]
Get Paid Media Ready for the Weekend, and the Quarter.
This chapter is an operating manual behind running the whole paid-media workflow across Meta, Google and TikTok. From how to size the next dollar, when to change the creative, and what to settle from the backend instead of the platform.
The Job, Not the Split.
Across the 2025 holiday accounts reviewed, the picture was less about a preset split and more about the role each channel played. Because margins, AOVs, and attribution windows vary by brand, a single blended ROAS would have been misleading. What mattered was each channel's job and how much room it still had to scale profitably.
| Meta | TikTok | ||
|---|---|---|---|
| Primary role | Prospecting + retargeting | Capturing existing demand | Conditional, creative-led |
| Why | The most consistent channel | Strongest at demand capture | Not a universal winner |
| Fits when | Almost always the base | Intent already exists | You have native short-form to keep testing |
The move that won wasn't a fixed split. It was shifting incremental budget toward whichever channel still had room to scale profitably. Attributed to Otis AI (Data Partner).
Directional, single-account BFCM figures anonymized. Attributed to Otis AI (Data Partner).
One System for the Whole Workflow.
Otis AI is the autonomous orchestration layer for paid ads. One system that plans, launches, optimizes, attributes, and reports across channels.
What makes that more than a dashboard is the vantage point. Otis is informed by more than 30,000 campaigns and 12M+ performance data points, channels, audiences, budgets, and creative. It uses those patterns to run the whole workflow in one place.
Named Otis case studies. General results, not BFCM-specific or portfolio-wide averages.
“The next dollar should go where it can still create profitable growth.”Miguel Guerrero, Founder & CEO, Otis AI
A Better ROAS Can Hide a Shrinking Campaign.
A small, well-tuned campaign can post the cleanest average ratio and still be nearly tapped out, while a lower-ROAS campaign has profitable room left to grow. The next dollar isn't judged on the prettiest average. It's judged on four things at once.
Speed only helps when the rules are set. Before BFCM, fix the margin, inventory, CAC, pacing, and escalation limits, and how much data is required before spend goes up or down. Then, when holiday demand moves faster than a scheduled report, Otis can pull back from weakening audiences, creatives, keywords, or channels while leaving profitable segments alone.
How Many, and When.
Creative is where BFCM is won or lost, and the answer isn't “more of the same.” Brands should enter BFCM with 12–16 launch-ready assets built around at least four distinct concepts, plus one held in reserve. A concept has to change the customer problem, proof point, offer, use case, or format. Not just the headline or the background color.
Fatigue should be judged by performance, not age. At modest spend, review creative every 10–14 days. Don't cut an ad just because it's old. Seven days is enough to cut only when an ad has had the delivery to be judged and is still producing weak results. High-ticket or low-conversion accounts need more time. The signal to watch is frequency or CPM rising while CTR, conversion rate, CPA, or marginal ROAS worsens.
A Focused Plan a Lean Team Can Actually Run.
Three Levels of Evidence. The Gap Is Information.
Every operator lives with the “Meta says X, blended says Y” gap. When you don’t crown a winner you can compare platform-reported performance against connected first-party customer, commerce, CRM, and transaction data. When the numbers disagree, the gap tells the team what to investigate rather than which number to believe. Platform attribution still steers the media. The business question gets settled on the backend.
Attributed to Otis AI (Data Partner).
Heading into BFCM 2026, agree in advance on the backend numbers that will settle performance questions, and feed the highest-quality conversion events back into optimization. A generic purchase event isn't enough for a subscription brand. Separate one-time orders from subscriptions and account for downstream retention and LTV. Where direct attribution falls short, lean on holdouts, geo tests, coupon codes, reservation events, or POS.
What Brands Get Wrong.
Five mistakes show up again and again across Otis's customer base.
| The mistake | The correction | |
|---|---|---|
| Prettiest ROAS | Chasing the best-looking ratio, which doesn’t show where you can scale. | Judge marginal growth and business economics before moving more budget. |
| Budget swings | Changing budgets too aggressively, or not at all. | Raise peak budgets in controlled steps, then pull back when post-peak demand falls. |
| Cosmetic “tests” | Calling ten versions of one idea ten new creatives. | Change the concept. In one holiday account, UGC video became the strongest format while older static and carousel lagged. |
| Age, not evidence | Cutting an ad before it has enough delivery, or keeping one that’s had its time and still won’t convert. | Give each ad enough data. High-ticket accounts need more patience than high-volume ones. |
| Split signal | Multiple managers on the same pixel and audiences, inconsistent conversion definitions, or changing offer, budget, channel, and landing page at once. | One pixel, consistent definitions, and one major change at a time so the results stay trustworthy. |
Five avoidable mistakes, and what to do instead. Attributed to Otis AI (Data Partner).
“BFCM performance changes throughout the event. Auction costs, customer intent, and inventory conditions all move. Give each test enough data, avoid changing every major variable at once, and move budget only while the next dollar remains profitable.”Miguel Guerrero, Founder & CEO, Otis AI
This chapter’s frameworks and data were supplied by Otis AI.