The 2026 paid ads stack: Meta, TikTok, Google, Microsoft — what works for what

We have run paid budget across Meta, TikTok, Google, and Microsoft for every business we operate, and across roughly twenty client accounts in adjacent verticals. The total annual spend across the portfolio is in the seven figures and the s

We have run paid budget across Meta, TikTok, Google, and Microsoft for every business we operate, and across roughly twenty client accounts in adjacent verticals. The total annual spend across the portfolio is in the seven figures and the spread of campaign types is wide enough — DTC, B2B SaaS, DACH consumer, US premium, EU mid-market — that we have stopped arguing about which platform is "best." There is no best. There are four platforms, each of which earns a specific role inside a stack, and the operator's job is to decide which roles are open this quarter. The 2026 paid ads stack works when each platform is hired for the job it is good at — Meta for catalog conversion, TikTok for demand creation, Google for intent capture, Microsoft for the bottom-of-funnel niche where Google's CPM has gotten lazy — and breaks when an operator picks a single platform and treats it as the entire stack. This is how we allocate, by buyer profile and product type, with the actual CPM and CPA bands we see in 2026.

The frame: platforms have jobs, not loyalties

Every operator we have ever talked to who is "doing paid wrong" is doing one of two things. They are running a single platform — usually Meta — and asking that platform to do every job in the funnel. Or they are running all four platforms with the same campaign goal and the same creative and grading them on the same first-click ROAS, and then concluding that "TikTok doesn't work for us" or "Google is too expensive."

Neither is the platforms' fault. The frame is wrong. The four platforms are doing four different jobs and need to be measured on four different things.

The CPM/CPC spread alone tells you the platforms are not interchangeable. A DTC brand whose unit economics survive a $14 Meta CPM may not survive a $90 Google CPC for the same keyword cluster, even if the keyword has clearer intent. A B2B SaaS brand spending $80 a click on Google may double its conversions by adding Microsoft at $1.20 a click against the same intent set, with no creative change. The arithmetic decides which platform earns budget; the creative decides whether that platform earns budget well.

Meta: the catalog conversion engine

Meta in 2026 is what Meta has been since the Advantage+ Shopping rollout stabilized in mid-2024. It is a catalog-driven conversion machine that performs best when fed a healthy product feed, broad targeting (the algorithm decides), and a high cadence of creative variants. The platform-side improvements over the last 18 months have moved budget further away from operator-controlled audience targeting and further toward creative volume as the primary lever.

The buyer profile Meta wins for: broad-consumer DTC under roughly €100 AOV, with a SKU library that maps cleanly onto a Shopify or WooCommerce catalog, and a creative pipeline that can produce at least 12–20 fresh variants per month. That last condition is the gating one. Without creative volume, Advantage+ Shopping has nothing to test, the algorithm narrows on a small set of winners, those winners fatigue, and CPA quietly climbs.

For our WondraKids account in DACH — kids accessories, €38 AOV, mid-five-figure monthly revenue — Meta is 75% of the paid stack. We documented the actual per-month creative cadence and the ROAS curve in the Meta ads playbook for kids products. The headline number is that ROAS climbed from 1.76 in January 2025 to 2.91 in December 2025, almost entirely on the back of going from 14 to 38 new creatives per month. Account structure barely changed. Audience barely changed. Creative cadence changed.

For a premium-priced brand — say €200+ AOV in fashion or wellness — Meta still earns its place but the role narrows. The creative spend rises (because each ad has to do more education work), the conversion window stretches to 14–28 days, and the right-frame metric is incrementality, not last-click ROAS. Most premium accounts under-spend on Meta because they grade it on a frame that suits impulse buyers and read the result as "Meta doesn't work for our category."

Meta does not work for: B2B targeting that needs job-title precision; products with regulatory copy restrictions (kids supplements, certain financial products, anything in the regulated GLP-1 space we touched on in the content automation pipeline writeup); and any brand whose buyer is over 55 and not on Instagram. The first two we work around with platform-side workarounds; the third we redirect to Google and Microsoft.

TikTok: the demand-creation surface, not a Meta substitute

TikTok ads vs Facebook is the wrong comparison. TikTok is not a substitute for Meta for most categories. It is a different role in the funnel — the one Meta doesn't do well anymore.

The buyer profile TikTok wins for: anything where the purchase decision has a discovery moment that doesn't currently exist in the buyer's life. Skincare brands targeting buyers who don't yet know they have the problem the product solves. Software targeting users who don't yet know there's a tool for the workflow they're hand-rolling. Wellness products with a "wait, this exists?" moment baked into the creative. Meta is good at converting demand that already exists. TikTok is good at creating demand that didn't exist this morning.

The CPM band is half of Meta's — typically $4–12 for the categories we run — and the first-click ROAS is correspondingly worse, because the buyer needs more touchpoints. The trap most accounts fall into is reading first-click ROAS at 0.6 on TikTok and concluding the platform doesn't work. The right read is to look at assisted conversions on a 14-day window. In the four DTC accounts where we have set up multi-touch attribution properly, TikTok contributes 18–34% of assisted conversions on Meta-attributed sales. Pulling the TikTok spend would pull the Meta ROAS down by a measurable amount. We have run that experiment twice and both times the answer was the same.

The creative requirement on TikTok is different from Meta's. Studio product shots fail. UGC-style native-creator video is the entry-level creative; the high-performing creative is something that looks like organic TikTok content with no clear "ad" framing. We documented the production economics of this in the Kling 3 for ad creatives writeup — the unit cost of a Kling-generated video ad is roughly 1/20th of a UGC-shot equivalent, with about 75% of the per-impression performance for the categories we have tested. That ratio makes TikTok's creative-volume requirement (12–20 fresh assets per month) economically tractable in a way it wasn't in 2024.

The buyer profile TikTok does not win for: anything over 50 in age. Anything where the purchase has a multi-stakeholder approval (most B2B). Categories with regulatory copy that doesn't translate to TikTok's casual format. And — counter-intuitively — most luxury and premium-priced product, where the platform's demographic mismatch is not solvable with creative alone.

Google: the intent capture surface

Google is the platform we under-rate the most when we are in a creative-heavy mode. The right way to think about Google is that it is not a paid-social channel at all. It is a query-intent channel. Every dollar spent on Google search is buying an intent moment that a buyer has already generated themselves — they have typed something into a search box. The platform's job is just to put the brand in front of that already-formed intent.

The buyer profile Google search wins for: bottom-of-funnel branded search (always; this is the cheapest, highest-converting spend we run on any platform); high-consideration purchases where the buyer is researching ("best CRM for solo consultant," "alternatives to X"); B2B SaaS demand capture; and any category with established Google Shopping coverage where the SKU is already in the feed.

Google's downside in 2026 is the cost. CPCs for B2B SaaS keywords have climbed every year of the last five and the floor on competitive keyword clusters is roughly $45 in our spend bands. The arithmetic that works on Google is volume × intent × conversion-rate, and the conversion-rate term has to do a lot of the work. A $90 CPC at a 12% landing-page conversion rate at €240 AOV is a customer at €60 CAC, which is a deal. A $90 CPC at 4% conversion at €120 AOV is a customer at €280 CAC, which is not.

The lesson from the last three years across our portfolio is that landing-page conversion rate is the variable that decides whether Google works. We documented the Meta-CPM impact of landing page quality in the GemPages-to-Horizon migration writeup; the Google-CPC impact is even more direct. Google grades the LP via the Quality Score system; a low Quality Score doubles or triples the effective CPC for the same auction position. The single highest-leverage paid-ads investment most B2B accounts under-make in 2026 is rebuilding the bottom-of-funnel landing pages.

The buyer profile Google does not win for: brand-discovery for unknown product categories (the user has to type the search; they can't search for what they don't know exists); broad-consumer impulse buys at low AOV (the CPC arithmetic doesn't survive); and most pure-content categories where intent is poorly correlated with purchase.

Microsoft: the underrated B2B converter

Microsoft Ads is the platform every operator we know either ignores or quietly over-loves. The split is roughly: agencies running consumer DTC ignore it; agencies running B2B SaaS treat it as an unfair advantage. Both reactions are correct, in their categories.

The buyer profile Microsoft wins for: B2B professional buyers age 35+; users on desktop browsers (Microsoft's traffic skews desktop more than Google's, which is closer to mobile-first); branded search in low-competition niches; and any campaign where the same Google ad set has been run and Google's CPC has gotten lazy. We typically copy the Google search campaign 1:1 into Microsoft and let it run for two weeks. In about 60% of the B2B accounts we have tested this on, Microsoft delivers comparable or better conversion volume at 30–60% of the Google CPC.

The catch is that Microsoft's audience volume is much smaller than Google's. The traffic ceiling is real — for most B2B SaaS verticals we run, Microsoft caps out somewhere between 10% and 30% of the conversion volume Google can produce, even at full saturation. The right way to think about Microsoft is as a CAC-reducer for a portion of the funnel, not a primary channel.

Microsoft does not win for: most consumer DTC (the audience demographic mismatch is too large); any creative-heavy category where the visual format matters (Microsoft's ad formats are more conservative); and brand-discovery campaigns (Microsoft's display network is smaller and less performant than Google's).

Microsoft is the channel everyone talks themselves out of running because the volume is small. The arithmetic doesn't care about the volume. It cares about the unit cost of the conversion.

— our paid lead, after the third B2B account where Microsoft beat Google on CPA

Allocation in practice: what we actually run

Across the portfolio, the default allocation we start with depends on the buyer profile and product type, not on the platforms themselves. Two profiles cover most of what we operate.

Profile A — broad-consumer DTC, sub-€100 AOV. The default split is roughly Meta 60%, Google 20% (mostly branded search and Shopping), TikTok 15%, Microsoft 5%. Meta does the catalog conversion work. Google captures the intent moments the brand has already created. TikTok creates demand the brand doesn't yet have access to. Microsoft is a small-volume CAC-reducer on branded search. This is roughly the WondraKids allocation, and roughly the allocation we have seen work for three other DACH consumer accounts in adjacent categories.

Profile B — B2B SaaS, multi-stakeholder buyer, $50+ MRR. The default split is Google 50%, Microsoft 25%, Meta 15% (LinkedIn-style retargeting and lookalikes off the customer list), TikTok 10% (only if the buyer is under 40 and the product has a discoverable workflow narrative). Google does the heavy lifting because the buyer is searching. Microsoft halves the CPA on a meaningful slice of that intent. Meta does retargeting, not prospecting. TikTok is optional and only earns budget if the creative works.

The portfolio split today across all the accounts we operate is roughly Meta 51%, Google 24%, TikTok 13%, Microsoft 7%, with the remaining 5% in experimental allocations across Reddit, Pinterest, and LinkedIn for accounts where those make sense. Reddit and Pinterest each take a small budget for the brands where they have specifically earned a spot; LinkedIn rarely earns its CPM for the spend levels we operate at.

Platform	Profile A (DTC)	Profile B (B2B SaaS)	Portfolio default
Meta	60%	15%	51%
Google	20%	50%	24%
TikTok	15%	10%	13%
Microsoft	5%	25%	7%
Other (Reddit, Pinterest, LinkedIn, etc.)	0%	0%	5%

The floor on any allocated channel is 5%. Anything below that is too thin to learn anything from — the spend doesn't generate enough events to escape statistical noise, and the operator ends up making conviction calls on three conversions a week. We have walked into accounts running Microsoft at 1% of total spend and explained why we'd either move it to 5% or kill it, and the answer is almost always to kill it and revisit when there is real budget to allocate.

What changes quarter to quarter

The allocation above is a starting point, not a static spec. The quarterly adjustments we run come from three signals.

Channel saturation. Every channel has a CPA elbow where the next dollar of spend produces a sharp decline in efficiency. On Meta, that elbow is usually at roughly 2.5–3x the daily spend that produces stable ROAS. On Google, it's the point where Quality Score holds but the auction position drops below the second result page. We watch for the elbow on each channel monthly and reallocate from saturated channels to under-saturated ones.

Creative pipeline health. Channels that earn more creative spend get more budget. Meta needs 15–20 new creatives per month to scale; if our creative pipeline is producing 30, Meta can absorb more spend, and we shift toward it. If the pipeline collapses (designer leaves, agency relationship breaks down), we shift away from creative-hungry channels and toward intent-driven ones temporarily.

Account-level events. A landing-page change. A new product launch. A regulatory shift in a category. Each of these changes the platform fit for the next 30–90 days. Meta penalizes landing-page changes for ~30–60 days through CPM elevation; Google penalizes them through Quality Score adjustments that resolve faster but cut deeper. We try to time spend allocation around known account events.

The number we watch most closely is blended CAC (cost of acquired customer, all channels combined, averaged over a 28-day window). When blended CAC moves more than 15% in either direction inside a month, we audit the allocation. Most months it doesn't, and we leave the allocation alone.

Where each platform breaks and why

Every platform breaks for some buyer profile. The breakage modes are surprisingly consistent across accounts.

Meta breaks for B2B with job-title targeting (the audience inference is too imprecise), for premium products graded on first-click ROAS (the conversion window is wrong for the product), and for any category where the creative pipeline can't produce 15+ variants a month (the algorithm starves).

TikTok breaks for buyers over 50, for any campaign graded on first-click ROAS only, for products that don't have a "discovery moment" in their narrative, and for accounts where the creative pipeline can't produce native-feeling video (which the Kling-style pipeline lowers the cost of, but doesn't eliminate the taste requirement for).

Google breaks for low-AOV impulse purchases (the CPC arithmetic doesn't work), for product categories with no established search demand (the user has to type the query; if no one types it, no one converts), and for any account with a poor landing-page experience (Quality Score multiplier kills the unit economics).

Microsoft breaks for consumer DTC at scale (the audience volume isn't there), for creative-heavy categories (the platform's ad formats are more text-driven), and for brand-discovery campaigns (the display network is too small to do prospecting work).

A campaign that's breaking on a platform is usually a sign of platform mismatch, not creative or budget mismatch. Operators who are six months into a Meta-only B2B account and frustrated with the result are running into the platform-mismatch problem, not a creative-execution problem. The fix is not better Meta creative; the fix is moving budget to the platform that fits the buyer.

What we'd do differently in the next portfolio

Two things, both small.

We would test Microsoft earlier on every B2B SaaS account. The test costs $2,000 in spend and two weeks of operator attention; the upside in the 60% of accounts where it works is a permanent 30–50% reduction in cost-per-conversion on a meaningful slice of the funnel. We have left this on the table on at least four client accounts.

We would set up multi-touch attribution before launching the second platform on any account. Most operator-graded "platform comparisons" we walk into are graded on last-click ROAS, which biases toward Meta and Google and against TikTok and Microsoft. Setting up a basic 14-day assisted-conversion model is a four-hour investment that changes which platforms get budget for the next quarter — and which ones get killed prematurely.

The general-case principle: platforms have jobs. The platforms have evolved into specialists. The job of the operator in 2026 is to know which buyer profile and product type each platform is the specialist for, and to allocate budget accordingly. The "best platform" debate is over. The "right role" debate is the one that earns the budget.