The question usually comes in the same format: "I have R$ 20,000 per month, where should I allocate it?" The honest answer is that no one knows before testing — but there is a method to find out quickly and without wasting budget.
First: separate testing budget and scaling budget
This is the most important and often ignored decision.
- 70% to 80% in proven channels: where you already know the cost per acquisition and the operation can meet the demand.
- 20% to 30% in testing: new channels, new formats, new audiences. This budget is planned spending, not waste — as long as there is a clear question and deadline.
Without this separation, the company either never tests anything or tests everything at once and fails to conclude anything.
Criteria for choosing a channel
Purchase intent
Paid search captures those already looking for a solution. Social media creates demand among those who are not yet searching. Both roles are necessary at different times — but search is usually the first place for budget in high-ticket, short-cycle businesses.
Capacity to serve
It’s no use generating 300 leads if the team can only handle 60. The channel needs to be scaled according to the commercial operation, not the desire to grow.
Product format
Visual products perform well on social media. Technical services perform well in search and LinkedIn. Impulse products perform well in short videos.
Data structure
A channel without reliable measurement turns into faith. If you can’t attribute sales to it, either create a controlled test or don’t scale.
A common starting point
For small and medium-sized operations, a reasonable initial division — not a rule, just a starting point:
- 40% paid search (Google, Bing): capturing existing demand.
- 35% paid social (Meta, TikTok): creating demand, volume of creatives.
- 15% remarketing: recovering those who have already interacted.
- 10% testing: a new channel or format every 60 days.
Adjust based on reality: local businesses tend to concentrate on search and Meta; high-ticket B2B tends to balance search and LinkedIn; visual e-commerce tends to lean towards social.
How to test without deceiving yourself
One variable at a time. Testing a new channel with a new audience and new creative at the same time does not generate learning.
Minimum period of 30 days and minimum volume of conversions. With 5 conversions, any number is noise.
Criteria defined beforehand. Write: "if the cost per acquisition stays below R$ X and the volume exceeds Y, I will scale by 30%". Without this, the decision becomes opinion.
Same attribution window. Comparing channels with different windows is like comparing different things.
Scale without losing efficiency
Increase in steps of 20% to 30% every 5 to 7 days, monitoring cost per result. Large jumps often move away from learning and hurt performance. And scale what already works before opening a new front: there is almost always room in audience, creative, and landing page before needing another channel.
Also read:
- paid media consulting
- investing in paid media with a method
- incrementality with high CAC
- what changed in campaign management
The classic mistake
Distributing budget equally "for everyone to see". A pulverized budget does not reach statistical volume in any channel, no one learns anything, and the conclusion becomes "paid traffic doesn’t work".
Concentrate, test, measure, and expand. Media is hypothesis management — and a hypothesis without volume cannot be proven.

