Why an ad needs a fair opportunity
A creative can only be judged after accumulating enough delivery to generate real market information. If you place many ads inside the same ad set, the platform tends to quickly concentrate delivery on one or two early winners, leaving the rest with almost no impressions, which distorts any conclusion you draw by the end of the week.
Especially on Meta, the main challenge with larger budgets is creative fatigue: frequency spikes, the same people see the ad too often, and results drop. You need an account structure that tests creatives at scale without some starving others. In that class, I recommended giving each ad its own dedicated ad set.
On Google Search this problem barely exists in the same way, because the auction and search-intent logic works differently. It is mostly at the top of the Meta funnel, where scale and repeated exposure are possible, that fatigue becomes decisive in judging an ad.
If your ad never received enough impression volume to draw conclusions, you don't actually know whether it works. You only know you never gave it a fair chance, and that difference should change how you decide what to pause and what to keep running.
Volume of variables, not volume of luck
With large budgets there is no way to rely on a single piece of copy, a single hook, or a single landing page and expect it to sustain scale. You cannot simply keep spending money using too few variables; scaling requires continuously producing new communication hypotheses.
This does not mean testing everything at once without criteria. It means that as budget grows, so does the need to keep producing new variables, so delivery does not stay locked onto the same two or three assets for too long, which would only accelerate the fatigue described above.
With small budgets the reality is different: there is less room to diversify, because spreading a small budget too thin leaves each test without enough critical mass to validate anything rigorously. With larger budgets, that critical mass is easier to reach, which allows real learning in less time, buying market information faster than a few euros a day could.
What I expect from a media buyer in the AI era
How to build a lead funnel that the sales team can actually work
Diversifying the funnel, not just the creative
A common mistake is concentrating the entire budget on the same product, platform, placement, or message. More budget should mean more risk diversification, not more concentration on a single asset: a large budget failing on one single front hurts far more than several smaller budgets spread across assets.
I favor having a ladder of products and services, with the largest share of budget going to entry-level offers that are easier to sell, even if they are not the most profitable on their own. Not every product or campaign needs to be profitable in isolation, but the business as a whole must be profitable by the end of the month, before other costs.
An entry product may have lower direct profitability and still be essential, because it feeds the funnel and assists sales of other, more profitable products. That is why it's worth looking at attribution models rather than being a hostage to a spreadsheet: some campaigns don't sell much directly but help other campaigns sell more, and cutting them just because they look unprofitable on their own can damage the entire funnel.
Testing as routine, not exception
With large budgets, testing is part of the daily routine. I never assume the account is already optimized, even when current results look satisfactory. With little money, testing too many hypotheses at once spreads the budget too thin and leaves each test without enough data.
With larger budgets, there is more capacity to run tests with real substance, because each variation receives enough delivery for the observed differences to better reflect market behavior rather than noise. That does not remove the need for discipline, it only changes the scale of the problem. A larger test spend reduces the risk of deciding based on noise, but on its own it is not statistical proof of anything.
The daily goal is simple to state and hard to execute: look for what works. Once you find it, you scale it for as long as it lasts, because some messages and products are only worth riding for days, weeks, or months before losing steam. Only after finding what works does scaling become easy, which is why I describe testing as a permanent process rather than a phase that ever concludes.
Broad audiences for large budgets
With large budgets, it is usually harder to scale using very narrow audience segments. On Meta, forcing tight segmentation with a lot of money tends to saturate that audience quickly, and results become diminishing: price goes up and volume goes down, an effect practically unavoidable past a certain investment level.
My recommendation is to go broad with most of the budget, perhaps keeping one thin layer of segmentation, such as a specific age range, without narrowing by interest, gender, or region within the country. The same reasoning applies to Google Shopping: instead of advertising just one or two products, it helps to run a broad product feed and a search campaign structure that covers not only the most obvious searches but also people searching by pain point or by solution.
This does not mean abandoning narrower segmentation altogether. I reserve strategies like sequential remarketing for a smaller slice of the overall strategy, even though that slice can carry a higher return, because those campaigns depend on already having enough volume of people who passed through earlier pages and ads. Most of the budget, however, still goes to the top of the funnel, where the broad audience sits.
Finding the point where profit stops growing
The final idea is arguably the most concrete: the sensitivity test on ROAS and profit, which I informally call the sweet spot, or in my own words, the nutella point. The premise is that doubling a campaign's budget rarely doubles its results, and accepting that separates reasonable decisions from naive ones.
Past a certain investment level, returns become diminishing, and that threshold is not fixed: it tends to drop as a campaign ages and creative fatigue builds up, as described earlier.
When advertising my book, I started at around 100 euros a day. Over time, having already reached most of my relevant audience, I reached a point where 10 euros a day produced nearly the same sales as 20 or 30 euros a day. At that stage, tripling the budget did not bring proportional returns; it was the point past which extra money stopped translating into equivalent sales.
The way to find that point is simple to describe though it requires discipline: raise and lower the budget over time and measure profitability at each level. Only by comparing different investment levels on the same asset can you tell whether doubling, holding, or cutting the budget makes sense. This logic of increasing or diminishing returns applies to small budgets too: if a campaign generates no return with one euro a day, it is unlikely to generate one with fifty.
What this means in practice
Putting these ideas together, the question of whether the ad failed or never got enough budget to be tested has a practical answer: before deciding a creative doesn't work, it is worth checking how much delivery it actually received within the account structure, not just how much the whole campaign spent overall.
It is also worth checking whether the budget was distributed so every variation had a real chance to generate clicks and conversions, or whether it stayed concentrated almost entirely on the first pieces to take off early. And before scaling or cutting an entire budget, it is worth testing gradual increases and decreases, rather than doubling or killing it outright, to find the point where profit is still growing, knowing that point shifts over time as the campaign wears out.
This way of thinking applies just as much to someone managing tens of thousands of euros a month as to someone working with modest budgets: the difference lies in how fast you buy data, not in whether ads deserve a fair chance to prove their worth.
For more on paid traffic and digital marketing, see the AI and Digital Marketing hub.
Source note
This reflection is based on my class Sextas Ímpares #91, "Como Gerir Grandes Orçamentos em Anúncios," which I recorded and published on YouTube: https://www.youtube.com/watch?v=SrJOaLpJG28. From 07:05 I talk about giving ads a real chance before judging them; around 23:58 I explain the sensitivity test on ROAS and profit; and near 26:30 I share the example of my book ads, where I found that point of diminishing returns.