Two Minute Reports With Causality Engine: Test the Blended Column

The short answer: a blended Conversions column in Google Sheets can total more orders than Shopify recorded, and the sheet will not say so. Two Minute Reports, the connector that pulls ad, analytics and store data into Google Sheets and Google Data Studio, documents that a blended custom metric takes one field from each connector and adds the values "into one total" (blending article). Its Facebook Ads connector also lets each query pick Meta's attribution window, from Default to 1 d click, without touching the settings in Meta (Facebook Ads connector). The column is three platforms' rules added up. For a reading built on the converting paths in the store's GA4 property instead, Causality Engine offers a first finding at no cost on its 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.

Three cells to fill before trusting the column:

1. The overshoot. For example, 268 Meta Website purchases, 233 Google Ads conversions and 58 TikTok conversions blend into 268 + 233 + 58 = 559. If the Shopify query returns 476 Orders, the column runs 559 - 476 = 83 ahead of the store.
2. The query you already run. Open the Facebook Ads query's Report Config and look at Conversion/Attribution Window(s). Default means Meta's default window (Facebook Ads connector).
3. A second window beside it. Select Default and 1 d click together and every conversion metric gets one column per window, such as Website purchases (1 d click) (Facebook Ads connector).

The sheet and the read, fact by fact

Meta's window, chosen query by query

Two Minute Reports leaves the attribution decision with Meta unless you take it. The Facebook Ads connector page says conversion metrics such as Website purchases use an attribution window that tells Meta which conversions to count: those within some days of a click, of a view, or both. Default uses Meta's default. To remove view-through conversions, choose a window with only a click part: 1 d click, 7 d click or 28 d click.

Three details from the same page matter for a client sheet:

• The window lives in the query, not in Meta. Changing it alters the numbers in that query alone.
• Several windows mean several columns. The Default column keeps the plain name; the others carry the window in brackets.
• Windows marked (deprecated) should not be used.

The window applies to Video views (3s) too, so a click-only query can show a small fraction of the views Ads Manager shows. That is the rule working, not a broken connection.

What a blended field adds up

Blending exists because every platform names spend differently. The blending article maps Meta's amount spent, Google's cost and TikTok's spend into one Spend field, and for any custom metric the values from each connector are added together. Spend blends cleanly; every currency unit is paid once.

Conversions do not. Map Website purchases, Conversions and TikTok's conversions into one field and an order that all three platforms claim appears three times. The article warns about a quieter version of the same thing: group a report by name alone and items sharing a name merge into one row with their metrics added, and "the report does not show that this occurred". Its fix, grouping by name and ID, solves the naming problem. It cannot solve overlapping claims, which are not a naming problem at all.

The store's figures in the same workbook

The Shopify connector gives the other side. Orders counts each order once, needs at least one product in it, and keeps refunded and cancelled orders, which show instead in Returned quantity and Refund amount. Two Minute Reports keeps a copy of the Shopify data refreshed every hour, so the last hour can be missing. Sales fields and payment fields cannot share a query, because the two lists do not align and the totals would multiply.

GA4 has a trap of its own. If the Google account that authorised the connection carries a No revenue metrics restriction, Purchase revenue comes back as 0 while sessions and key events still match (GA4 connector). A zero there is a permission, not a bad month.

Worked example: one client, one month, three windows

For example, with invented numbers that describe no real store: an October sheet for a footwear client, built from three ad queries and one Shopify query.

With Default, the blend overshoots the store by 559 - 476 = 83. With 7 d click it overshoots by 494 - 476 = 18. With 1 d click it falls 476 - 462 = 14 short. Meta's own column moves by 268 - 171 = 97 purchases on the choice of window alone.

The 1 d click sheet looks reconciled. It is not. Google and TikTok still claim orders by their own rules, Meta now claims fewer by a stricter one, and the near-match is an accident of which rule was picked. Choosing the window that makes the sum fit is not a method.

What a causal read puts beside these columns is described on its how-it-works page: a caused share for every channel set against a last-click count rebuilt from the property's converting paths, a health grade between 0 and 100, and a suggested move to try first in one region. If it disagrees with the blended column, test before the client's budget shifts.

The free check, in four cells

All of this runs in the sheet the client already gets:

1. Run the Facebook Ads query with Default and 1 d click selected together.
2. Next to the blended column, divide it by Shopify Orders in a plain spreadsheet formula. Above 1.0, the platforms overlap.
3. Subtract the 1 d click column from the Default column. The remainder is what Meta credits to views and to buyers who took longer than a day.
4. Keep both ratios on the summary tab, month by month.

That check is free and stands on its own: it tells you the claims overlap and how much of one platform's figure rests on its window. What it cannot tell you is which channel the store would miss.

Where each tool stops

Two Minute Reports describes blending as the thing that makes cross-channel dashboards clean, and one Conversions column is clean; nothing in the field builder marks that the column adds three platforms' separate rules together, so the tidiness a subscriber pays for also hides the overlap.

The read has edges of its own. It needs a year or more of GA4 data and reads by channel, so a single TikTok campaign or a creative test stays a question for the platform (how it works). Small channels, under 1% of credit by both methods, are flagged too small to act on. Its pricing page for agents also states that a platform's reported ROAS never feeds the estimate and that nothing has to be migrated first. The sheet stays; the read sits beside it.

Next, export the Attribution paths CSV from the client's GA4 property and drop it on the start page where a first finding is produced free, in the browser tab itself. Causality Engine, which I founded, asks €99 for one full read, paid once, excluding VAT, with 30 days to ask for the money back (pricing, checked on 11 November 2026).

See also

• Two Minute Reports: connectors, plans and the Hub
• How to use Two Minute Reports: one Google Sheet per client
• Two Minute Reports on the marketplace: Two Minute Reports's affiliate program terms
• Causality Engine: a causal read and its price
• How to use Causality Engine: the export, the read and a regional test
• Causality Engine on the marketplace: how referrals of the read are paid
• Two Minute Reports vs Supermetrics: two ways to fill a sheet
• How to export GA4 data for attribution: the CSV the read needs

FAQ

Can Two Minute Reports push data into Causality Engine?

No. The two are not connected in any direction. Causality Engine reads a CSV downloaded from GA4's Attribution paths report, and a GA4 query in Two Minute Reports, however detailed, is a different shape of data (integrations page).

Does changing the window in the query change my Meta campaigns?

No. The Facebook Ads connector page says the window applies only to the data in the query, and the attribution settings of your ads in Meta stay as they are.

Should the blended column use Default or a click-only window?

My recommendation: neither alone. Keep Default so the figure matches Ads Manager, and print the 1 d click column beside it so the reader sees how much of Meta's number depends on the rule chosen.