How to Allocate Your Marketing Budget: The 70-20-10 Rule in Practice
A practical breakdown of the 70-20-10 budget split - what counts as 'proven', when to flip the ratio, and the exact signals that tell you it's time to move money.
You've got RM 20,000 to spend this month and everyone in the room has an opinion. Sales wants more on Google because leads closed well last quarter. Marketing wants to test TikTok because a competitor's ad is everywhere. The founder wants to double down on whatever worked in Q2. Nobody in that room is wrong - they're just optimizing for different things. This is exactly why most conversations about how to allocate marketing budget by channel go nowhere: people argue about the percentages before anyone agrees on what each percentage is actually for.
70-20-10 is a risk framework, not a spending formula
The numbers aren't the point. The point is that your budget should be split by how much risk you're willing to take with each portion of it. 70% goes to channels you already know will bring a return - this is the money that keeps the lights on. 20% goes to structured tests with a clear win/lose line drawn before you spend a ringgit. 10% goes to bets - new formats, new platforms, new positioning - where you've already accepted the money might not come back, and you've written down exactly what 'not coming back' looks like.
The split isn't 70-20-10. The split is: money you know will return, money you're betting will return, and money you've already accepted might not.
What actually qualifies as your '70%'
This is where most teams get it wrong. They call a channel 'proven' because they've run it for two years, not because it's actually stable. Longevity is not the same as stability. A channel earns its way into the 70% tier when its CAC or ROAS has held inside an acceptable band - say, within 15% either direction - for at least 3 consecutive months. Not one good month. Three.
- CAC or ROAS has stayed within a defined range for 3+ straight months, not just one strong sprint
- You can predict roughly what this month's spend will return before you've spent it
- The result doesn't collapse when one creative, one person, or one algorithm update changes
- You've deliberately stress-tested it - cut budget, swapped an audience, paused a top ad - and it recovered instead of falling apart
If your brand is under 12 months old, flip the ratio
Early-stage brands love to apply 70-20-10 to a channel that isn't actually proven - it's just the one they like, or the one someone on the team is comfortable running. Putting 70% of your budget into an unproven channel isn't stability, it's guessing with more money attached. If you don't have a channel that meets the criteria above, your job right now is to find one, not to defend one. Flip the ratio - something closer to 20% held on your current best guess, 60% into structured testing across two or three channels, 10% into a genuine long shot. The goal of this phase isn't efficiency, it's discovery.
The signal to move money is data, not the calendar
Most teams review budget allocation once a quarter because that's when the meeting is scheduled, not because that's when the numbers actually changed. The better approach to how to allocate marketing budget by channel is to define your triggers in advance and check them monthly, so the decision is already made before the meeting starts.
- Proven channel's CAC rises for 2 consecutive months - move next month's incremental budget into the test tier instead of defending the old spend
- A test channel hits its target CAC or ROAS twice in a row - promote it toward the proven tier with a real budget increase, not a note in a slide deck
- A bet channel misses its pre-set kill criteria - cut it that week, don't extend it 'one more month' because someone's attached to the idea
Budget allocation isn't a planning exercise you revisit once a quarter - it's a live decision that should move every time a channel's numbers move. If your 70-20-10 split hasn't changed in six months, that's not evidence you've found stability. It's evidence nobody's been checking.
Frequently asked questions
What percentage of marketing budget should go to paid ads versus other channels?
There's no fixed percentage that applies to every business - the honest starting point is 70-20-10, where 70% goes to whatever channel is already delivering a stable, predictable return (often paid ads, sometimes not), 20% goes to structured tests of new channels or formats, and 10% goes to genuine long-shot bets. Which channel earns the 70% depends entirely on your own CAC and ROAS history, not on what's typically true for your industry.
How often should I review my marketing budget allocation?
Check the underlying numbers monthly, not quarterly - the split itself might not change every month, but the data that should trigger a change needs to be reviewed that often. Set the trigger in advance: for example, if a proven channel's CAC rises two months in a row, that's your cue to shift new spend elsewhere, rather than waiting for a scheduled quarterly meeting to notice it.
Should I hire an agency to manage my marketing budget allocation, or handle it in-house?
Most agencies are reasonably good at running the 20% test tier - launching campaigns, iterating creative - but far fewer are disciplined about defining kill criteria for the 10% bet tier or being honest when a 'proven' channel is quietly decaying. Before hiring anyone, ask them directly what their exit criteria are for a test before they launch it. If they can't give you a specific number, that's worth noting. Pricing for this kind of ongoing budget strategy work varies widely and usually depends on whether you're paying for media management, strategy, or both.
What CAC or ROAS number makes a channel 'proven'?
There is no universal benchmark - a CAC that's healthy for one business would be unsustainable for another. The practical test isn't hitting an industry average, it's stability: has your CAC or ROAS stayed within a defined range, roughly ±15%, for at least 3 consecutive months without you having to intervene. A channel that occasionally spikes to a great number isn't proven - it's inconsistent, and inconsistent channels don't belong in your 70%.