Meta ads for kids products on Shopify

We have run Meta ads for WondraKids — a kids-accessories Shopify store in the DACH region — since 2022. Four years, several million in Meta spend, roughly €178,000 in 2025 alone. Over that period the ad-platform has changed twice, our creat

We have run Meta ads for WondraKids — a kids-accessories Shopify store in the DACH region — since 2022. Four years, several million in Meta spend, roughly €178,000 in 2025 alone. Over that period the ad-platform has changed twice, our creative strategy has rebuilt itself three times, and the specific mechanics of what works for kids products have settled into a pattern we can now describe with some confidence. The meta ads kids products playbook for a mid-price DACH Shopify brand in 2026 is roughly: one Advantage+ Shopping campaign as the main spend vehicle, a small manual prospecting campaign as the creative-testing lab, a 3:1 static-to-video creative mix, parent-facing messaging (never kid-facing), and a creative-volume cadence of roughly 30 new assets per month — shipped at the expense of budget micro-adjustments, which no longer move the needle. This is the full year's playbook with the numbers.

The macro picture

In 2025 we spent €178,100 on Meta. Blended ROAS ended the year at 2.34, which means Meta delivered €416,800 of last-click-attributed revenue. Blended ROAS counting organic and email as a boost ran 2.61. Neither number is heroic; both are workable for our margin structure.

Spend was concentrated in Q4 — November and December alone accounted for 38% of the annual spend and a correspondingly elevated 41% of annual Meta-attributed revenue. Q1 had the worst ROAS of the year (1.87 blended) and the smallest spend. Q3 was the steadiest quarter, both in spend and performance.

The account structure we run

After three years of testing every structural configuration the internet recommended, we've converged on a small, simple account:

• One Advantage+ Shopping campaign. 75% of spend. Catalog-driven. Broad targeting (ages 25–50, DACH, no interest stacking). This is where the scale lives.
• One manual prospecting campaign. 15% of spend. Tight audience — lookalikes seeded from 90-day purchasers. Used primarily for creative testing, not for scale.
• One retargeting campaign. 10% of spend. Audience: site visitors in last 14 days who did not purchase. Single carousel ad, updated monthly, pointed at the current hero bundle.

Three campaigns. Nothing more. We had fourteen campaigns at one point in 2023; every additional one was a justification looking for a reason. We now treat the three-campaign floor as the account's opinion.

Inside ASC, we run one ad set with two ads for each major product cluster. That's five ad sets, ten ads, maybe fifteen creative variants at any time. Meta's machine does the rest. We stopped micro-managing budget at the ad-set level after a six-week test in April that showed zero performance lift from daily manual budget adjustments — and a measurable performance drag from the algorithmic instability our adjustments introduced.

Creative: the only thing that still moves the number

Ninety percent of account performance in 2025 came from creative choices. Audience, targeting, placement, device — all decisions we deferred to Meta's algorithm with better results than when we'd tried to optimize them ourselves. The creative is the only lever left that matters.

For kids products specifically, we've landed on a taxonomy of four creative types that produce the bulk of our converting ads.

Type 1: Studio product shot + parent-benefit headline. A clean image of the product, a short headline directed at the parent ("Under 30 seconds to get her out the door"). This is our highest-volume creative type. Cheap to produce (we shoot a batch of 40 in an afternoon), easy to iterate, and consistently in the middle of the ROAS distribution.

Type 2: UGC-style video of a parent using the product. A real parent (often a micro-influencer we pay for a short clip) demonstrating the product in use. 15–22 seconds. Vertical. The highest-performing ad type on a per-impression basis but the most expensive to produce. We run about 6–8 per month.

Type 3: Lifestyle photography with a kid in frame, parent-voice copy. A staged photo of the product in a family context, with copy written for the parent buying it. Mid-performance, mid-cost. Workhorse category.

Type 4: Problem-solution narrative. A longer-copy ad (three or four short sentences) that names a specific parenting frustration and positions the product as the fix. This is a slow-burn type — builds familiarity, doesn't click hard day one, shows up well in view-through attribution over two weeks.

Across 2025 we ran these roughly at a ratio of 4:2:3:1 of types 1:2:3:4. The heavy reliance on type 1 is a volume story — we can produce it fast — but on a per-ad ROAS basis, type 2 is the single best performer. If we could shoot UGC at scale, we would.

The 2025 creative treadmill, month by month

Month	New creatives	ROAS	Spend
Jan	14	1.76	8,900
Feb	16	1.98	9,800
Mar	18	2.28	12,100
Apr	19	1.89	14,200
May	22	2.11	13,900
Jun	24	2.42	14,800
Jul	20	2.51	13,600
Aug	22	2.38	14,400
Sep	28	2.47	15,200
Oct	32	2.54	17,100
Nov	36	2.68	23,400
Dec	38	2.91	20,700

Two patterns stand out. April was the worst month of the year by ROAS, despite higher spend than March, because we pushed through a landing-page migration that Meta's algorithm needed 30–45 days to re-learn. Covered in more detail elsewhere; the short version is that landing-page stability matters as much as creative stability, and changing both at once is a two-month performance hole.

Q4's creative volume is not coincidental. Meta rewards creative freshness during auction-heavy seasons because more brands are spending and the algorithm needs more variety to optimize. We ramped creative production by 40% versus summer, and the ROAS lift was real.

What doesn't work for kids products in 2026

Interest-stacked audiences. We tested "parents + baby brands + toy-store competitors" versus broad 25–50 DACH. Broad won by 12% ROAS. Meta's audience targeting from third-party signals is a decade behind the algorithm's native optimization; stop curating audiences.

Carousel ads with five SKUs. Carousels used to be our highest-click format. In 2025 they dropped below single-image ads on every metric. Our hypothesis: the placement surfaces (Reels, Stories) don't reward carousels the way feed used to. We now run exactly one carousel ad, in retargeting only.

Generic holiday messaging. "Give the perfect gift this Christmas" style creative. Below-baseline on every test we've run. A product-specific holiday ad ("Sara's Christmas outfit, €34, ready in 2 days") out-performs the generic version by 40–60%.

Kid-facing copy and voice-over. Covered above. The whole category error.

Cheap static with stock imagery. Better a real, imperfect photo than a perfectly-lit stock composite. Meta's algorithm seems to reward authenticity signals and punish stock imagery; we haven't seen a stock-based ad win in 18 months.

Expensive cinematic video over 30 seconds. We spent €4,800 on a cinematic 45-second brand video in Q2 2025. It performed at 60% of the blended baseline. The algorithm doesn't reward production value beyond a certain floor; it rewards scroll-stopping opens in the first second and a clear product-understanding within the next four.

The dead-campaign cleanup

Every quarter we run a brutal cleanup. Any ad set with less than 50 purchases in the last 30 days gets killed. Any creative that hasn't produced a purchase in 14 days gets killed. Any campaign with a ROAS below 1.5 on the last 30 days gets killed.

The number of things that get killed in a given quarter: roughly 60% of what's live. This feels aggressive. It's correct. The account gets healthier after cleanup than before, every time.

The counterintuitive part: after cleanup, total account performance usually improves for the next 30 days, not drops. Cutting underperformers lets Meta's algorithm concentrate budget on what's working, and the net effect is a lift.

We killed forty ads and spent the same money the next week. ROAS went from 2.1 to 2.4. The account had been spending on things that didn't deserve it for three weeks and nobody noticed.

— our media buyer, after the Q3 2025 cleanup

Attribution: what we trust and what we don't

We run three attribution systems in parallel and don't fully trust any of them.

Meta's in-platform attribution is the default view inside Ads Manager. 7-day click, 1-day view. In 2025 this reported €486,000 of revenue on €178,000 spend — a reported ROAS of 2.73. We treat this as the optimistic ceiling.

Shopify's attribution model counts orders from users who arrived from Meta during a session window. It reported €412,000 from Meta — a ROAS of 2.31. We treat this as the pessimistic floor.

Triple Whale blends the two and adds server-side conversion data. It reported €425,000 — a ROAS of 2.39. We treat this as the working-number, and budget off it.

The three numbers differ by about 18%, which is painful. We've stopped trying to reconcile them exactly. The signal is in the change — when all three move up together, Meta got more effective. When one rises and the others don't, we're looking at an attribution artifact rather than a real performance lift.

For the blended ROAS of 2.34 we cite at the top of this piece: that's the Triple Whale number rounded down by two points to leave margin for the attribution uncertainty. We recommend this practice to anyone running paid at our size: pick the middle attribution source, subtract a conservative buffer, manage to that.

The budget-pacing rule

After six years of ad-account work, the pacing rule that's served us best is embarrassingly simple: spend inside your ROAS floor and do not chase the ceiling.

We set a floor at 2.1 for blended ROAS. Any week where the trailing 7-day falls below it, we hold spend flat and push creative production. We don't slash (Meta penalizes sudden cuts) and we don't increase (Meta penalizes sudden increases). We stabilize at the level the algorithm was calibrated to.

Any week where the trailing 7-day is above 2.5, we increase spend by 15% and let the algorithm re-find its level. Usually ROAS drops a little at the higher spend; usually it recovers inside a week; occasionally it doesn't, and we unwind the increase.

This rule is not sophisticated. It has worked more consistently than any of the optimization techniques I've learned from ad-account experts. Meta's auction dynamics reward boring.

Black Friday: what we learned in 2025

Black Friday week 2025 was the biggest week we've ever run. We spent €14,900 across five days and brought in €42,300 of attributed revenue — a ROAS of 2.84, better than any other week of the year.

Things that contributed:

• We froze creative for the 10 days before BFCM. Every ad running BFCM had been live for at least two weeks with proven performance. No new creatives introduced in the final sprint.
• We increased retargeting share. Retargeting jumped from 10% of spend to 18% during BFCM week, because the traffic we'd built up in Q3–Q4 was primed to convert with one more touch.
• We ran a specific BFCM landing page. Product-aware copy, single CTA, no generic "sale" messaging. The landing was cut as a page in Shopify Horizon two weeks before BFCM.
• We capped the discount at 20%. Heavier discounting in prior years produced bigger headlines and worse unit economics. 20% turned out to be the sweet spot between conversion lift and margin preservation.

Things that hurt:

• One of our most-trusted creatives burned out on day 3. ROAS on the hero ad dropped from 4.1 to 1.8 in 72 hours. We killed it and replaced with its second-best variant. The lesson: even frozen creative has a decay curve, and at BFCM volumes that decay happens fast.
• iOS Safari checkout conversion was 30% lower than Chrome on mobile. We couldn't fix it mid-campaign. Something to investigate for 2026.

What a typical month costs to run

Fully-loaded ad operations on the account look like:

Line	Monthly €
Media spend (Meta + Google)	13,500–17,000
Creative production (photographer, UGC creators)	1,200
Media-buyer contractor	1,800
Shopify + Klaviyo + Triple Whale	520
Landing-page changes (contracted dev)	300
Total operating cost of paid media	17,320–20,820

Against a Q4 revenue contribution from Meta of roughly €42,000 per month, this produces a gross contribution after ad-operations cost of around €24,000 per month, before COGS, fulfillment, and everything else downstream. The account earns its keep. Just barely, in some quarters.

What to do differently in 2026

Produce more UGC. Our best-performing creative type is the one we ship least of. We've booked a UGC creator to run four sessions a month for 2026, which should lift the type-2 creative output from 6–8 to 14–16 per month.

Investigate the Safari checkout issue. A 30% mobile conversion gap between browsers is money on the floor. Our Horizon theme rebuild should fix most of it; we'll instrument specifically.

Test creative inside the Meta AI prompt layer. Meta's generative tools have evolved; we've tested them perfunctorily. Our suspicion is that they're still behind a custom creative workflow for a brand with taste, but the gap is closing and we should retest quarterly.

Reduce dependency on Advantage+ Shopping. ASC is 75% of our spend. If Meta changes the mechanics (which it will), the account is vulnerable. We're building out manual-campaign structure to serve as insurance without hurting current performance — more a resilience move than an optimization one.

Keep the 3:1 static-to-video ratio under pressure. The temptation in 2026 will be to over-shift to video because TikTok-style content dominates the conversation. For kids products in DACH, at this brand's margin structure, the ratio still holds. We'll re-check monthly.

The honest version

Meta ads for kids products on Shopify, in 2026, work. They do not work magically. They work when the creative is good enough, in enough volume, with a parent-voice discipline, shipped into a stable landing and supported by a clean account structure. The treadmill is relentless. The numbers are small. The margin is thin. And yet, for a brand our size with our distribution, this is the channel that pays the rent, and any playbook that doesn't start from that fact is writing fiction.