How DTC Marketing Spend Changed 2020-2026: iOS 14, Attribution Loss, Recovery

Public DTC marketing spend as a percentage of revenue moved from a 16.5% pooled median in 2021 to 10.6% in 2023 and back to 13.3% in 2025. Three eras drove the arc: pre-iOS14 efficiency, attribution-shock divergence, post-2023 rationalization. The trend line tells one story. The variance under the trend line tells the actual story.

We pulled marketing spend as a percentage of revenue from SEC EDGAR 10-K filings for 11 public DTC brands across eyewear, beauty, pet, apparel, food and beverage, and outdoor — Warby Parker, e.l.f. Beauty, Bark Inc., Revolve, FIGS, Beauty Health, Yeti, Honest Co, Beyond Meat, Lululemon, and Celsius Holdings. The pre-iOS14 paid-media model was a gift: acquire on Meta at attribution-validated ROAS, scale until contribution margin breaks, repeat. Then April 2021 happened. Apple's AppTrackingTransparency framework rolled out in iOS 14.5, opt-in collapsed below 25%, and last-click on Meta — the foundation the entire DTC paid-media stack was built on — broke.

The DTC marketing spend trend from 2020 to 2026 isn't a clean V-shape, and it isn't a clean U. It's an asymmetric compression: the median fell, but the spread doubled. The brands that found the iOS14 era hardest were the ones who tried to spend their way through it. The brands that won pulled back, rebuilt the measurement stack, and reallocated to channels and motions that didn't depend on Meta's last-click pixel firing in real time.

Pooled median by year: the headline trend

This is the pooled view across all 11 brands by fiscal year. Sample size (n) varies because brands IPO'd at different times — FIGS in 2021, Warby Parker in 2021, Bark in 2021, Honest Co in 2021. Lululemon and Yeti go back further. Celsius pre-IPO data goes back to 2018. The pooled median is calculated across whatever brands had reported 10-K data for each year.

Fiscal year	Brands (n)	Median S&M % revenue	Mean S&M % revenue	P25	P75
2018	1	40.3%	40.3%	40.3%	40.3%
2019	2	29.1%	29.1%	28.1%	28.1%
2020	2	20.7%	20.7%	14.8%	14.8%
2021	8	16.5%	20.4%	14.0%	23.8%
2022	10	14.7%	20.6%	7.0%	16.5%
2023	9	10.6%	13.2%	7.7%	16.1%
2024	9	12.4%	13.6%	7.7%	14.8%
2025	8	13.3%	14.2%	5.1%	14.3%
2026	2	6.7%	6.7%	5.6%	5.6%

Three things to notice. First, the gap between median and mean — in 2021 and 2022 the mean is 4-6 points above the median, which means a few brands were spending way above the pack and pulling the average up. That's the iOS14 escalation pattern in the data. Second, the P75 in 2021 is 23.8%, but the P75 in 2023 is 16.1% — the upper-quartile brands compressed harder than the median did. Third, 2024-2025 saw a modest rebound in the median to 12-13%, then 2026 pulled lower again, but the 2026 sample is only 2 brands so far — Yeti and Lululemon, both mature, both below 8%. Don't read too much into the 2026 line yet.

What did the pre-iOS14 era (2020-2021) actually look like?

The 2020-2021 numbers look low because of what they don't include. Most of the brands that became public DTC stories — FIGS, Warby Parker, Honest Co, Bark, e.l.f.'s post-relaunch growth — were either pre-IPO or in their first 10-K when the data starts. The brands that did exist publicly and were aggressive on paid social — e.l.f. at 31.2% in 2021, Celsius at 23.8% in 2021, Beauty Health at 42.9% in 2021 — drove the upper-quartile lines.

What the pre-iOS14 era really looked like, channel-side: brands could acquire on Meta at attribution-validated ROAS in the 2.0-3.5x range, scale paid social as a percentage of revenue with a clear payback window, and treat the resulting marketing-spend-to-revenue ratio almost like a dial. Push spend up, revenue followed in a measurable way. The dial worked. Most operators believed it would keep working.

The pre-iOS14 stack was an arbitrage. Meta's pixel firing reliably plus a saturated organic base meant the cheapest customer acquisition the DTC era will ever see. A lot of operators thought that was the new normal. It was always temporary — we just didn't know we were living through the cheap window until it ended.

Across 2020 and 2021, the pooled median sits at 16.5-20.7%. That's the baseline you should compare any iOS14 disruption argument to — not a hypothetical pre-iOS14 number that includes brands like FIGS and Warby Parker who hadn't filed yet. The honest pre-iOS14 number is "low to mid teens for mature, twenty-plus for scaling."

Why didn't iOS14 attribution loss push the median higher?

This is where most DTC narratives get the data wrong. The conventional take is "iOS14 broke attribution, so DTC marketing spend went up." It didn't, in pooled median terms. The 2021 median was 16.5%, the 2022 median was 14.7%, the 2023 median was 10.6%. The pooled spend percentage fell through the entire iOS14 disruption window.

What did go up: the mean, in 2022. Pulled up by brands like Beauty Health (43.8% of revenue in 2022) and Celsius (54.0% in 2022), both of which escalated paid spend hard trying to compensate for lost attribution signal and accelerate growth. Beauty Health peaked at 43.8% before compressing back to 31.1% by 2025. Celsius's 53.97% in 2022 was the apex of escalation — when last-click broke, the temptation was to keep pushing dollars in and hope contribution margin tolerated it. For a year or two, it did. Then it didn't.

The brands that compressed hardest in 2022-2023:

• e.l.f. Beauty: 31.2% (2021) → 33.4% (2022) → 16.7% (2023). A 17-point compression in two years.
• Honest Co: 17.0% (2021) → 15.2% (2022) → 10.6% (2023). 7 points down. Disciplined.
• Bark: 15.9% (2021) → 12.6% (2022) → 10.4% (2023). Nearly halved.
• Beyond Meat: 2.6% (2021) → 4.9% (2022) → 5.0% (2023). Already low; held.
• Warby Parker: 14.1% (2022) → 11.7% (2023). Small but real compression.

Why the median fell when attribution got harder: the brands that survived the iOS14 era are the brands that cut. The escalators (Beauty Health, Celsius in its peak year) either compressed later or, in Celsius's case, restructured so aggressively that they fell out of the public DTC sample. The dataset of survivors is biased toward operators who pulled back. That's the actual lesson of the iOS14 era — efficiency wins, escalation buys you time you usually can't afford.

What killed brands in 2021-2023 wasn't attribution loss. It was spending against attribution loss. If your CAC went up 30-40% and you tried to volume your way through it, you ate two years of contribution margin trying to hit a top-line number that was structurally harder to reach. The brands that pulled back, rebuilt measurement, and lived with slower growth came out the other side intact.

The channel mix shifted underneath the headline number. Meta's share of US DTC ad spend fell from 34.9% in Q1 2021 to 27.0% by Q1 2022 as performance worsened and brands diversified into Google, TikTok, offline media (direct mail, podcasting), and Pinterest. Multi-touch attribution adoption climbed from 31% in 2023 to 47% by April 2026. Marketing mix modeling went from 9% adoption to 26%. The measurement stack rebuilt itself in real time, and the brands that invested in it — instead of spending more on Meta — compressed their reported marketing-to-revenue ratios while keeping unit economics intact.

What changed structurally in the 2024+ rationalization?

By 2024 the pooled median had stabilized around 12-13%. That's not the bottom — that's the new equilibrium for public DTC. What changed structurally:

Retention took share of revenue

Mature DTC brands moved roughly 60% of revenue to repeat customers by 2024-2025. When repeat revenue grows faster than acquired revenue, the marketing-spend-to-revenue ratio mechanically falls — the denominator grew from a source that didn't require new acquisition spend. The compression in the median is partly a story about denominators, not about brands cutting paid media in absolute dollars. Yeti at 7.8% in 2024-2026, Lululemon at 4-6% across the same window — those numbers look small because retail and repeat are doing the heavy lifting.

Brand and top-of-funnel started taking share of marketing budget

Pre-iOS14, the playbook was direct response. Run paid social, measure last-click, optimize ROAS. Post-2023, sophisticated DTC operators started moving 5-15% of their marketing budget — sometimes more — into top-of-funnel and brand. Video views, podcast sponsorships, OOH, content, awareness creative. None of this clears in last-click attribution. All of it grows revenue with a lag. The marketing-to-revenue ratio reported in the 10-K still includes the spend; the revenue from the spend shows up in 90-day windows or longer.

Top of funnel is where the arbitrage moved. No one trusts it because no one can measure it cleanly — so the few operators willing to invest there are competing in a market with very few buyers. The brands that built brand equity through 2023-2025 are the ones who'll have the cheapest performance media in 2026-2027, because their Google search volume is up, their direct traffic is up, and their CPMs on retargeting are softer.

AI-driven creative production compressed creative cost per variant

By 2025, roughly 86% of DTC brands surveyed had adopted AI tools for creative generation — short-form video variations for TikTok and Reels, ad copy variants for Meta, image variations for static creative. The downstream effect on the financials is subtle but real. The unit cost of producing a Meta-ready ad variant fell roughly 60-70% from 2021 to 2025 as AI tools matured. Brands that were spending 15-20% of marketing budget on creative production agencies in 2021 brought that down to 8-12% by 2025, freeing dollars for paid media or retention investment without raising the headline marketing-to-revenue line.

The framework shifted from ROAS to MER to contribution margin

Last-click ROAS lost credibility in 2021-2022. Marketing efficiency ratio (MER) — total marketing spend divided by total revenue — became the dominant high-level metric in 2023. By 2024-2025, the most disciplined operators had moved further: MER as a directional check, contribution margin as the actual operating metric. "What's our CM2 after marketing?" became the question that drove budget decisions. The reported marketing-to-revenue ratio is the output of that calculation, not the input.

Per-vertical evolution: apparel held high, CPG food went leaner

The headline median masks a real category split. Here's the same dataset cut by vertical, looking at the most recent fiscal year and the trajectory.

Vertical	Brands	2021 range	Latest range	Direction
Apparel DTC	Revolve, FIGS, Lululemon	4.8% - 15.8%	4.5% - 14.3%	Held flat; mature compresses
Beauty CPG	e.l.f., Beauty Health	31.2% - 42.9%	21.4% - 31.1%	Compressed 10-15 points
Eyewear DTC	Warby Parker	14.1% (2022)	12.6% (2025)	Modest compression
Pet DTC	Bark	15.9%	12.8% (2025)	Compressed 3 points
Personal care DTC	Honest Co	17.0%	13.8% (2025)	Compressed; rebounded slightly
Outdoor DTC	Yeti	n/a (later filer)	7.8% (2026)	Mature, low intensity
Food/Beverage CPG	Beyond Meat, Celsius	2.6% - 23.8%	2.2% (BYND latest)	Compressed hard; retail-led

Apparel held high spend. Revolve at 14-16% across the entire 2021-2025 window, FIGS at 14-30% early on, Lululemon as the floor at 4-5% (pulled lower by retail and wholesale revenue). Apparel can't lean on consumable repeat the way beauty and food can — replacement cycles are 6-18 months for apparel versus 30-60 days for consumables, which forces structurally higher marketing intensity to feed the funnel.

Beauty CPG compressed but stayed above 20%. e.l.f. went from the 31-33% peak in 2021-2022 down to 16.7% in 2023, then back up to 21.4% by 2025 — a controlled rebuild from a more efficient base. Beauty Health came down 12 points from peak. The category supports higher marketing intensity because of margin (65-72% gross) and consumable repeat (90% of revenue from repeat for mature brands).

Food and beverage CPG went leanest. Beyond Meat at 2.2% in 2025 — that's a wholesale-and-retail revenue base with very thin DTC paid media on top. Celsius peaked at 54% in 2022 then consolidated into a retail-distribution model. Food and beverage at scale is a shelf-velocity story, not a DTC paid-media story — the marketing-to-revenue ratio reflects that.

What private DTC should learn (and what to ignore)

Most private DTC brands look at the public median and reach the wrong conclusion. They see "13% of revenue" and think they should be running 13% of revenue. They shouldn't. Here's the framework I run with my clients in our fractional CFO engagements.

Stage and gross margin set your marketing-to-revenue band, not the public median

A private DTC brand at $5M-$15M revenue, scaling on paid acquisition with 55-60% gross margin and a 90-day payback target, should be running 25-40% of revenue on marketing. That's not a failure mode. It's the cost of scaling pre-retention-curve. The public median compressed because the public sample is dominated by brands at $200M+ revenue with 40-60% repeat rates and meaningful retail/wholesale revenue underneath. Comparing your $12M brand's 35% marketing ratio to Lululemon's 5% is comparing two completely different financial structures.

The right peer is a brand 12-18 months ahead of you, not a brand 5 years ahead

If you're running a $20M apparel DTC, your reference point is FIGS in 2020 (14.8%) or Revolve in 2018 — not Lululemon today. If you're a $30M beauty CPG, your reference is e.l.f. in its 2018-2020 window, not e.l.f. today. The public datasets are useful for the trajectory and the mechanics; the levels are only useful if you're matching stage.

Watch the spread, not the median

The most useful number in the pooled-by-year table isn't the median — it's the gap between P25 and P75. In 2021 that gap was 9.8 points. In 2022 it was 9.5 points. In 2023 it tightened to 8.4 points. By 2024 it was 7.1 points. The public sample is converging on a tighter band because the operators with broken unit economics either compressed hard or got delisted. If you're running far outside the band — either way — you're probably running on borrowed time or running on a model that doesn't compound.

Don't borrow the headline number; borrow the framework

The framework that drove the 2023-2025 compression in public DTC was the same one I run with clients privately:

• Marketing efficiency ratio (MER) as the directional check — total marketing divided by total revenue, run weekly
• Contribution margin after marketing as the actual operating metric — what's left after CoGS, fulfillment, and marketing
• Retention curve and repeat rate as the leverage point — every 5-point lift in 90-day repeat rate compresses your marketing-to-revenue ratio by 2-4 points without changing paid spend
• Top-of-funnel allocation at 5-15% of marketing budget as the long-tail investment, with separate measurement (incrementality testing, MMM, brand search lift)
• Channel diversification — Meta capped at 50-60% of paid spend, with the remainder in Google, TikTok, retention, and brand

That's the framework that delivered the public-DTC compression. Run it privately and you'll converge on the same band — but at a stage and at a pace that matches your unit economics, not someone else's.

If you can't measure top of funnel cleanly, you probably should still be allocating to it. The brands that won the 2023-2025 compression cycle aren't the ones who got the best at last-click measurement — they're the ones who got comfortable spending where they couldn't measure perfectly, while keeping the measurable channels disciplined.

The 2026-2028 outlook for private DTC marketing intensity

Three forces are shaping the next 2-3 years. First, AI search and zero-click results are eroding organic traffic for many DTC categories — meaning paid media share of acquisition is structurally rising, not falling, even as marketing-to-revenue ratios in public DTC stay flat. Second, third-party cookie deprecation widens dark-funnel attribution gaps, making MMM and incrementality testing table stakes rather than nice-to-have. Third, retention software (Klaviyo, Yotpo, post-purchase platforms) keeps maturing — meaning the retention-driven denominator effect that compressed public-DTC ratios in 2023-2025 keeps working for private brands that build the stack.

The private DTC operator who treats 2026 as "iOS14 is over, paid media is back" will lose the next two years. The operator who treats it as "the measurement stack is permanently rebuilt, retention is the leverage point, and brand is the long-tail bet" will compound through it.

Frequently Asked Questions

What was the DTC marketing spend trend from 2020 to 2026?

Pooled across 11 public DTC brands, median marketing spend as a percentage of revenue moved from 20.7% in 2020 (small sample) and 16.5% in 2021, to 14.7% in 2022, then dropped to 10.6% in 2023 as brands rationalized post-iOS14, and rebounded modestly to 12.4% in 2024 and 13.3% in 2025. The arc is: pre-iOS14 efficiency, an attribution-loss shock that masked rather than raised median spend, and a post-2023 rationalization where mature brands compressed spend and immature brands stayed elevated.

How did iOS14 change DTC marketing spend?

iOS14's AppTrackingTransparency framework, rolled out in April 2021, broke last-click attribution on Meta. Public DTC brands responded in two ways. Some increased spend trying to compensate for lost signal — Beauty Health hit 43.8% of revenue in 2022, Celsius hit 54.0%. Others rationalized hard — Honest Co cut from 17.0% in 2021 to 10.6% in 2023, Bark from 15.9% to 10.4%. The median fell, but the spread widened. Variance, not the median, is the iOS14 story.

Why did DTC marketing spend drop after 2023?

Three reasons. First, brands that escalated through 2022 hit the limit of what their unit economics could support and pulled back. Second, brand-and-performance shift — mature DTC brands moved budget from pure direct response to top-of-funnel and brand, which generates revenue with a delay rather than dollar-in-dollar-out. Third, retention took share — by 2024 around 60% of revenue at scaled DTC was repeat, which structurally lowers marketing-to-revenue ratios because the denominator grows from non-acquired revenue.

What is a healthy marketing spend percentage for a private DTC brand in 2026?

It depends on stage, gross margin, and channel mix. Most $5M-$50M DTC brands run 25-40% of revenue when scaling on paid acquisition. As they cross $50M and retention ramps, the public-brand band of 12-22% becomes plausible. Below 12% is rare for a brand that hasn't built out wholesale or retail revenue underneath the DTC layer. The framework: pick a target marketing-to-revenue ratio that's consistent with your CM2 and your retention curve, not one borrowed from a public peer with different unit economics.

Did public DTC brands cut marketing spend or just shift channels?

Both. The pooled median dropped from 16.5% in 2021 to 10.6% in 2023 — a real cut. But under the median, the channel mix shifted. Meta's share of US DTC ad spend fell from 34.9% in Q1 2021 to 27.0% by Q1 2022. Brands reallocated to Google search, TikTok, offline media (direct mail, podcasting), and increased retention spend from roughly 5-10% of marketing budget pre-iOS14 to 15-30% post. The headline number compressed; the composition behind it changed even more.

Sources

• SEC EDGAR 10-K filings, fiscal years 2018-2026, for: Warby Parker (WRBY), e.l.f. Beauty (ELF), Bark Inc. (BARK), Revolve (RVLV), FIGS (FIGS), Beauty Health (SKIN), Yeti (YETI), Honest Co (HNST), Beyond Meat (BYND), Lululemon (LULU), Celsius Holdings (CELH).
• Eightx benchmark database, "Public DTC Marketing Spend % of Revenue 2018-2026," pooled by fiscal year.
• Rockerbox, "iOS14 and the State of DTC Ad Spend One Year Later," covering Meta share of DTC paid spend Q1 2021 vs Q1 2022.
• Modern Retail, "How 3 DTC Brands Are Reallocating Their Advertising Spend After Apple's iOS14 Update."
• Compiled 2026 marketing attribution statistics (multi-touch adoption and MMM tooling spend).
• Apple Developer documentation, AppTrackingTransparency framework, iOS 14.5 (April 2021).