In short
- Is: an ad platform such as Meta, Google, Amazon or TikTok that owns a logged-in audience, sells ads on its own inventory and measures them with its own tools, while user-level data stays inside
- You can take out: results per campaign, ad set and ad, and from Google each click with its ID. Who saw an ad and which of your orders a platform credited stay inside, or come out only as aggregates from clean rooms
- Grades itself: each platform credits its own ads by its own rules, views and modeled conversions included, and none removes orders another platform also claims, so their sum exceeds your shop's revenue
- Measured outside: in our panel, Meta Ads returns a median ROAS of 1.01 on last touch (September 2026), and Ads Manager claims more for the same brands
- Regulated: the EU's Digital Markets Act lists the ad services of Alphabet, Amazon and Meta as gatekeeper services and entitles advertisers to the data they need to verify their ads independently
What a walled garden is
A walled garden in advertising is a platform that owns a logged-in audience, sells ads on its own inventory and measures their results with its own tracking, while the user-level data stays inside. You buy reach and read results, but you can’t take out who saw an ad, who clicked or who bought.
Meta, Google and Amazon run the largest ones. What they let you take out:
| Platform | What you can take out | What stays inside |
|---|---|---|
| Meta | Results per campaign, ad set and ad, totals for up to 37 months; impressions and spend per ad set, day and region for marketing mix models | Who saw or clicked an ad, and which of your orders a reported purchase is |
| Campaign reports and each click with its gclid, one day at a time, up to 90 days back. Ads Data Hub joins your first-party data with Google’s event-level ad data | User-level data: Ads Data Hub returns only aggregates and drops rows with too few users | |
| Amazon | Amazon Marketing Cloud, a clean room free for eligible advertisers, joins your pseudonymized data with Amazon Ads impressions, clicks and conversions | Shopper-level data: the clean room returns only aggregated, anonymous results |
Sources: Meta’s Insights API, data limits and MMM export; Google’s click view, Ads Data Hub and its privacy checks; Amazon Marketing Cloud.
Why walled gardens grade their own homework
Inside its wall, each platform credits its own ads, by its own rules:
- Meta counts purchases within one or seven days of a link click, depending on the attribution setting, within one day of an engagement such as a share, a save or a 5-second video view, and within one day of an impression nobody clicked, as a view-through conversion (Meta, Meta).
- Google Ads uses data-driven attribution by default for most conversion actions, counts clicks for 30 days by default and mixes observed and modeled conversions in its Conversions column (Google, Google, Google).
- TikTok credits conversions up to 28 days after a click and up to 7 days after a view (TikTok).
None of them sees the others’ ads, so a buyer who clicked a TikTok ad, saw a Meta ad and then searched on Google counts in all three accounts. Each number holds by its own rules. That is the core of platform-reported conversions: the seller of the ads also judges their results.
Three walled gardens, one shop: a worked example
A shop takes 2,000 orders in a month at €50 net of VAT each, €100,000 in total, and spends €25,000 on Meta, Google Ads and TikTok, whose pixels receive the same net values. In the shop’s own tracking, 1,400 orders came after at least one paid click, 500 of them after clicks on two platforms; 600 came only through email, organic search or direct visits.
| Spend | Revenue the platform reports | Platform ROAS | Last touch, shop data | ROAS, shop data | |
|---|---|---|---|---|---|
| Meta | €12,000 | €50,000 | 4.17 | €26,000 | 2.17 |
| Google Ads | €8,000 | €45,000 | 5.63 | €34,000 | 4.25 |
| TikTok | €5,000 | €15,000 | 3.00 | €10,000 | 2.00 |
| Email, organic, direct | €30,000 | ||||
| Total | €25,000 | €110,000 | 4.40 | €100,000 | MER 4.00 |
The platforms claim €110,000 for a shop that took in €100,000, and leave nothing for email, organic search and direct visits. Their extra €40,000 over the €70,000 the shop’s data credits to them on last touch has two sources: 500 orders touched two platforms and count twice (€25,000), and 300 conversions came after an impression or were modeled, with no click in the shop’s data (€15,000). Only the shop-data column adds up to the shop’s revenue.
What you can measure outside the walls
What happens on your own domain is yours, and so are your orders:
- Clicks and orders. An ad click lands on your site with its click ID and URL parameters. Recorded as first-party data and matched to your orders, it lets one attribution model credit each order once across all platforms. Sent back server-side, through Meta’s Conversions API or Google’s conversion upload, your conversions feed their bidding.
- The sum. Shop revenue divided by total ad spend, MER, can’t be inflated by any platform.
- What clicks can’t show. An impression nobody clicked leaves no trace on your site. Whether views caused sales takes an experiment: a holdout test that switches ads off for a random group or some regions measures incrementality on your own sales. Marketing mix modeling estimates it from spend against sales over time, but only with years of weekly data in which each channel’s spend moved on its own. Meta’s Conversion Lift holds out a control group too, but Meta draws the groups and computes the result (Meta).
Walled gardens and the Digital Markets Act
In September 2023 the European Commission designated six gatekeepers under the EU’s Digital Markets Act, with six months to comply. Their designated services include the ad services of Alphabet, Amazon and Meta; for ByteDance and Microsoft, they include TikTok and LinkedIn, not the ad businesses (European Commission). Two rules of the regulation concern advertising data:
- Article 6(8): on request and free of charge, advertisers and the third parties they authorise get access to the gatekeeper’s performance measuring tools and to the data for their own independent verification, aggregated and non-aggregated.
- Article 5(2): combining personal data across a gatekeeper’s services needs the user’s consent. In April 2025 the Commission fined Meta €200 million for not giving EU users the choice of a service that uses less of their personal data (European Commission).
Meta and Google measured from outside: our panel
Our benchmarks measure the two largest walled gardens from outside: first-party tracking matched to the shop orders of the D2C brands in our panel, each order credited once, on gross order revenue before returns. On last touch, the median ROAS for September 2026 is 1.01 on Meta Ads and 2.66 on Google Ads without brand search.
For the same brands, Ads Manager reports a higher median ROAS than our last touch, and Google’s own reporting gives much of its credit to brand search, which these figures leave out. The Meta Ads benchmarks and Google Ads benchmarks show both side by side.
How adtribute works with walled gardens
- Your domain. Tracking runs first-party on your domain, through two DNS records and the tracking code on every page (docs: Connect your domain for first-party tracking). URL parameters on Meta, Google and TikTok ads carry the ad, ad set and campaign IDs into each visit (docs: Meta Ads).
- Each order once. Your attribution model splits each order’s value across the channels its journey touched, so the channels add up to your orders, except in Full Impact and Channel Impact, which give every touchpoint, or the first touchpoint of each channel, the full value (docs: Attribution models in adtribute).
- Platform numbers differ by design, and the platform-reported conversions entry’s section on how adtribute handles them shows what to compare them with.
- Back through the wall. Actions send conversions you define, such as new customers’ orders only, to Meta’s Conversions API (CAPI entry: how adtribute sends events to Meta) and to Google Ads, which shows only those it can tie to a Google Ads click (docs: Google Ads Conversion Upload).
Our view: verify the walled gardens, don’t avoid them
Our view: the walled gardens are where the reach is. A brand that leaves Meta or Google to escape their reporting gives up its strongest channels to fix a measurement problem. And part of what stays inside is personal data that no platform could simply hand over.
So keep the gardens and move the verdict outside. Use each platform’s numbers to compare its own campaigns with each other, and your shop’s orders to compare platforms and set budgets: MER for the total, one attribution model for the split, a holdout test where impressions do most of the work. The walls are opening a little: Meta’s click-through attribution now counts only link clicks, to align with tools like Google Analytics, as the attribution window entry’s section on click vs. view windows and platform defaults explains, and its Conversions API accepts your attribution model’s credit per purchase (Meta).
FAQ about Walled Gardens
Which platforms are walled gardens?
Meta (Facebook and Instagram), Google (Search, Shopping and YouTube) and Amazon are the largest; TikTok, Pinterest, Microsoft and LinkedIn run smaller ones. All of them sell ads on their own inventory and measure the results themselves.
Why do Meta, Google and TikTok together claim more revenue than my shop made?
Each credits its own ads by its own rules, including conversions after a view without a click and modeled ones, and none removes orders another platform also claims. A buyer who clicked ads on two platforms counts in both. Judge the total on MER and the split on one attribution model across all channels.
Is Ads Manager good enough as our only source of truth?
For comparing Meta campaigns with each other, yes: it shows the conversions Meta’s bidding works with. For budgets across channels, no: it sees only Meta’s ads and can’t tell what Google, email or a direct visit did for the same order. Keep your shop’s orders as the source of truth and read Ads Manager against them.