Marketing Budget Allocation: How to Split Your Budget Across Channels
Budget allocation is the most consequential marketing decision — and most businesses do it based on convention rather than data.
The 70/20/10 split is the best starting allocation I know, for a simple reason: it resolves the tension between what works today and what will work in a year.
How it splits
70% — what you know works. Channels with proven returns, where you have at least three months of data. You don't experiment here, you execute.
20% — what looks promising. A new channel, a new format, a new audience that showed signs in a small test. Enough budget to find out whether it scales.
10% — what you've never tried. Something that might fail completely. This budget is meant to be lost, sometimes. When it isn't, it graduates to the 20%.
Why not 100% on what works
Because every channel saturates. Audiences exhaust, costs rise, competition arrives. An account performing perfectly today will return less in a year, guaranteed.
The 30% is insurance. Companies that run out of options cut exactly that part first, in a weak quarter.
How to split the first 70%
Not by platform reports — they all claim the same sale. Split by switch-off tests: turn a channel off for two weeks and watch total sales.
It's uncomfortable and feels wasteful. It's the only method that can't be fooled.
What gets left out of the maths
Content production. Ads need assets, and assets burn out. If your budget covers only distribution, in two months you're distributing something stale.
Tools. They accumulate unnoticed and become a serious line item.
Internal time. If someone in-house spends 10 hours a week on this, that's a real cost even if it doesn't appear on an invoice.
Adjustment cadence
Monthly: small moves within the 70%. Quarterly: reassess the split across the three buckets. Annually: reconsider whether the channel mix still makes sense.
Moving budget weekly, chasing report fluctuations, does more harm than good. Algorithms need stability to learn.
When 70/20/10 doesn't apply
A new company with no data. You have nothing to put in the 70%. Start with two channels, test properly, build the base.
A cash crunch. Then everything goes to whatever recovers fastest, and experiments wait. That's a correct decision, but it should be made deliberately, not by drift.
How many hours a week do you lose to repetitive work?
In 20 minutes on Zoom we work out which of them can run on their own. You leave with the answer, whether we work together or not.
Free call, no obligations. You leave with a concrete list of problems and automation opportunities, whether or not you choose to work with us.