YouTube and Meta both distribute visual advertising through algorithmic auctions, but the consumer context differs. Meta ads appear within social feeds and stories; YouTube ads appear around chosen video, search, Shorts and feed experiences. Demand Gen can also extend Google creative to Discover, Gmail, Maps and Display.
The allocation decision should not be based on which platform reports the higher ROAS. Each platform observes different interactions, uses different attribution rules and claims credit independently. The question is which channel is likely to produce the best marginal contribution for the next pound, given the brand's creative and measurement constraints.
The operating comparison
| Area | YouTube / Demand Gen | Meta |
|---|---|---|
| Primary contexts | In-stream, in-feed, Shorts and other selected Google visual surfaces | Facebook and Instagram feeds, Stories, Reels and other Meta inventory |
| Intent inputs | Google and YouTube behaviour, custom segments, first-party data | Meta activity, audience suggestions, first-party data |
| Ecommerce catalogue | Merchant Center product feed in Demand Gen | Meta catalogue and sales campaign tools |
| View attribution | Engaged-view and other ad-event types under Google settings | View-through and click attribution under Meta settings |
| Creative strength | Demonstration, education, long and short video context | High-volume native social concepts and rapid iteration |
These are tendencies, not hard borders. YouTube Shorts can resemble a social feed, and Meta can support longer video. Evaluate the actual placement and creative rather than the platform logo.
Compare current campaign products
For conversion-focused YouTube activity, Video Action campaigns have been replaced by Demand Gen. Demand Gen supports video and image creative, channel controls and Merchant Center product feeds. Google's migration guide and product-feed guide describe the current setup.
Meta's Advantage+ audience treats many audience selections as suggestions and can expand beyond them, while retaining controls such as location, minimum age, language and custom-audience exclusions. Meta's official Advantage+ audience overview is the primary source for that behaviour.
Both platforms therefore mix advertiser input with expansion. A custom segment on Google and an audience suggestion on Meta are not deterministic lists of only those people.
When YouTube deserves the test
The YouTube ecommerce playbook covers setup and measurement in detail. YouTube is a strong candidate when the product benefits from demonstration, explanation or comparison; when Google intent signals are strategically useful; or when the brand wants to connect video exposure with Search, Shopping and Merchant Center operations.
Examples include unfamiliar mechanisms, considered purchases, products with visible results and categories where buyers research on YouTube. A longer in-stream explanation and a vertical Shorts edit can express the same proposition at different depths.
The risk is overvaluing post-view attribution or producing one expensive brand film with no testing system.
When Meta deserves the test
Meta is a strong candidate when the team can produce a steady volume of native social concepts, the product is discovered visually, community or creator proof is abundant, and the account has enough conversion data and catalogue quality for sales optimisation.
The risk is assuming historic average ROAS will persist as spend expands, or treating broad delivery as proof that selected interests caused the result.
Use one proposition, native executions
The YouTube creative-testing guide provides a concept matrix for this work. Share the strategic idea, not necessarily the exported file. A comparison concept might become:
- A 30-second YouTube in-stream demonstration with mechanism and proof
- A vertical Shorts edit that reveals the result immediately
- A Meta Reels creator version with faster cuts and social proof
- A feed carousel comparing concrete product attributes
Track the common concept and the platform-specific execution separately. If YouTube wins with a demonstration and Meta loses with a testimonial, the result does not isolate the platform.
Make attribution comparable before reading ROAS
- Use the same purchase or qualified-lead definition.
- Align revenue, currency, refunds and duplicate handling.
- Document click and view windows on each platform.
- Segment Google's conversions by ad event type.
- Compare new-customer and backend contribution using one warehouse or analytics source.
- Measure total business movement and overlap, not just two platform columns.
Google's engaged-view conversion documentation explains its post-view event. A Google conversion including engaged views and a Meta conversion under another window are not equivalent observations.
Allocate on marginal contribution
Historical average ROAS includes cheap early reach, remarketing and brand demand. The next £1,000 may reach colder users at a lower return. Estimate contribution after product cost, fulfilment, discount, returns, media and incremental creative cost.
Hypothetical example: Meta spends £20,000 at 350% attributed revenue ROAS and YouTube spends £5,000 at 280%. Meta is not automatically the next-budget winner. If its last £5,000 produced 180% while the YouTube test improved with a new demonstration, the expected marginal result may favour YouTube. The numbers are illustrative, not benchmarks.
Increase budgets in measured steps and monitor customer mix and total contribution. A change in platform-reported credit can occur without the same change in total sales.
A fair cross-channel test
Use the same product set, offer, geography and broad time window. Build creative from the same proposition while respecting each placement. Predefine spend, primary backend outcome and minimum test duration. Exclude existing customers consistently where possible.
A direct platform split is not randomised because each system finds different people and optimises differently. For a causal budget decision, use matched geographic cells or another holdout design, check for spillover and keep other media stable. Platform experiments answer within-platform questions more cleanly than they compare two separate auctions.
What to do with a small budget
Do not split £100 per day across two platforms, four audiences and ten creatives merely to "be everywhere." Fund the channel with the clearest customer hypothesis and strongest creative capacity. Use the result to improve the proposition, then introduce the second channel when it can receive a fair test.
If profitable Search or Shopping is budget constrained, capturing existing demand may take precedence over either prospecting channel. Allocation begins with opportunity cost across the whole account.
A decision scorecard
- Can the product benefit from explanation or social discovery?
- Which platform has the stronger current creative pipeline?
- Are feed and purchase data reliable on both?
- Can attribution definitions be aligned?
- Which channel has more unsaturated marginal opportunity?
- Can the team measure new customers and contribution outside the platform?
- Is the available budget enough to learn?
A cross-channel reconciliation template
Build one weekly table outside both platforms. Rows should include spend, attributed orders by click and view treatment, net revenue, first-time customers, gross profit, contribution after media, refunds and creative cost. Add platform-specific reach and frequency as diagnostics, not common outcome metrics.
Freeze a reporting date so delayed conversions have the same maturity. Use one order system to deduplicate revenue. If both platforms claim the same order, keep each platform's attributed view for optimisation but count the order once in the business total. Document currency conversion, tax, shipping and refund treatment.
Hypothetical example: Google reports £18,000 conversion value and Meta reports £16,000, while the store records £27,000 total revenue from the promoted products. Adding platform values to £34,000 would double-count overlapping journeys. The observed fact is that each platform assigned credit under its own rules. The incremental contribution of either channel remains unknown until a holdout or credible budget test supplies a counterfactual.
Cross-channel failure cases
- Different customer definitions: one platform reports "new" using a model while the warehouse uses first-ever order. Reconcile to the warehouse definition.
- Creative asymmetry: Meta receives 20 native concepts and YouTube receives one horizontal brand film. The test compares execution capacity, not only channels.
- Remarketing imbalance: one channel includes warm visitors and the other excludes them. Match policy or report the roles separately.
- Different offer exposure: a Meta code discounts the order while YouTube uses full price. Compare contribution, not revenue ROAS.
- Budget starvation: each channel receives too little conversion volume, then noisy averages are used to crown a winner.
When these cannot be aligned, do not force a league table. State the asymmetry and make the narrower decision the evidence supports.
What this comparison cannot settle
Audience systems and auctions change, platform reports overlap and no standard dashboard shows the counterfactual. Creative quality can dominate the comparison. A result for one product, season and budget level may not generalise.
Treat channel choice as a portfolio decision reviewed at the margin. The winner is not the platform with the best screenshot; it is the investment that adds the most defensible profit and learning.