Key takeaways
- Allocate marketing budget by the job each channel does in the funnel, demand capture or demand creation, not by last year's split or platform ROAS.
- Fund demand capture up to the point where extra budget stops buying extra sales, then move the rest to demand creation and testing.
- Check incrementality with holdouts, geo splits or platform lift tests, because platform ROAS rewards channels that harvest sales other channels created.
- Keep a fixed share for testing and rebalance monthly on CAC and MER; any percentages are starting points to adapt, not benchmarks.
Good marketing budget allocation starts with giving each channel a job in the funnel, then funding it according to how well it does that job. Fund demand capture (search, marketplace ads, retargeting) until extra budget stops buying extra sales, put the rest into demand creation (Meta, TikTok, content, KOL), and keep a fixed share for testing. Then rebalance monthly on CAC and MER, and check incrementality so the channels that harvest demand do not take credit for the ones that create it.
We do not start from a fixed percentage. We start from the process below.
Why last year's split is the wrong starting point
Most budget plans we see are last year's numbers plus or minus a bit. Or they follow whichever platform shows the best ROAS in its own dashboard.
Both fail for the same reason. Platform ROAS rewards channels that sit closest to the sale. Brand search, retargeting and Shopee search ads will almost always report the best numbers, because they meet buyers right before they pay. Cut the channels that created those buyers and, a few weeks later, the "best" channels start shrinking too.
So we start with roles, not reports.
How do you allocate budget by funnel role?
Every channel does one of two main jobs. Some do a bit of both.
“Give every channel a job, then make it prove it did the job.”
| Funnel role | Typical channels | What it does | How to judge it |
|---|---|---|---|
| Demand capture | Google Search, Shopee search ads, brand search, retargeting, Apple Search Ads | Converts people already looking | CPA, ROAS, impression share; watch for diminishing returns |
| Demand creation | Meta prospecting, TikTok, YouTube, KOL, content and social | Makes new people want what you sell | New customer CAC, MER, branded search growth, lift tests |
| Conversion support | Landing pages, CRM, WhatsApp follow-up, email | Turns interest into sales and repeat sales | Conversion rate, cost per qualified lead, repeat rate |
The order of funding:
- Fix conversion support first. A slow landing page or unanswered WhatsApp chats waste every channel above them. This is usually the cheapest gain.
- Fund demand capture up to its ceiling. Raise search and marketplace budgets while cost per sale holds. When each extra step buys noticeably fewer sales, you are near the ceiling. More budget there just buys the same buyers at a higher price.
- Put the rest into demand creation. This is what grows the pool that capture channels harvest next month.
- Ring-fence a testing share. More on this below.
A Bali resort selling to Australians and a Jakarta app chasing installs will end up with very different mixes. The process is the same.
Take the resort. Very few Australians search for a specific villa development in Nusa Penida until something has put it in front of them. So search is small and fills up fast, and most of the plan has to sit in creation: Meta and YouTube in the target market, content that shows the place, retargeting for people who watched. For a Jakarta delivery app in a category people already search for, it flips. Search and app store ads can absorb a lot before they hit their ceiling, and creation spend is there mainly to win new users from competitors.
Same framework, opposite answers. That is why copying another brand's split rarely works.
What does incrementality mean for budget decisions?
Incrementality asks one question: what sales would we lose if this channel stopped? Platform reports cannot answer it, because each platform counts the sales it touched, not the sales it caused.
Ways to test it, from simple to rigorous:
- Pause and watch. Turn a campaign off for two to three weeks and watch total backend sales and MER, not the platform dashboard. Crude but honest.
- Geo holdouts. Run ads in some cities or regions and not others for a period, then compare. Works well in Indonesia, where Jabodetabek, Surabaya and Bali behave like separate markets.
- Platform lift studies. Meta and Google offer conversion lift tests that hold back a control group. Eligibility and minimums vary, so check what your account can run.
- Watch the spillover. If branded search and marketplace searches rise after a TikTok or Meta push, the creation channel is working even when its own ROAS looks modest.
Run at least one incrementality check a quarter on your largest channel. The answer often moves more budget than any dashboard review.
How much should go to testing?
Enough that testing actually happens, and fixed in advance so it does not get cut the first slow week.
A common starting structure is 70/20/10: about 70% on proven channels, 20% on promising ones you are scaling, 10% on experiments. Treat these as illustrative starting points, not benchmarks. A new brand with no proven channel might put far more into testing. A mature account with stable demand might put less.
The principle that matters: every test needs a hypothesis, a success metric and an end date. "Try TikTok" is not a test. "Can TikTok GMV Max bring new buyers at a CAC within our ceiling over six weeks?" is.
How do you rebalance across channels?
Monthly, on business numbers:
- CAC by channel, against a ceiling set from margin and LTV
- MER across all channels, to check that platform claims add up
- Diminishing returns on capture channels: is cost per sale rising as budget rises?
- Test results, graduated into the proven bucket or shut down
Move money in steps rather than big swings. Most algorithms need time to settle after a large change, and you want to see the effect of each move. If you need the metric definitions, see ROAS, CAC, CPA and LTV explained.
How we plan channel budgets at FOXA
We map every channel to a funnel role in the first month, agree CAC and MER targets with the client, and fix a testing share before anything goes live. Then we review monthly and change structure quarterly.
For Bonvie, a marketplace haircare brand, that meant Shopee ads for capture, TikTok GMV Max and Meta CPAS for creation and recapture, judged together on marketplace sales. The result: 107x peak Shopee ads ROAS, 35x Meta CPAS ROAS and 5.7x orders from TikTok GMV Max (Bonvie story). The full channel playbook is in our Shopee ads and TikTok Shop guide.
For HappyFresh, budget followed new-user CAC rather than platform ROAS. New-user CAC fell -69% in 3 months, and new users grew +94% in a record month (HappyFresh story).
Next step
List your channels, write each one's funnel role next to it, and mark which ones you have never tested for incrementality. Then book a call and bring that list; it is the first thing we would ask for.
Questions we get asked
How should I split my marketing budget across channels?
Start by giving each channel a funnel role. Fund demand capture, such as search and marketplace ads, until extra budget stops adding sales. Put the rest into demand creation, such as Meta, TikTok and content, and keep a fixed testing slice. Then rebalance every month using CAC, MER and incrementality checks rather than platform ROAS alone.
What is the 70 20 10 rule in marketing budgets?
It is a common heuristic: roughly 70% on proven channels, 20% on promising ones being scaled, and 10% on experiments. It is useful as a starting structure because it forces a testing budget to exist. It is not a benchmark, and the right shares depend on your stage, category and how much demand already exists for what you sell.
How do I know if a channel is incremental?
Turn it off or down for a comparable group and see what happens to total sales. Options include geo holdouts, where some regions get no ads for a period, platform conversion lift studies, and pausing a campaign for a set time while watching MER and backend orders. If sales barely move without the channel, it was mostly taking credit for demand created elsewhere.
How often should I rebalance my channel budget?
Review monthly and make structural changes quarterly. Weekly shifts react to noise, and most channels need a few weeks to settle after a budget change. Move budget in steps rather than all at once, and rebalance immediately only when something breaks, such as tracking failing or a channel's CAC jumping well above your ceiling.
About FOXA DigitalFOXA Digital is a performance marketing agency in Jakarta, Indonesia. Since 2018 we have run paid media, tracking, SEO and content for apps, consumer brands, hospitality and attractions, judged on CAC, ROAS and leads.
Want us to look at your numbers?
30 minutes. No deck. You leave with three things to fix this week.



