Meta today published the findings of a structured creative research study conducted with Toluna, the global insights company, identifying which specific creative elements drive performance on Instagram Reels across both brand and direct response objectives. The results, shared in Episode 9 of Meta's "You're Better Off Knowing" podcast and accompanied by a LinkedIn post from Torsten Mueller-Klockmann, Marketing Science Manager for the DACH and MEA regions at Meta, offer some of the most granular publicly available data on short-form video creative effectiveness to date.

The study tested 100 advertisements against 65 distinct creative variables, covering four markets - Germany, the United Kingdom, Italy, and Turkey - and two advertising verticals: consumer packaged goods (CPG) for brand objectives and e-commerce for direct response (DR). Each ad was shown to 100 participants through a mobile testing tool that simulated an Instagram Reels feed environment. Viewing behavior was measured passively, and participants were subsequently surveyed on memory and brand metrics including comprehension, likability, and purchase intent. Performance was assessed by identifying which creative variables correlated most strongly with ads ranking in the top 20% for each metric compared to country-specific benchmarks.

Why country-specific benchmarks mattered

One methodological nuance stands out. Rather than comparing all ads against a single aggregate benchmark, Toluna used country-level reference points to control for systematic differences in audience response patterns. According to Bex Lauenstein-Dean, Associate Research Director at Toluna, audiences in Turkey tended to rate ads more positively overall, while German and UK audiences skewed more critical. Applying country-specific benchmarks allowed the analysis to surface cross-market creative principles without those results being distorted by local response-style tendencies.

This is a meaningful technical choice. Cross-market creative studies frequently aggregate data in ways that mask such variation, producing findings that appear universal but are actually driven by one or two dominant markets. The decision to normalize first at the country level, and then identify patterns that held across all four markets, adds credibility to the findings.

The Reels audio shift

Among the clearest signals in the study was the role of sound. According to Meta, over 80% of Reels played on Instagram each day are played with sound on. This is a direct reversal of the conditions that characterized social feed advertising for much of the past decade. Dan Moller, Creative Strategist at Meta, noted that during earlier years of feed advertising, the standard guidance was to design for a sound-off environment, ensuring ads were comprehensible without audio. That principle no longer applies to Reels.

The study found that delivering a message through both audio and visual cues increases the likelihood of an ad ranking in the top 20% for brand interest by 1.8 times. For brand advertisers using speech and music together, the multiplier rises to 2.0 times for brand interest. The finding aligns with Reels' structural difference from feed: users arriving in a full-screen, immersive format are far more likely to have sound engaged, creating a genuine opportunity for audio-led storytelling that was simply not available in earlier feed environments.

Brand and message placement: the five-second window

The study found strong evidence that early brand and message introduction is a significant performance driver. For brand advertisers, showing both the brand and the main message within the first five seconds of a Reels ad makes it 1.7 times more likely to rank in the top 20% for purchase intent.

This finding challenges a persistent creative instinct: that branding early in an ad risks turning off viewers before they are sufficiently engaged. According to Moller, the playthrough curve for any Reel shows declining viewership over time as users retain the option to swipe past. If both brand and message do not appear early, a significant share of viewers who scroll away will have received neither. The data suggests that when content is genuinely engaging and includes a strong hook, early brand identification does not suppress viewing - it coexists with it.

A separate but related finding involves dynamic branding: creatives that display the brand more than once, whether in different scenes or different screen positions, are 1.8 times more likely to rank in the top 20% for purchase intent among brand advertisers. The principle is not just "show the logo early" but "show the brand repeatedly and differently."

Direct response findings: product, story, and native elements

The DR results follow a distinct logic. Product visibility is the primary driver. Showing the product more than once makes a DR Reels ad 2.7 times more likely to rank in the top 20% for purchase intent. That figure rises further when contextual information is layered in - unique selling points, features, or other supporting details - bringing the multiplier to 5.3 times for purchase intent when the ad tells a fuller story around the product. These are notably large effect sizes for a creative variable study.

Brand presence in DR contexts requires careful calibration. Featuring the brand in the creative, but limiting that presence to no more than 25% of the total ad duration, increases the likelihood of ranking in the top 20% for purchase intent by 4.8 times. Extending brand visibility beyond that threshold, according to the data, starts to reduce impact rather than amplify it. For categories like CPG where product packaging inherently carries heavy branding, this represents a concrete and somewhat counterintuitive guideline.

Native elements - specifically emojis - also produced a notable result. DR ads that incorporate emojis as a visual element are 2.5 times more likely to rank in the top 20% for purchase intent. Emojis function here as an indicator of format-native design: they signal to viewers that the content belongs to the Reels environment rather than being repurposed from another channel. A related finding involves call-to-action elements: including a CTA visually and/or through audio increases the likelihood of ranking in the top 20% for purchase intent by 1.9 times.

Hooks and the first two seconds

Across both brand and DR categories, hooks - elements designed to arrest scrolling immediately, using both visual and audio cues - increase the likelihood of ranking in the top 20% for purchase intent by 1.5 times. The mechanism is straightforward: users of short-form video formats retain full control to scroll past at any moment. An ad that fails to establish relevance or interest within the first one to two seconds will lose the viewer before any subsequent messaging can register.

The podcast discussion noted that previous research into TikTok hooks reached consistent conclusions - the need to capture attention within the first five seconds applies broadly across short-form formats, not specifically to Instagram Reels. Combined with findings from other studies examining Reels, Shorts, and TikTok videos, this creates a consistent cross-platform principle around attention capture.

Slice of life and human presence

The brand findings also include a result around everyday content. Featuring scenes of people in ordinary, relatable situations increases the likelihood of ranking in the top 20% for purchase intent by 1.5 times for brand advertisers. Earlier Meta research had already identified human presence in an ad as a general performance driver; this study refines that into a more specific directional signal - it is not just the presence of a person but the authenticity and relatability of the scenario they occupy.

This reflects a broader characteristic of the Reels environment. According to Martin Ash, Media and Digital Measurement Specialist at Toluna, users of short-form video expect content that feels real, relevant, and consistent with how they use the platform. Ads that succeed often look and feel closer to user-generated content than to polished broadcast production. The format creates its own grammar.

The synthetic persona validation

Perhaps the most technically significant disclosure in the study involved a parallel validation exercise. According to Toluna, after conducting the original research with 100 human respondents per ad, the company retested the same ad set - comprising 75 to 76 ads - using synthetic personas rather than live human panels. The exercise was conducted as a validation of whether AI-driven synthetic respondents could replicate the directional findings produced by real people.

According to Ash, the results were notably aligned. Ads that ranked in the top quintile among human respondents also tended to rank highly among synthetic personas. The same pattern held at the bottom: consistently weak performers showed up as weak in both evaluations. Divergence appeared most frequently in the middle tier of the distribution, which is broadly consistent with what would be expected between two different human samples from the same population.

According to Niels Bovee, AI creative testing expert at Toluna, whose comments appear in a May 5, 2026 LinkedIn article by Ash: "Synthetic personas don't just scan the surface of an ad - they process it the way a real viewer would. They pick up emotional tone and storytelling quality. They recognise brand presence in all its forms, whether that's a logo on screen, a voiceover, or the way a product is woven into a narrative. They respond to pacing, structure, and the overall coherence of the creative. What they're doing, in other words, is reflecting back the patterns of human response that have been built into them - and that's precisely why the learning they generated tends to align with what real audiences told you."

Toluna's synthetic personas are built from its proprietary panel, which the company states includes over 79 million human participants across global markets. The personas draw on anonymized first-party data encompassing demographic, psychographic, lifestyle, and consumption attributes accumulated across more than two decades of panel operations.

The practical implication is significant for how creative testing programs operate at scale. A traditional lift study or consumer panel test requires considerable time and cost for each ad creative evaluated. Synthetic persona testing, if directionally aligned with human results, allows substantially more creative variations to be evaluated at a fraction of both the time and cost. Toluna has since launched a commercial product - ACT Instant AI - that extends this methodology into a self-serve ad testing offering, announced October 28, 2025.

20 dos and 20 don'ts for Reels advertising

The study and podcast discussion together produce a set of concrete directional signals. These are drawn directly from the research findings and expert commentary - not universal rules, but patterns that held across 100 ads, four markets, and two advertising verticals.

Dos

  1. Show both the brand and the main message within the first five seconds. For brand advertisers this raises top-20% purchase intent likelihood 1.7 times.
  2. Display the brand more than once, in different scenes or screen positions throughout the creative. Dynamic branding raises top-20% purchase intent likelihood 1.8 times.
  3. Use both speech and music together. For brand advertisers this increases the likelihood of ranking in the top 20% for brand interest by 2.0 times.
  4. Deliver the message through both audio and visual cues simultaneously. Combined delivery increases top-20% brand interest likelihood 1.8 times.
  5. Feature scenes of people in ordinary, relatable everyday situations. Slice-of-life content raises top-20% purchase intent likelihood 1.5 times for brand advertisers.
  6. Design the creative with sound on as the expected state. Over 80% of Reels on Instagram are played with audio enabled.
  7. Open with a hook that uses both visual and audio elements in the first one to two seconds. Hook adoption raises top-20% purchase intent likelihood 1.5 times across brand and DR.
  8. Show the product more than once in DR creatives. Repeated product visibility raises top-20% purchase intent likelihood 2.7 times.
  9. Add context around the product - unique selling points, features, or additional information beyond the main message. Doing so raises top-20% purchase intent likelihood 5.3 times for DR.
  10. Include a call to action visually and/or through audio in DR creatives. CTA inclusion raises top-20% purchase intent likelihood 1.9 times.
  11. Use emojis as a native visual element in DR creatives. Emoji use raises top-20% purchase intent likelihood 2.5 times.
  12. Keep brand presence in DR creatives to no more than 25% of the total ad duration. Staying within that threshold raises top-20% purchase intent likelihood 4.8 times.
  13. Build the creative in 9:16 vertical format with audio enabled and key content within the safe zones - these are the foundational requirements before any other optimization applies.
  14. Treat the creative as genuinely audio-visual rather than visual-first. Ads that do not engage the audio channel are giving up a significant share of the available attention.
  15. Think about the hook as a combined audio-visual moment, not just a visual one. Both channels contribute to arresting scroll behavior in the first two seconds.
  16. Ensure the brand is identifiable within the creative from audio cues, not only visual ones - voiceover, a brand name spoken early, or music that carries brand association all contribute.
  17. Start a test-and-learn program using these findings as a baseline. According to Meta and Toluna, the data provides directionality rather than guarantees - individual brand testing is the intended next step.
  18. Consider lift studies or A/B tests to validate which specific elements work best for a particular brand, product, and market combination.
  19. Treat findings as applicable beyond Instagram Reels. The podcast discussion noted that studies of YouTube Shorts and TikTok videos have reached broadly similar conclusions on attention capture and branding timing.
  20. Validate creative decisions through testing across different markets separately rather than relying on aggregate cross-market data, given the systematic differences in audience response styles the study documented.

Don'ts

  1. Do not save the brand reveal for the end of the creative. Viewership drops throughout the duration of any Reel, and viewers who scroll past before a late-stage logo reveal receive no brand impression.
  2. Do not show the brand only once. A single logo appearance, particularly a brief one, is less effective than brand identification distributed across the creative.
  3. Do not design for sound off. The behavior pattern that drove sound-off guidance on feed placements does not transfer to Reels, where sound is the default engaged state.
  4. Do not repurpose feed or broadcast creative without adapting it to the Reels format. Ads that feel native to the platform - in pacing, format, and elements like emojis - consistently outperform repurposed content.
  5. Do not flood DR creatives with branding. Exceeding 25% brand presence duration in a DR ad is associated with reduced purchase intent performance rather than higher brand recognition.
  6. Do not show the product only once in DR creatives. A single product appearance substantially underperforms relative to repeated product visibility.
  7. Do not omit context from DR ads. Showing a product without supporting information - features, USPs, or a narrative around it - leaves significant performance on the table.
  8. Do not skip the call to action in DR creatives. A CTA, whether visual, audio, or both, meaningfully raises the likelihood of conversion-related outcomes.
  9. Do not treat the hook as optional. Short-form video viewers have full agency to scroll past at any moment; an ad without an immediate attention-capture mechanism in the first two seconds risks losing the viewer entirely.
  10. Do not assume a strong hook alone is sufficient to hold viewers through a brand-light opening. The data suggests early branding coexists with strong creative rather than conflicting with it.
  11. Do not ignore the audio dimension of the hook. A visually arresting opening that lacks a corresponding audio cue misses the combined effect the data identified.
  12. Do not apply DR creative principles to brand objectives or vice versa. The study found meaningfully different patterns for upper-funnel and lower-funnel creative, and treating them as interchangeable risks underperformance on both.
  13. Do not position key content outside the safe zones. Safe zone violations remain one of the most common technical errors in Reels creative, and content outside the guaranteed visible area may not be seen regardless of aspect ratio or placement.
  14. Do not treat the research findings as a guaranteed recipe. According to Wedekin, they represent a starting point - not every element will perform identically for every campaign, product, or market.
  15. Do not skip cross-market benchmark normalization when evaluating multi-market creative studies. Aggregate comparisons without country-level controls can produce misleading findings, as the Turkish, German, and UK response-style differences in this study illustrate.
  16. Do not assume synthetic persona results are interchangeable with human panel results at the individual ad level. The Toluna validation showed directional alignment and stable extremes, but mid-tier ad rankings diverged between human and synthetic evaluations.
  17. Do not use synthetic testing as a wholesale replacement for incrementality measurement. According to the podcast discussion, synthetic personas and lift studies serve different functions; the former is suited to creative screening at scale, the latter to measuring in-market causal impact.
  18. Do not conflate brand interest and purchase intent as equivalent metrics when evaluating creative decisions. The study found that some elements drove one measure more than the other, and the distinction matters for campaign objectives.
  19. Do not over-brand CPG creatives on the assumption that visible packaging equals effective advertising. The 25% brand duration guideline applies even in categories where logos are embedded in the product itself.
  20. Do not wait until a campaign is live to begin testing creative elements. The study's methodology - controlled pre-market testing of creative variables - is designed to inform decisions before media spend is committed.

Why these findings matter to the market

Reels now account for over 50% of time spent on Instagram and are re-shared 4.5 billion times per day, according to Meta. Video time spent on Instagram increased more than 30% year-over-year in 2025. The format has moved from an experimental placement to one of the central surfaces for advertising investment on the platform. Yet despite this growth, systematic creative research for Reels remained comparatively thin before this study.

The research fills a specific gap. Meta's existing creative guidance - the "creative essentials" framework calling for 9:16 vertical format, audio enabled, and key content positioned within safe zones - established foundational requirements. Ads that met those essentials outperformed both image ads and standard video for both brand and performance objectives. But those requirements are platform-level hygiene, not creative strategy. This study was specifically designed to go deeper: identifying which creative decisions, within an already technically compliant ad, move performance from average to top-quintile.

The study also connects to a broader question that has gained urgency in 2025 and 2026: to what extent can AI tools replicate or augment human judgment in creative evaluation? Meta has been pushing hard toward AI-automated advertising, including the use of generative tools for creative production, automated delivery optimization, and recommendation system improvements that drove a 5% increase in ad conversions for Instagram. The Toluna validation exercise sits in that same current: it tests whether AI can participate meaningfully in the pre-flight evaluation side of creative, not just the delivery side.

For media buyers and creative strategists, the practical takeaways are fairly concrete. Brand campaigns benefit from early and repeated brand identification, multimodal message delivery, and relatable human content. DR campaigns benefit from repeated product visibility, a compelling product narrative, controlled brand presence, native format elements, and explicit calls to action. Both benefit from audio-visual hooks in the first two seconds. None of these findings are absolute guarantees - Sonia Wedekin, Marketing Science Partner at Meta, described them as a starting point for individual brand testing rather than a universal recipe. The research provides directionality; whether a specific element performs optimally for a specific brand in a specific market requires testing against that brand's own baseline.

The research was conducted across Germany, the UK, Italy, and Turkey in 2024, commissioned by Meta. The Meta for Business blog post summarizing the findings was originally published December 2, 2025. Episode 9 of the "You're Better Off Knowing" podcast, which features the full discussion between Mueller-Klockmann, Ash, Lauenstein-Dean, Wedekin, and Moller, launched today on YouTube.

Timeline

Summary

Who: Meta and Toluna, the global insights company, alongside podcast guests Martin Ash and Bex Lauenstein-Dean from Toluna, and Sonia Wedekin and Dan Moller from Meta.

What: A structured creative effectiveness study testing 100 Instagram Reels ads against 65 coded creative variables, measuring brand and direct response performance across metrics including purchase intent, brand interest, and comprehension. A parallel synthetic persona validation study using the same ad set was also conducted, with broadly aligned results. Key quantified findings include: brand and message shown within the first five seconds raises top-20% purchase intent likelihood 1.7 times; dynamic branding raises it 1.8 times; product shown more than once raises DR purchase intent likelihood 2.7 times; emojis as a native element raise it 2.5 times; and brand presence capped at 25% of duration raises it 4.8 times.

When: The Toluna research was conducted in 2024. The Meta for Business blog post summarizing findings was published December 2, 2025. Episode 9 of the "You're Better Off Knowing" podcast launched today, May 18, 2026, on YouTube, with the LinkedIn announcement published the same day. The Toluna synthetic persona validation article by Martin Ash was published May 5, 2026.

Where: The study covered four markets - Germany, the United Kingdom, Italy, and Turkey - spanning CPG (brand) and e-commerce (direct response) verticals. Testing used Toluna's mobile creative testing tool simulating an Instagram Reels feed environment.

Why: Despite Reels accounting for over 50% of time spent on Instagram and being re-shared 4.5 billion times daily, systematic data on which specific creative elements drive performance in the format remained limited before this research. The study aimed to go beyond foundational technical guidance - vertical format, sound-on design, safe zone compliance - and identify granular creative decisions that move brand and DR performance from average to top-quintile. The synthetic persona validation tested whether AI-driven respondents could replicate the directional findings of human research, with implications for the speed and scale at which creative testing programs can operate.

Share this article
The link has been copied!