Black Friday 2025 was one of the most informative periods I have experienced in a long time, not because everything went well, but because so many assumptions were tested at once.
From the outside, it may have looked like a typical BFCM cycle. Inside the accounts, it behaved very differently.
What Changed This Year
Thursday was almost as strong as Friday for several large brands. That alone was unusual. Thursday has historically been a buildup day, not a peak. This year, demand showed up earlier and with more intent.
Then Cyber Monday ended, and demand fell off fast. Faster than I have ever seen it drop. For some brands, the days immediately after Cyber Monday were record lows. The transition from high volume to near silence was abrupt.
That pattern tells a clear story. Consumers were not browsing. They were waiting for a deal, and once that deal disappeared, so did their willingness to buy. This aligns with broader economic pressure and increased price sensitivity.
Fewer purchases, more intention, and far less tolerance for ambiguity.
That context matters, because it framed everything that followed.
When the Platform Becomes the Risk
At the same time, Meta behaved in ways that forced difficult decisions. November was filled with bugs, delivery issues, and behavior that made very little sense at scale.
We saw spend concentrate into single ads inside large ad sets, even when those ad sets contained dozens of strong creatives. We saw overspend early in the day that could not be explained by real user behavior. We saw features toggle on without intent and flex formats behave unpredictably.
When you are spending at scale, these are not minor inconveniences. Meta is no longer just a marketing platform. It is a financial system, and millions of dollars flow through it.
No other financial system would be allowed to operate this way. You cannot imagine a trading platform where instruments fail, money continues to move, and explanations arrive days later. Yet this is effectively what advertisers accept.
Why We Had to Rebuild
Going into Cyber Five, we already knew that two high spend accounts needed to be rebuilt. The problem was timing. You do not rebuild multi seven figure accounts a week before Black Friday.
What saved us was preparation.
Before Cyber Five, we built a clean backup account with what I would consider a dream structure. Only the strongest ads from the year. Clean data flow. No legacy clutter. We warmed it up briefly, confirmed delivery and signal, then shut it down.
On Saturday of Black Friday weekend, one of our main accounts experienced severe delivery distortion. Entire ad set budgets began flowing into a single ad, not because it was outperforming, but because something inside the system broke.
We did not receive clarity from Meta until the following week, long after the sale had ended.
By Sunday, we made the decision to deploy the backup account and begin scaling it immediately, while managing damage control in the original account. It felt less like optimization and more like landing a plane with failing instruments.
What the Rebuild Revealed
What surprised me was how quickly the new account stabilized. Within days, spend distribution normalized, signal became consistent, and performance returned to levels we had seen before November.
That experience reinforced something important.
Account structure matters, but not in the way most people think.
I no longer think in terms of campaigns. I think in terms of signal flow. A campaign is simply a container for data. The real work is ensuring that data flows cleanly, consistently, and at sufficient volume for the system to learn.
Right now, consolidated CBO structures are working well for high volume accounts. One primary CBO, broken into ad sets by major product category when necessary, with spend minimums applied to ensure each category receives enough signal.
In practice, that often means committing roughly half of the budget to guaranteed signal, while allowing the remaining budget to float. The goal is not control for its own sake. The goal is coherence.
When dashboards show even distribution, stable CPAs, and predictable behavior, the system is usually healthy.
Why Legacy Thinking Breaks at Scale
What matters less today is overly complex segmentation, legacy retargeting stacks, or structures built for a version of Meta that no longer exists.
Retargeting is a good example. Many brands run it simply because they always have. In reality, unless you have a very large customer file, it often adds cost without meaningful incrementality.
Meta already retargets at the ad level, whether you explicitly ask it to or not. Leakage exists by design, and at scale, that leakage often provides the signal the system needs.
The Shift Toward Creative Systems
The biggest change heading into 2026 is where leverage actually lives.
The role of the media buyer is changing. Less time is spent rearranging ad sets. More time is spent understanding why a specific creative works, how it communicates, and how it can be iterated quickly.
Performance now comes from creative velocity, from new hooks, new edits, new combinations of message and format, delivered consistently into a clean signal environment.
Legacy structures often remain active because they are good enough to survive. But if most accounts were rebuilt from scratch today, they would look very different.
The Real Takeaway
Black Friday still works. Meta still works. But the margin for error is smaller, the systems are less forgiving, and the cost of outdated assumptions is higher than it used to be.
The teams that win going forward will be the ones who treat platforms as complex systems rather than black boxes, who respect signal as the foundation of performance, and who invest in creative as an operating discipline, not a side project.
That is where the real leverage is now.