Creatify With Causality Engine: Check the Blended ROAS It Reports

The short answer: a blended ROAS of 3.3 in a morning brief and a store whose week does not support it are two different sums, not a contradiction. Creatify, known for turning product pages into AI video ads, now sells an agent that connects to Meta, Google Ads, TikTok and other ad accounts, and its daily brief reports "spend, revenue, blended ROAS and CPA" pulled from those platforms (Creatify Agent page). Every input is a platform's own claim. Causality Engine, a causal attribution tool, reads the store's GA4 export instead and estimates what each channel caused; the first finding is free on the start page.

Disclosure: I founded Causality Engine and own it, and this page recommends it. Every claim about it links to Causality Engine's own pages.

What to settle before approving a raise:

• Redo the blend with the store's revenue. For example, $80,850 of platform revenue on $24,500 of spend is a blended ROAS of 80,850 / 24,500 = 3.3. If the store booked $63,700 that week from every source, its own figure is 63,700 / 24,500 = 2.6, and that one includes email and organic sales.
• Ask the agent which rule it used. Creatify's Meta agent page shows it naming the conversion event and window in its reports ("Reported on Purchase, 7-day click attribution"). Note both for every account.
• Find the matching rows in GA4. Meta clicks belong in Paid Social and Google's in Paid Search; the Traffic acquisition report shows whether they arrive there.

The two tools, fact by fact

How Creatify decides what won

The agent keeps no purchase counter of its own. Its creative testing page is direct about this: "It reads your ad platform's own results; there's no separate A/B tool." What it adds is discipline over those results, and the discipline is worth having:

• A rule set before launch, with a floor: no call under 10 conversions per variant and none before day 7.
• A range, not a point. Each variant gets a 95% range on cost per purchase, and the winner's whole range must sit under the target.
• Raises in steps. Scaling goes up about 20% at a time; a raise over 30% on Meta, 20% on Google Ads or 50% on TikTok is flagged for an explicit OK (agent page).
• Approval cards. Pauses, budget changes and launches wait for a person; a routine can be switched to act alone.

Every one of those guards works on the purchase column the platform filled, and on Meta that column follows the attribution setting chosen per ad set (Meta's attribution page). A tight range around a platform's count is still a range around the platform's count.

On the pages read on 11 November 2026, the agent's connections are ad platforms only, with no store and no GA4 among them (Creatify glossary entry), so the revenue in the brief is the platforms' revenue.

How a Creatify ad's click reaches GA4

Creatify launches on approval; the tagging sits in the ad account. Meta's URL parameters help says ads may have Campaign source set to Facebook or Instagram and Campaign medium set to paid added automatically since 1 January 2024, and that parameters typed in Ads Manager override them.

That pair lands where it should. Google's default channel group files a visit under Paid Social when the source matches its list of social sites and the medium matches cpc, ppc, retargeting or anything starting with paid. A click with no tags at all arrives as a social referral, which the same rules file under Organic Social, and the ad loses credit it earned. The same page gives TikTok as an example of a video site under Paid Video. Meta also warns that a parameter covers the first page only: a buyer who leaves and comes back later arrives untagged.

The claim gap, as a worked example

For example, with figures made up to show the sums, not taken from any store or any Creatify account: one week, three accounts, all read by the agent's daily brief.

The store booked $63,700 that week from every channel. The three platforms claim 80,850 - 63,700 = $17,150 more than the store took, before email, organic search and Direct have claimed anything. Divided the store's way, the week returns 2.6, not 3.3.

The brief is right by its own rule: it sums three rules that each count a buyer who touched more than one ad.

A causal read takes the GA4 export from the same week and earlier and, per its how-it-works page, puts each channel's caused share next to a last-click rebuild, with a data-health score from 0 to 100 and a next step: raise, cut or hold, tested in one region first. If the read rates Meta well below its claim while the agent proposes a 20% raise on Meta, approve the raise for one region and keep the rest flat.

The free check, every Monday

Divide the store's revenue for seven days by the brief's total spend for the same days, and log it beside the blended ROAS. Over a month the two lines move together or drift apart; drift is the signal. The blended ROAS calculator does the division.

Then make one test cheap. Before an approval card scales a winner everywhere, apply it to ad sets that target one region and leave a similar region unchanged for three weeks. Compare store orders by shipping region, before and after. That holdout is by hand; Creatify's own pages document no geo test, and they say plainly that the agent has no A/B engine of its own.

Where the read gives no answer

• No link between them. Creatify connects to ad accounts. Causality Engine reads a file you export from GA4, and its integrations page describes no connector at all.
• Channels, not creatives. The read reports Paid Social as one row. Which hook won, which avatar fatigued, which ad to render next: those stay Creatify questions, and its testing rules answer them inside the platform's count.
• Data floors. Twelve months or more of GA4 data reads better (how it works), and a channel under 1% of credit on both methods is called too small to act on. The home page says that when spend is too small to find a signal, it declines.

The agent's loop has a price attached at each turn: when a winner fatigues, the same chat "writes the new hooks and renders fresh video and image ads to test", and the FAQ says the agent uses credits as it works, so every fatigue verdict drawn from a platform's purchase column also ends in renders that Creatify's credits pay for.

When the two lines keep drifting apart, Causality Engine, which I founded, reads the store's GA4 export on the start page; the first finding costs nothing and stays in your browser, and the full read is €99 once with a 30-day refund (pricing, checked on 11 November 2026).

See also

• Creatify: the video maker, its credits and complaints
• How to use Creatify: product page to test-ready video ads
• Creatify on the marketplace: Creatify's affiliate program terms
• Causality Engine: what the causal read covers and costs
• How to use Causality Engine: an export, a read, a regional test
• Causality Engine on the marketplace: the owner's affiliate program terms
• Creatify vs Arcads: product-link videos against script-first actors
• Blended ROAS calculator: store revenue over total ad spend

FAQ

Can Creatify's agent read my Causality Engine results?

No. Nothing connects the two. Causality Engine takes a GA4 export and returns a read you can share by link; Creatify reads ad accounts. Paste figures into the agent's chat by hand if you want them side by side (integrations).

Does Creatify use my store's orders when it calls a winner?

Not on its documented pages. The creative testing page says the agent reads the ad platform's own results, and its Meta launch example picks the Purchase event on your pixel.

Is a 95% range not proof enough?

It is proof about the platform's purchase count, not about the store. A variant can beat its rivals on Meta's 7-day click count and still bring few orders Meta would not have had anyway; only a holdout or a causal read separates the two.