Supermetrics With Causality Engine: Read the Union Blend Total

The short answer: a union blend that adds every ad platform's conversions into one row will usually show more purchases than the store shipped, and Supermetrics is doing its job when it does. Supermetrics, a reporting platform that connects ad, analytics and CRM data to Sheets, Looker Studio, warehouses and AI chats (entry, homepage), passes each platform's count through. In AI chats, Facebook Ads conversions use the window already set on the ad account or ad set, and "there is no Supermetrics-side default" (conversion window guide). The question of what each channel caused needs a different file: the GA4 Attribution paths export, read for free on Causality Engine's start page. Causality Engine estimates each channel's caused share from that one file (entry).

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 do before the next report goes out:

• Count orders, not value, once. Total purchases across platforms for a month and set the order count from the store beside it. Order counts make the double claim easy to see, since one order cannot be two.
• Start from the queries you have. Each Facebook Ads query has a Conversion window dropdown under Options in Sheets, Excel, Power BI and Query Manager, and under Edit connection in Data Studio (conversion window guide). Write down which window each query uses.
• Check the GA4 refresh. Supermetrics recommends a 12-day refresh window for GA4, so modelled conversions can settle (GA4 notes). A total read on day two of the month is still moving.

The two tools in one table

How Supermetrics carries each platform's count

Supermetrics' fix guide traces Facebook mismatches to the attribution changes Facebook made after iOS 14, and fixes them query by query. The setting has different names on each side: Supermetrics calls it a conversion window, Ads Manager calls it an attribution setting. At campaign level it shows "Multiple Attribution Settings" when the ad sets differ (conversion window guide).

Google Ads arrives differently. It reports one Conversions metric split by a Conversion type dimension, while Facebook Ads exposes each custom conversion as its own metric, so lining them up takes a custom field per conversion type (custom conversions guide). Once aligned, the union blend puts both in one column, which the guide calls "a unified table" ready for "aggregated KPIs".

That aggregated KPI is where the overcount sits. Each row is right. The sum is a total of claims.

GA4 through Supermetrics, and the file the read wants

The ROAS guide joins the ads union to GA4 by mapping GA4's source and medium to channel names with a lookup table, then divides GA4 revenue by cost. GA4's revenue there follows whatever the Data API returns, and Supermetrics says plainly that the API "does not appear to support" getting or setting the attribution model (GA4 notes).

Causality Engine does not take that table. It takes GA4's Key event attribution paths report, downloaded as a CSV from the Advertising section, with the paths that ended in purchase (Google's report help). Since the API exposes no attribution functionality, download it from GA4 itself, 12 months or more of it (how it works); the export walkthrough covers the clicks.

Worked example: the orders that were claimed twice

For example, using made-up counts that belong to no store. October, one Supermetrics union blend:

1,214 + 1,087 + 166 + 93 = 2,560 purchases claimed. The store shipped 1,902 orders. 2,560 - 1,902 = 658 orders were counted by more than one platform, or by a platform that only showed an ad. And email, organic search and Direct have not yet spoken for any of the 1,902.

Change only the Facebook query's Conversion window to 1-day click and suppose it returns 803. Then 1,214 - 803 = 411 of Facebook's purchases depend on the longer window or the view. That number is not waste. It is the part of Facebook's claim that its own rule decides.

Beside these, a causal read sets out each channel's caused share against last-click from the GA4 export, a data-health score from 0 to 100, and a next step to try in one region first (how it works). Where the read and Facebook's column disagree, a regional test settles it before budget moves.

A check that needs no new tool

Two passes in the sheet you already have. First, run each platform's purchases under its tightest click-only setting and again under the setting you report on; the gap per platform shows how much rests on the window. Second, by hand, since no holdout feature appears in these Supermetrics guides: pause one platform in one country for 14 days, keep a similar country running, and compare orders by country from the store. For example, 412 orders where it ran and 389 where it paused, after two weeks of matching volumes: 412 - 389 = 23 orders that the platform's presence made.

That answer stands without buying anything. For every channel in one pass, Causality Engine, which I founded, reads the GA4 export from the start page; the first finding stays in your browser, and the full read is €99 once, with a refund inside 30 days (pricing, checked on 11 November 2026).

Limits on both sides

The two do not connect. The integrations page says Causality Engine reads every channel the GA4 export records, so no Supermetrics destination feeds it. It reads channels, so a campaign or ad set question stays with the Supermetrics query. It reads the key events GA4 holds (one-off read), and where spend is too thin for a signal, the homepage says it declines. A channel under 1% of credit on both methods is called too small to act on (how it works).

Supermetrics does much the read never will: scheduled reports into Sheets and Looker Studio, and, per its homepage, asking Claude or ChatGPT what is working and acting on the answer in the same chat.

The pricing page lists Data Transformations, which covers data blends, in the Enterprise package (pricing). So the table that adds every seller's conversions into one figure is the feature that justifies the largest contract, and it adds them exactly as each seller reported. A bigger blended total does not make that contract harder to renew.

See also

• Supermetrics: packages, destinations and what each tier holds
• How to use Supermetrics: a reporting job from connection to sheet
• Supermetrics on the marketplace: Supermetrics' affiliate program terms
• Causality Engine: the read, its prices and where it declines
• How to use Causality Engine: the export, the read, the regional test
• Causality Engine on the marketplace: the affiliate terms, stated as the owner's
• Causality Engine vs Supermetrics: the two compared as purchases
• How to find what Direct traffic is hiding: where Direct's buyers came from

FAQ

Can a Supermetrics query produce the file Causality Engine reads?

Not by any route the guides describe. Supermetrics' GA4 notes say the Data API offers no attribution modelling, and the read wants the Attribution paths CSV downloaded from GA4. The two products have no link.

Which Facebook conversion window should the union blend use?

My recommendation: the one your Ads Manager columns use, written into the sheet, so the blend matches what the media buyer sees. Then run it once at 1-day click to see how much of the claim the window carries (conversion window guide).

Why do my GA4 totals in Supermetrics change for days afterwards?

Modelled conversions settle late. Supermetrics recommends refreshing the last 12 days, and its discrepancy guide adds that API data can lag the GA4 interface by 24 to 48 hours.