Allocate budget to channels based on their marginal return, not their share of last-click conversions. Start with your best-performing channel and add spend until returns flatten, then move to the next.
Most marketing teams split budget based on what worked last year, what the channel rep recommended, or what feels balanced. None of those approaches connects spend to revenue. This guide covers a practical method for allocating budget across channels using measurement data, so every pound goes where it earns the most.
Why last-click data gives you the wrong answer
Last-click attribution (where 100% of credit goes to the final touchpoint before a sale) makes paid search look brilliant and social look useless. But the customer who searched your brand name was probably already convinced by a display ad or a YouTube video. Allocating purely on last-click starves the channels that do the convincing and over-funds the ones that just collect the credit.
A better starting point is incremental contribution: how many sales did each channel generate that would not have happened without it? Marketing Mix Modelling (a statistical technique that uses your historical spend and sales data to calculate how much each channel contributed) gives you this figure for each channel in your plan.
The principle of diminishing returns
Every channel has a saturation point. Your first 10,000 on paid search might return 40,000 in revenue. The next 10,000 might return 25,000. The 10,000 after that might return 12,000. At some point, adding more spend returns less than it costs. This is diminishing returns.
Budget allocation is the exercise of finding where each channel sits on that curve and shifting money from saturated channels to under-invested ones. You are not chasing the channel with the highest average return. You are chasing the channel with the highest marginal return on the next pound you spend.
A step-by-step allocation process
- Gather at least 18 months of weekly spend and sales data per channel.
- Run a Marketing Mix Model or work with a partner who can, to get the response curve for each channel.
- Rank channels by their current marginal return: what does the next pound in each channel earn?
- Move budget from the lowest-ranked channels to the highest until the marginal returns equalise.
- Set a floor for brand-building channels (TV, display, video) that have a long payback period.
- Review the allocation every quarter, not just at annual planning.
Studies consistently show that 10 to 20 percent of marketing budgets sit in channels past their saturation point. Reallocating that spend to under-invested channels can lift revenue by 15 to 25 percent with no increase in total budget.
Setting channel floors and ceilings
Pure marginal-return optimisation would tell you to put everything into one channel until it saturates. In practice, you need a spread for resilience (algorithm changes, auction volatility, platform outages) and for brand building, which takes time to show returns. Set a minimum floor for each strategic channel based on what share of voice you need to maintain, then optimise above that floor.
What to do without a Marketing Mix Model
If you do not yet have MMM data, run geo-based experiments. Switch off or scale back spend in a test region and compare sales to a matched control region. This gives you a rough read on incrementality before you have a full model. Pair that with platform-reported ROAS broken down by spend bracket and you can spot which channels look less efficient at higher spend levels.
When to revisit your allocation
Your allocation from six months ago is already out of date. Competitor spend changes, seasonality shifts, and platform auction dynamics all move the saturation curve. Build a quarterly review cadence where you update your model inputs and recheck whether your allocation still matches the data.
How do I allocate budget if I have no historical data for a new channel?
Start with a small test budget, typically 5 to 10 percent of what you would allocate to a proven channel. Run for 8 to 12 weeks to gather enough data to estimate a response curve. Do not scale until you have evidence of positive incremental returns.
Should brand spend and performance spend come from the same pot?
Ideally, yes. Separating them creates a false choice and makes it harder to measure the full journey. If your finance team insists on separate budgets, at minimum ensure your measurement framework captures how brand spend supports downstream performance conversions.
Our MMM shows TV has a strong return but our CFO wants digital-only. How do I handle that?
Present the model output alongside a simple scenario: here is what our projected revenue looks like with TV in the plan versus without it, at the same total budget. Put the decision in revenue terms, not channel terms. CFOs respond to revenue forecasts.