Video vs Image Ads on Meta: CTR, CPM, CPA and ROAS Benchmarks from 74 DTC Brands (2026)
Video gets the clicks and image gets the purchases. Across the Meta ad accounts of 74 DTC brands in the 12 months to 28 October 2026, median link CTR was 1.98% for video against 1.38% for image (57 and 54 brands, 8 January to 7 September 2026), median CPA $63.15 against $58.66, and median first-purchase ROAS 1.45 against 1.64 (74 and 71 brands). Video took 53% of spend, image 23%. Source: MHI Meta Benchmarks, DTC, October 2026 edition, read from MHI Media's warehouse of client Meta ad accounts on 29 October 2026, data settled to 28 October 2026.
Every figure on this page is a median across the Meta ad accounts of 74 DTC brands, read from MHI Media's warehouse of client ad accounts on 29 October 2026 with data settled to 28 October 2026. Facebook and Instagram only. The two windows are the 90 days from 31 July to 28 October 2026 and the 12 months from 29 October 2025 to 28 October 2026. A brand is in when it spent at least £3,000 and recorded 30 purchases at ad level in the window. A cell is published only with 8 or more brands, and it also needs £1,000 of spend and 10 purchases for CPA and ROAS, or 50,000 impressions and 200 link clicks for CTR, CPM, CPC and conversion rate. Each brand counts once; the median is the middle brand and the interquartile range (IQR) in brackets runs from the 25th to the 75th percentile. USD at £1 = $1.3289 (multiply by 0.7525 for GBP). The attribution setting, the exchange rate and the rest of the method are in the methodology on the benchmarks hub.
The metrics do not all cover the same dates. CPA, ROAS, spend and purchases run to 28 October 2026. CTR, CPM, CPC and conversion rate exist only for 8 January to 7 September 2026, so the 90-day delivery rows below cover 31 July to 7 September 2026 and the 12-month rows 8 January to 7 September 2026. Frequency exists only to 3 August 2026 and placement only for 8 June to 13 July 2026; neither is split by format on this page. ROAS is Meta-attributed first-purchase revenue divided by spend. The dataset does not measure blended or lifetime revenue, profit, repeat orders or any channel other than Meta, and it cannot say which ad inside a format did the work.
How format is assigned, and why 22% of spend has no format
Format comes from Meta's creative metadata first and the ad name second. A video id or a video object type makes an ad Video; an image object makes it Image; an asset-feed format marks it Flexible (Meta's automatic format); carousel child attachments mark it Carousel. Where the metadata is missing, a token in the ad name decides: video, UGC, static, image or carousel. Anything left is Unknown.
The warehouse holds creative metadata only for ads in the archive, and most ads created after mid-July 2026 have none, so they carry no format unless their names say it. That is why Unknown is 22% of 12-month spend and 34% of 90-day spend across all brands. The hub's beauty-only Table C1 shows the same effect at 28% and 48%, and I link it from the benchmarks hub rather than repeat it here.
Unknown means the format could not be read. Its rows stay in Table H1 so the shares of spend add up. Its CPA is the highest in both windows, and the plain reason is that these are the newest ads, many still in testing. I would not read a format effect into it.
Video vs image vs flexible vs carousel: CPA and ROAS across 74 DTC brands
Image ads had the lower median CPA and the higher median ROAS in both windows: $58.66 and 1.64 against video's $63.15 and 1.45 over the 12 months to 28 October 2026 (71 and 74 brands), and $64.73 and 1.61 against $68.58 and 1.51 over the 90 days to the same date (70 and 72 brands).
| Format | Window | Share of spend | Brands | CPA: median (IQR) | ROAS: median (IQR) |
|---|---|---|---|---|---|
| Video | 12 months | 53% | 74 | $63.15 ($41.19 to $115.90) | 1.45 (0.87 to 1.93) |
| Video | 90 days | 46% | 72 | $68.58 ($42.29 to $106.88) | 1.51 (0.85 to 1.93) |
| Image | 12 months | 23% | 71 | $58.66 ($42.42 to $101.26) | 1.64 (0.94 to 2.22) |
| Image | 90 days | 19% | 70 | $64.73 ($41.90 to $97.32) | 1.61 (0.92 to 2.13) |
| Flexible (Meta automatic format) | 12 months | 2% | 35 | $50.08 ($38.04 to $91.69) | 1.81 (1.14 to 2.29) |
| Flexible (Meta automatic format) | 90 days | 2% | 27 | $63.75 ($36.40 to $87.28) | 1.80 (1.24 to 2.32) |
| Carousel | 12 months | under 0.5% | 12 | $52.80 ($35.81 to $78.69) | 1.59 (1.36 to 2.26) |
| Carousel | 90 days | under 0.5% | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
| Unknown (no format metadata) | 12 months | 22% | 59 | $64.40 ($46.70 to $108.11) | 1.50 (0.91 to 1.86) |
| Unknown (no format metadata) | 90 days | 34% | 55 | $74.87 ($51.29 to $111.46) | 1.46 (0.79 to 1.77) |
Source: MHI Meta Benchmarks, DTC, October 2026 edition (benchmark cuts, 29 October 2026).
The gap is small and the interquartile ranges overlap almost completely. Spend-weighted over 12 months, image still leads on ROAS (1.59 against 1.45) and video edges CPA ($60.80 against $62.03), which tells me the biggest spenders buy video more efficiently than the median brand does. Half the brands sit inside a video CPA band of $41 to $116, so where your own account falls matters more than the medians. This reproduces the beauty-only finding in the hub on 70-plus brands.
CTR, CPM, CPC and conversion rate by format
Video had the higher median link CTR and the higher CPM: 1.98% and $27.41 against image's 1.38% and $19.42 over 8 January to 7 September 2026 (57 and 54 brands), which left median CPC almost level at $1.48 against $1.52.
| Format | Window | Brands | CTR (link): median (IQR) | CPM: median (IQR) | CPC (link): median (IQR) | Conversion rate: median (IQR) |
|---|---|---|---|---|---|---|
| Video | 12 months | 57 | 1.98% (1.34% to 2.68%) | $27.41 ($17.65 to $47.44) | $1.48 ($0.97 to $2.17) | 2.48% (1.53% to 3.51%) |
| Video | 90 days | 54 | 1.89% (1.37% to 2.76%) | $36.56 ($18.16 to $51.26) | $1.64 ($1.15 to $2.24) | 2.74% (1.63% to 3.74%) |
| Image | 12 months | 54 | 1.38% (0.97% to 1.71%) | $19.42 ($12.98 to $29.82) | $1.52 ($1.08 to $2.21) | 2.71% (1.68% to 3.84%) |
| Image | 90 days | 54 | 1.36% (1.07% to 1.89%) | $21.88 ($13.07 to $33.74) | $1.73 ($1.06 to $2.28) | 2.44% (1.56% to 3.80%) |
| Flexible (Meta automatic format) | 12 months | 36 | 1.44% (0.97% to 1.92%) | $22.75 ($12.94 to $40.72) | $1.54 ($1.15 to $2.39) | 2.95% (2.04% to 4.05%) |
| Flexible (Meta automatic format) | 90 days | 27 | 1.30% (1.05% to 1.60%) | $20.13 ($14.15 to $39.15) | $1.57 ($1.32 to $2.69) | 3.51% (2.13% to 4.53%) |
| Carousel | 12 months | 9 | 1.44% (1.10% to 2.37%) | $29.94 ($12.41 to $39.68) | $1.54 ($0.92 to $1.87) | 2.13% (1.18% to 2.85%) |
| Carousel | 90 days | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
| Unknown (no format metadata) | 12 months | 51 | 1.82% (1.29% to 2.51%) | $26.19 ($15.30 to $45.09) | $1.44 ($1.08 to $2.06) | 2.22% (1.44% to 2.98%) |
| Unknown (no format metadata) | 90 days | 47 | 1.86% (1.37% to 2.44%) | $30.12 ($16.02 to $48.31) | $1.57 ($1.19 to $2.22) | 2.24% (1.64% to 3.39%) |
Source: MHI Meta Benchmarks, DTC, October 2026 edition (benchmark cuts, 29 October 2026).
Video's extra clicks cost more to buy and convert slightly less often: 2.48% of video link clicks became a purchase against 2.71% for image over 12 months. That difference, with the higher CPM, accounts for the CPA gap. The 90-day video CPM of $36.56 covers only 31 July to 7 September 2026, so read it as a late-summer figure. The funnel cut measures the click gap a different way: on ad-days with at least one 3-second video view, the median brand's link CTR was 2.04% against 1.40% on non-video ad-days (60 and 57 brands, 8 January to 7 September 2026), and the median hook rate was 30.5% (IQR 23.2% to 35.4%, 60 brands). Those rows and the rest of the on-platform funnel are in the funnel benchmarks.
Video vs image by vertical: beauty, fashion, health and wellness, supplements
Image beat video on median CPA in beauty and skincare ($46.63 against $61.21, 9 brands), fashion ($60.72 against $70.46, 14 brands) and health and wellness ($52.52 against $60.94, 14 and 15 brands) over the 12 months to 28 October 2026; in supplements the order reversed ($80.35 against $74.10, 9 brands) with ROAS level at 0.83.
| Vertical | Format | Share of vertical spend | Brands | CPA: median (IQR) | ROAS: median (IQR) | Brands (delivery) | CTR (link): median (IQR) | CPM: median (IQR) |
|---|---|---|---|---|---|---|---|---|
| Beauty and skincare | Video | 54% | 9 | $61.21 ($47.37 to $62.73) | 0.98 (0.80 to 1.22) | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
| Beauty and skincare | Image | 16% | 9 | $46.63 ($43.35 to $50.08) | 1.14 (0.90 to 1.33) | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
| Beauty and skincare | Unknown | 30% | 8 | $57.64 ($46.76 to $70.79) | 0.85 (0.72 to 1.03) | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
| Fashion | Video | 65% | 14 | $70.46 ($53.21 to $92.57) | 2.13 (1.41 to 2.42) | 12 | 1.74% (1.34% to 2.07%) | $34.84 ($14.33 to $49.71) |
| Fashion | Image | 30% | 14 | $60.72 ($47.70 to $91.13) | 2.38 (1.68 to 2.62) | 12 | 1.40% (1.13% to 1.59%) | $20.94 ($11.54 to $36.12) |
| Fashion | Flexible | 3% | 11 | $78.16 ($49.13 to $102.29) | 2.37 (1.88 to 2.63) | 11 | 1.44% (1.32% to 1.70%) | $17.07 ($13.33 to $46.62) |
| Fashion | Unknown | 2% | 11 | $72.80 ($55.04 to $111.07) | 2.02 (1.54 to 2.89) | 10 | 1.62% (1.45% to 1.78%) | $20.42 ($13.14 to $32.92) |
| Health and wellness (other) | Video | 38% | 15 | $60.94 ($42.25 to $123.07) | 0.84 (0.64 to 1.53) | 10 | 2.06% (1.56% to 2.86%) | $27.39 ($19.08 to $29.83) |
| Health and wellness (other) | Image | 30% | 14 | $52.52 ($38.71 to $85.00) | 1.37 (0.83 to 1.79) | 9 | 1.08% (0.97% to 1.88%) | $18.99 ($16.84 to $21.57) |
| Health and wellness (other) | Unknown | 30% | 10 | $80.50 ($58.48 to $151.46) | 0.91 (0.65 to 1.46) | 8 | 1.52% (1.02% to 2.24%) | $21.45 ($15.71 to $31.95) |
| Supplements | Video | 33% | 9 | $74.10 ($68.23 to $121.09) | 0.83 (0.65 to 0.91) | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
| Supplements | Image | 30% | 9 | $80.35 ($69.08 to $105.60) | 0.83 (0.70 to 0.91) | under 8 | not reported (fewer than 8 brands) | not reported (fewer than 8 brands) |
Source: MHI Meta Benchmarks, DTC, October 2026 edition (benchmark cuts, 29 October 2026).
Beauty here is the same nine brands as the hub's Table C1, read a week later, and the pattern holds. Fashion is the clearest case: video takes 65% of spend, yet image converts at a lower CPA and a higher ROAS on a median CPM of $20.94 against $34.84. Health and wellness has the widest spread, with a video CPA IQR from $42 to $123, so its medians are soft. Supplements is the exception, and both formats sit under 1.0 first-order ROAS, which only works with repeat orders behind it. Beauty and supplements delivery cells are under 8 brands, so there is no CTR or CPM by format for them.
Flexible format and carousel: small samples, read with care
Flexible format (Meta's automatic format) had the lowest median CPA of any format over the 12 months to 28 October 2026, $50.08 (IQR $38.04 to $91.69), with a ROAS of 1.81 (1.14 to 2.29), on 35 brands and 2% of spend.
Over the 90 days to the same date it was $63.75 and 1.80 on 27 brands, so the CPA edge shrank to about level with image while the ROAS held. Flexible format lets you load several images and videos into one ad and leaves Meta to choose which to show each person, so part of its result is Meta picking the cheaper asset per placement and part is the kind of account that switches it on. Thirty-five brands at 2% of spend is a test result. I would not move budget on it. I would put one flexible ad in the next test round and read the asset-level breakdown afterwards.
Carousel is under 0.5% of spend in these accounts in both windows. Its 12-month CPA ($52.80, IQR $35.81 to $78.69) and ROAS (1.59, 1.36 to 2.26) rest on 12 brands, and the 90-day cells are not reported (fewer than 8 brands). That tells you these brands have mostly stopped running carousels; it says little about whether they should. The carousel vs single image post covers when a carousel still earns its slot.
What this means for your account
Run both formats and let the account tell you the split. The 74 brands here put 53% of 12-month spend into video and 23% into image, and image ended with the lower median CPA, so these tables do not require a video-heavy split.
- If your video CPM sits above the 75th percentile ($47.44, 8 January to 7 September 2026), test the same offer as a static and judge the pair on CPA.
- If your image conversion rate is under the 25th percentile (1.68%, same window), the cheap impressions reach the page and the page or offer loses them. Fix that before buying more video.
- If you have no image ads live, start there: median image CPM was $19.42 against $27.41 for video. In the Icy Bear Dental case study the first direct-response statics tripled revenue in month one, before founder video came in.
- If a third or more of your recent spend has no readable format in your own reporting, put the format in the ad name. The hub's beauty cut had 48% of 90-day spend unclassified; this page has 34%.
- Flexible format is worth one slot in each test round. Carousel is worth one only where the product has several variants or a sequence to show.
Our static vs video post covers format by campaign type, the ad formats round-up covers length and placement, and native static story ads explains one image format in detail. To have both formats made and tested against each other on your account, see our ad creative production team.
FAQ
Do video ads or image ads perform better on Meta for ecommerce?
It depends on the metric. Across 74 DTC brands over the 12 months to 28 October 2026, video had the higher median link CTR (1.98% against 1.38%, 57 and 54 brands, 8 January to 7 September 2026) while image had the lower median CPA ($58.66 against $63.15) and the higher first-purchase ROAS (1.64 against 1.45, 71 and 74 brands). Image also had the cheaper median CPM, $19.42 against $27.41. Source: MHI Meta Benchmarks, DTC, October 2026 edition.
What is a good CTR for video ads vs image ads on Meta?
The median link CTR was 1.98% for video ads (IQR 1.34% to 2.68%, 57 brands) and 1.38% for image ads (IQR 0.97% to 1.71%, 54 brands) across DTC brands over 8 January to 7 September 2026. A video CTR above 2.68% or an image CTR above 1.71% sits in the top quarter of these accounts. CTR alone does not predict CPA: image converted more of its clicks, 2.71% against 2.48% over the same window.
Do video or image ads work better for beauty brands on Meta?
Image had the lower CPA. Across 9 beauty and skincare brands over the 12 months to 28 October 2026, median CPA was $46.63 for image ads (IQR $43.35 to $50.08) against $61.21 for video ($47.37 to $62.73), with first-purchase ROAS 1.14 against 0.98, while video took 54% of beauty spend and image 16%. CTR, CPM and engagement by format for beauty are not reported (fewer than 8 brands).
Is Meta's flexible ad format better than a single video or image ad?
It looks promising on a small sample. Flexible format had a median CPA of $50.08 (IQR $38.04 to $91.69) and a first-purchase ROAS of 1.81 (1.14 to 2.29) across 35 DTC brands over the 12 months to 28 October 2026, the best of any format, on 2% of spend. Over 90 days the CPA was $63.75 on 27 brands. Give it one slot in every test round and move budget only once your own account confirms it.
How much do DTC brands spend on video vs image ads on Meta?
Across the Meta ad accounts of 74 DTC brands over the 12 months to 28 October 2026, video took 53% of ad spend, image 23%, flexible format 2% and carousel under 0.5%; the remaining 22% went on ads whose format could not be read from Meta's metadata or the ad name. Over the 90 days to 28 October 2026 the shares were 46% video, 19% image, 2% flexible and 34% unknown.
Sources
- MHI Meta Benchmarks, DTC, October 2026 edition, MHI Media; hub post with methodology, 22 October 2026
- MHI Media warehouse of client Meta ad accounts, benchmark cuts edition, read-only extract 29 October 2026, data settled to 28 October 2026