How to Compare Meta, TikTok and Google Ads

You give each platform $600. Same product, same offer, the same set of creatives that already sold on Meta last quarter. Three weeks later Google shows a 3x return, TikTok shows cheap clicks and no sales, and Meta reports conversions you can't find anywhere in your checkout. So which platform won?

Short answer: None of them, yet. A three-way split from one budget can't crown a winner because each platform measures conversions on its own attribution window, the cheapest-looking platform is often the one inflating its numbers with junk traffic, and $600 is rarely enough to clear any platform's learning phase. You design the test to remove those three distortions before you read it.

The takeaways

• Platform-reported ROAS is not comparable across Meta, TikTok and Google: each defaults to a different attribution window, so you are reading three different rulers.
• A thin split (say $600 per platform) often dies inside the learning phase, which Meta puts at roughly 50 conversions per ad set, so you are judging a half-trained model.
• The honest verdict comes from your own downstream record (CRM, checkout, repeat orders), not from three dashboards that each credit themselves.

Why can't you compare platform-reported ROAS directly?

Because each platform counts a conversion with a different stopwatch. Meta and TikTok default to a short click window with a one-day view window; Google Ads leans on a longer click window and last-click logic. The same sale can show up on one platform and vanish on another purely because of when and how each one looks back. You are not comparing performance, you are comparing measurement settings.

The tell is simple: add up the conversions all three platforms claim, and the total usually beats your actual revenue. Each platform sees only its own touch in a multi-step journey and credits itself by default. So the platform that "won" your test may just be the one with the most generous attribution window, not the one that drove the sale.

Is the same budget on each platform a fair test?

Equal dollars feel fair, but they rarely buy an equal test. Every platform runs a learning phase where delivery is unstable and cost is inflated until it gathers enough conversion signal. Meta's documentation puts that threshold at roughly 50 conversions per ad set; TikTok and Google have their own versions. A $600 split can expire before any of the three exits learning, so you end up ranking three half-trained algorithms against each other.

There is a creative problem hiding in here too. A static that sold on Meta is not native to TikTok's full-screen sound-on feed, and a feed image is not a Google search ad at all. When you reuse one platform's winner everywhere, you are testing fit-to-platform, not platform-versus-platform. Give each one a budget that can realistically clear learning and a creative built for its native format, or accept that the test is biased before it starts.

Which platform is quietly inflating its own numbers?

Usually the one with the cheapest clicks. Low CPMs and high reported conversion rates can come from low-quality placements and invalid traffic: bot clicks, accidental taps, audience-network junk that converts on paper and never in your bank account. The buyer who split $1,800 three ways and watched one platform burn most of the budget on traffic that produced zero sales was not unlucky, they were reading raw platform metrics instead of traffic quality.

This is where a single headline number betrays you. A platform can win on CPM and CTR while losing on the only metrics that touch money. A scoring approach that blends several signals (hook rate, CTR, cost per result, return, revenue per result) instead of one keeps a cheap-but-empty platform from topping the table on volume alone. The cheap click is only good if it survives all the way to revenue.

How do you design a head-to-head you can trust?

Standardize the ruler first. Align attribution windows where each platform lets you, then stop trusting the dashboards as the final word and reconcile every claimed conversion against one neutral source you control: your checkout, your CRM, your repeat-order data. That single source is your tiebreaker.

Then give each platform a fair shot: a budget that can clear its learning phase, a creative built native to its feed, and enough time that one viral or dead day doesn't decide it. Early platform scores swing wildly on small samples, so treat a wild first week with suspicion. The honest fix is the same one good testing systems use on noisy early data: pull that early number toward what the format normally does before you act on it, so a lucky Tuesday on TikTok doesn't win the whole budget.

Adscalr normalizes Meta, TikTok and Google into one dataset and scores creatives on a composite of metrics rather than a single platform-flattering number, which is the same discipline a fair head-to-head needs: read every platform on the same ruler, against your own downstream truth. If you want the framework behind that, see how Adscalr reads ad tests. For the metrics that belong in the composite, meta ads metrics that matter covers why no single number decides a test, and once you have an honest read, budget allocation across platforms covers what to do with it.

The platform that deserves your budget is the one still profitable in your own records after a fair test. Anything a dashboard tells you before that is a hypothesis, not a verdict.