
Brazil invested R$ 42,7 billion in digital advertising in 2025, a 12,7% increase over the previous year, with social media accounting for 55% of that amount , according to the Digital Adspend 2026 study by IAB Brasil in partnership with Kantar Ibope Media . A significant portion of this money goes to Meta Ads campaigns , and it is precisely here that many advertisers are losing efficiency without understanding why. The platform has changed its architecture, but much of the market still operates with the logic it learned in recent years.
The change happened in the most structural place possible: in the system that decides who sees each ad. And this reorganizes the work of those who manage paid traffic in retail and digital marketing.
In this article you will read:
Meta Ads in 2026 in five points
- The delivery engine has changed: Meta completed the global launch of in October 2025. Project Andromeda, a system that began defining the audience based on the creative content, rather than manual segmentation.
- The creative side turned to segmentation: Stacked interests, lookalikes, and fragmented ad sets have lost traction because they compete with a system that reads the ad itself to predict who will respond.
- Advantage+ has become the standard: Since February 2026, Meta has unified campaign creation and enabled automation by default; manual structures are being discontinued.
- Volume and variety of creative They became the main operational investment, not audience refinement.
- The quality of the conversion signal (Pixel + Conversion API) is no longer a technical detail; it now determines the speed at which campaigns learn.
From audience to creative: what automation has changed in Meta Ads
For more than a decade, managing traffic in Meta Ads was, in practice, about choosing who the ad would appear to. The work revolved around interests, behaviors, age ranges, and lookalike audiences; the creative was only the second step.
However, this order has been reversed. With Andromeda, the system reads the creative signals (image, video, text, format) and predicts who is most likely to convert, distributing the ad accordingly. Meta itself , in communications since 2025, has started recommending broad audiences and treating creative diversification as the main driver of performance.
The consequence is straightforward: those who maintain campaigns built on layers of lookalikes and dozens of segmented ad sets are not only using a less efficient tactic, but are also delivering a fragmented signal to the algorithm, which hinders its learning process instead of helping it.
It's worth comparing the two operating methods:
| Factor | Model until approximately 2024 | Model in 2026 |
| Main lever | audience segmentation | Creative diversity |
| The role of the manager | Define who the ad appears to. | Feeding the system with creative concepts and a clean signal. |
| Ideal structure | Many segmented sets | Established campaign, broad reach. |
| lookalikes | Central strategy | Supporting input, not mandatory. |
Advantage+ as standard and the end of fragmented manual structures.
In February 2026, Meta merged manual and automated creation workflows into a single interface, with AI optimization enabled by default in all new campaigns . In March 2026, reinforcements followed: consolidated budget controls, expanded audience controls within Advantage+ campaigns , new creative optimization signals, and the ability to set a budget cap for existing clients.
This ceiling is useful; without it, the algorithm tends to concentrate investment on those who have already purchased, because they convert more easily, which inflates the apparent ROAS and stifles the acquisition of new customers . Being able to control the ratio between acquisition and retention within the same campaign gives the manager back a strategic decision that automation alone would poorly resolve.
There's also a technical requirement that changes budget planning: to move out of the learning phase, Meta Ads requires around 50 optimization events per week per ad set . Budgets that don't generate this volume get stuck in "limited learning," and performance doesn't stabilize. Focusing budget on a lean structure has become a prerequisite, not a preference.
Volume and diversity of creative talent: the new currency.
The system needs raw material for testing, and the demand for volume has increased. Market recommendations today revolve around 15 to 50 active creatives, with a library of 10 to 20 truly distinct concepts. The key word is distinct: changing only the text on the same image tends to be read by the algorithm as repetition, and accounts without real variation can be penalized with higher delivery costs.
True diversification means covering different angles and intentions: social proof with real testimonials, problem and solution for those still discovering the need, direct offers with urgency for those already familiar with the brand, educational content in native Reels and Stories formats. Each approach generates a different signal, and it is this variety that allows Meta to find audiences that manual segmentation would never reach.
The conversion signal as a structural advantage
No amount of creative volume can compensate for a poor conversion signal. With browser cookie restrictions , relying solely on the pixel leaves the algorithm making decisions with incomplete data. The Conversions API (CAPI) sends events directly from the server to Meta, without depending on the browser, and recovers a good portion of the conversions that the pixel misses.
The indicator that summarizes the health of this signal is the Event Match Quality (EMQ) , measured from 0 to 10 in the Event Manager. Accounts with pixel and CAPI running together and an EMQ above 7 give Andromeda faster learning and a consistent advantage over competitors that still operate only with the browser pixel. It's an unglamorous job with a very high return.
Metrics that are becoming more important
Monitoring only ROAS and cost per lead reveals the symptom after it has already appeared in the results. CPM (cost per thousand people reached) acts as an early warning of fatigue: when it rises steadily, it signals that the algorithm is running out of responsive audiences for the current creatives. The correct response, in this case, is not to adjust the bid, but to renew the creative.
Useful tracking in 2026 combines four metrics: CPM per creative (fatigue), ROAS and CPA (results), EMQ (signal quality), and ad library renewal rate . The decision to change creatives becomes based on performance, not calendar.
The new role of the traffic manager.
Here's the least noticeable change: the traffic manager has come to handle three roles: creative strategist, signal engineer, and interpreter of the buyer's journey.
The part that the algorithm doesn't do (and won't do) is understand the business. It doesn't know the client's margins, it doesn't know which product needs turnover, it doesn't define which pain point to communicate to those who haven't yet decided to buy. Translating this knowledge into creative concepts that the system can use is, today, the most valuable work. Market analyses from 2026 already indicate that the quality of the creative content accounts for more than half of an account's performance in Meta Ads.
It's not a loss of control, but rather a shift of it: from the audience tab to the creative tab, to the conversion signal, and to the journey analysis. Whoever understands this first captures the advantage that always exists when an architectural change hasn't yet been assimilated by the majority of the market.
For those operating complete sales funnels, this logic directly relates to what already happens on other media platforms. It's also worth checking out our analysis of automation in Google Ads , where the same automation trend redefines the role of the manager.
Talk to people who have already made this transition.
Meta Ads in 2026 rewards those who feed the system with diverse creative, clean conversion signals, and an accurate reading of the buyer's journey. Setting up this operation requires method, from campaign structure to creative production and CAPI and EMQ configuration.
If your account is still operating under outdated logic, or if results have dropped without a clear explanation, it's worth talking to the paid traffic experts at e-Plus . With over 13 years of experience in e-commerce and performance projects, we've helped retailers make this transition without losing predictability of results. And if you want to delve deeper, explore the other blog content about media, data, and conversion.
FAQ – Frequently Asked Questions about Meta Ads in 2026
What is Project Andromeda in Meta Ads?
Project Andromeda is the ad delivery system that Meta completed globally in October 2025. It replaced audience-based targeting with a creative-based model: instead of using interests and lookalikes, the system reads the ad and predicts the right audience from behavioral signals.
Is it still worthwhile to use lookalike audiences in Meta Ads?
Lookalike audiences are no longer a primary strategy for most accounts. Under the new system, they function as a supporting input, not a rigid limit. The exception is usually new accounts with small budgets and limited conversion history, where a similar audience still helps as a starting point.
How many creatives do I need to have active in Meta Ads by 2026?
Market recommendations suggest 15 to 50 active creatives, with at least 10 to 20 genuinely different concepts. Volume matters because a few ads concentrate the majority of results, and the system needs real variety to find those who convert.
What is EMQ and why does it matter in Meta Ads?
EMQ ( Event Match Quality ) is a score from 0 to 10 that measures the quality of conversion data sent to Meta. An EMQ above 7, with the pixel and Conversion API running together, accelerates the algorithm's learning and reduces the cost of acquisition compared to accounts that rely solely on the pixel.