monthly review built in. I chose the rebuild, since documenting flawed assumptions more clearly wouldn’t have fixed the inconsistent data sources underneath them.
Prioritisation
Rebuilding the model around clearer assumptions and better historical data integration came first, since the monthly review process would only catch drift if the underlying model was sound.
Delivery
I rebuilt the forecasting model around clearer assumptions, better integration of historical data, and a more rigorous sign-off process, then introduced monthly forecast reviews with variance analysis built in and trained the team on the new model.
Decisions made
Introducing a monthly review cadence, rather than relying on the rebuilt model alone, was the decision that mattered most, it caught drift early instead of letting it compound silently over the year.
Trade-offs
The monthly review added a recurring commitment the team hadn’t had before. I judged that worthwhile given the commercial uncertainty a 10% variance was creating.
Business outcome
Forecast variance came down from 10% to 1%, and the model became the standard approach for marketing budget planning across the brand portfolio, still in use after the team structure later changed.
Customer outcome
Finance partners could plan against numbers they trusted, rather than treating marketing forecasts as directional at best.
Lessons learned
A forecasting model is only as reliable as the assumptions and data sources feeding it, a review cadence on top of a flawed model just catches problems later rather than preventing them.
What I'd improve today
I would automate the historical data integration from the outset rather than relying on manual pulls in the early cycles, since that manual step was where drift crept back in fastest, before the monthly review had a chance to catch it.
