Specs and format: it's not short-form
The 2026 Facebook Feed video spec: 4:5 vertical (1080×1350) or 1:1 square (1080×1080) are the working aspect ratios; 4:5 outperforms square on mobile because it occupies more vertical real estate and consistently delivers higher CTR in placement testing. Length runs from 1 second to 240 minutes nominally; the practical sweet spot is 15–30 seconds for cold traffic, with watch-through collapsing past 30. In-stream video (mid-roll ads inside publisher / creator videos) tightens that to 5–15 seconds because the viewer reads the slot as an interruption.
The hard format break with Reels: Reels are 9:16 full-screen, immersive, sound-on by default. Feed video sits inside a scrollable Feed, plays sound-off by default, and shares the screen with text, comments, share controls, and the rest of the Facebook UI. The viewer is in a different posture — scrolling, glancing, looking for context — not committed to immersive watching. A Reels-cut 9:16 dropped into Feed loses the bottom and top to UI chrome and reads as a misfit.
As of March 2026, Meta consolidated Facebook Stories, Facebook Reels, Instagram Stories, and Instagram Reels into a single unified 9:16 safe zone — one vertical asset works across all four placements. Feed is explicitly not in that consolidation. 4:5 is its own creative spec, and the operator who treats it as "just a cropped Reel" leaves performance on the table.
Audience: older, more considered, higher intent
Pew's 2025 generational survey has Boomers at 88% Facebook adoption (with 57% of adults 65+ concentrating their social activity on Facebook) and Gen X at 84%. Feed is where that audience actually lives — they didn't adopt Reels at the same rate, and the in-feed video unit is the format they interact with most. The downstream implication: Feed video targets a demographically and behaviorally different cohort than Reels, even inside the same Advantage+ campaign.
The intent quality on Feed runs higher per impression for two reasons. First, Feed CPMs are higher than Reels CPMs (Meta's own placement breakouts and third-party 2026 benchmarks both confirm this), so the auction is selecting for higher-LTV buyers. Second, the Feed viewer is reading caption text, hovering on the dwell, and engaging deliberately — Facebook explicitly uses dwell time as a ranking and retargeting signal on Feed, treating hover-time as proof of interest even when no click fires.
For products with longer consideration cycles — DTC home, B2B services, financial products, considered DTC at AOV $200+ — Feed video often out-converts Reels at lower volume but higher margin per conversion. 2026 industry-wide Facebook ad conversion benchmarks land around 9% on average, with strong accounts above 10%; Feed-first campaigns disproportionately populate the higher tier.
Feed dwell vs. short-form watch-completion
The signal Facebook's Feed ranker rewards is dwell time: how long the viewer hovered over the post inside the Feed, whether or not they completed the video. That's a fundamentally different signal from short-form watch-completion. On Reels and TikTok, a 95% completion rate on a 12-second hook- demo is the win condition. On Feed, a 4-second dwell on a 60-second video with engaged scroll-stop counts as success even if the viewer never watched to the end.
That changes hook design. Short-form's 3-second hook mandate (covered in sound-off and the cold-start work) still applies — if the first frame doesn't register, dwell goes to zero. But the second-half of a Feed video doesn't need to retain like a Reels close does. The rubric on Feed weights opening-attention and mid-roll dwell-extenders (a second hook, a payoff reveal, a comparison mid-frame) more than seamless looping or repeat-view design.
The Ad Bench currently scores Feed video with the short-form rubric plus Facebook-Feed calibration. The pacing score weights tighter on Feed cuts than on TikTok cuts (Feed viewers tolerate a slower opening because the format permits it), and the CTA architecture weights closer-frame CTAs more on Feed than on Reels because the loop seam doesn't carry the conversion path. The save-bait copy patterns that score on Reels — see the save economy — underperform on Feed where comments and the explicit click carry more conversion weight.
Advantage+ Shopping with Feed-first creative
Inside an Advantage+ Shopping campaign, Meta's placement optimizer picks where the asset lands. Agency-reported 2026 breakouts have Feed taking 60–70% of efficient spend, Reels 20–30%, Stories 10–20%. That distribution shifts by vertical — higher-AOV considered purchases skew further toward Feed; impulse / Gen-Z DTC skews toward Reels — but Feed is the heavier slice for most catalog-driven accounts.
The Andromeda-era ranker rewards creative diversity at the campaign level. Public guidance from 2025–2026 puts the working target at 8–12 conceptually distinct concepts per campaign with 2–3 variations each, and pushes brands toward 10–15 distinct assets to feed the retrieval engine. Feed-first creative is one of the most valuable variety axes inside that asset set, because a Feed-tuned 4:5 with sound-off design and a Reels-tuned 9:16 with sound-on, music-led design read as structurally distinct to the diversity scorer — they don't collapse into the same Entity ID and don't compete with each other for retrieval.
The operator move: ship at least one Feed-first 4:5 cut into every Advantage+ Shopping campaign, even when the rest of the set is Reels-first. The placement optimizer will find the spend the variant unlocks, and the diversity score keeps both versions in active rotation. See platform rules for the spec-and-policy floor each variant needs to clear before it ships.
When to ship Feed-first vs. Reels-first
Three signals point to Feed-first. One: target audience skews 45+ — the demographic concentration is on Facebook Feed and the format is what they actually use. Two: AOV is high and the consideration cycle is longer than 24 hours — Feed dwell time and considered scroll behavior over-index for this buyer. Three: the creative leans on text-heavy comparison, multi-product showcase, or before-after with explicit claims — Feed's larger frame and longer permitted runtime carry these formats better than 9:16 does.
Three signals point to Reels-first. One: target audience is 18–34 — Reels' reach and engagement skew younger and Instagram pulls disproportionate share of that cohort. Two: impulse-driven, low-AOV, single-SKU offers where the conversion path is hook-demo-CTA inside 15 seconds. Three: the creative depends on trending audio, a creator's native voice, or a Reels-format reference that won't carry into Feed's sound-off-by-default environment.
Where The Ad Bench's rubric currently has a limit worth naming: the short-form-grounded scoring catches hook design, sound-off comprehension, pacing, native-feel, and CTA architecture on Feed video well, but it under-weights the dwell-time signal and over-weights loop design. A Feed video that scores 70 on the rubric and converts at 12% in-market is doing something the rubric isn't fully reading — the dwell mechanic. The sibling article Facebook Reels vs. Instagram Reels covers the short-form half of the Facebook surface, where the rubric's calibration is on firmer ground.