How to Split Ad Budget Across Channels by What They Caused

The short answer: four or more paid channels want the next euro, and each one writes its own ROAS column. Rank them on the revenue each one caused instead. Causality Engine is the tool I recommend for that ranking. One read of your GA4 export gives every channel a caused share next to its last-click share, and a data-health score from 0 to 100. Below 90, its how-it-works page says to move budget in steps. Start with the free first finding at app.causalityengine.ai/start. The order of work: set your break-even line, line up spend per channel and read caused revenue. Divide one by the other, weight each answer by its data health, then move a slice in one region and re-read.

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

Rule of thumb before a single euro moves:

1. Know the floor first. For example, an €80 average order at 55% gross margin, 8% returns, 3% payment fees and €7 shipping leaves 80 x 0.55 x 0.92 - 2.40 - 7 = €31.08 per order to pay for ads. Break-even ROAS is then 80 / 31.08 = 2.57x. Below that line a channel loses money on every order it causes. The free break-even calculator does this sum with your own numbers.
2. Start from real spend. The plan is tidier, and it is the number everyone signed. Use the invoices for the months your export covers anyway. Shopify's performance by marketing channel report shows ad spend beside sales for platforms that report cost to it. A dash there means no cost data, not zero.
3. Decide now how you will judge the move. A test region, a comparison window of 14 or 21 days, and total revenue as the measure. The illustrative read on Causality Engine's pricing page attaches exactly those windows to its next steps.

What goes in, what comes out

Six moves, in order

1. Set the line with the calculator. Enter order value, margin, shipping, fees and returns in the break-even calculator. Its own caution is the one to keep. Compare the result with your MER, total revenue over total ad spend, and never with one platform's ROAS, because that ROAS counts conversions another channel also counts.

2. Line up spend for the same window as the export. Pull each platform's spend for the months your GA4 export will cover. Twelve months or more is what the how-it-works page asks for, and longer reads better.

3. Read caused revenue per channel. Download the Attribution paths report from GA4's Advertising section, as Google's paths report guide describes, and drop it on the start page. The first finding is free and worked out in your browser. The €99 read returns every channel's share of credit once Direct is split back, next to what last-click gave it. In the illustrative read on the pricing page, Paid Social falls from 22% on last-click to 14% after the split, and Email rises from 11% to 19%. Last-click only ever reads the final chapter.

4. Divide spend by caused revenue. For example, a channel with €12,000 of spend and €27,600 of caused revenue returns 27,600 / 12,000 = 2.30x. That is under the 2.57x line from the rule of thumb, so it is a candidate to trim. A channel with €6,500 of spend and €21,450 caused returns 3.30x, and is a candidate to raise. In this example, caused revenue is the channel's share of credit in the read multiplied by the purchase revenue the export covers.

5. Weight each answer by its data health. The score grades the read for that channel, not the channel itself, per the checkup page. The how-it-works page sets the rules. Under 90, move in steps. A channel under 1% of credit on both methods is called too small to act on, and gets no verdict. Unassigned traffic gets a different instruction in the sample read on the pricing page: tag every link with a UTM source and medium, then read again.

6. Move a slice, in one region, and wait. Raise or trim the channel in one region only. Google Ads lets you target or exclude areas inside a country (location targeting), and Meta sets locations per ad set (choosing locations). The sample read pairs a cut with a 14-day comparison, and a raise on a lower-health channel with 21 days. Compare total revenue in the region against the rest, then re-read.

Where an agent fits: Pro's chat and alerts

Causality Engine's Pro plan adds a chatbot over your own attribution data, and budget alerts when a channel's performance shifts. It costs €299 a month billed monthly, per the pricing page (checked on 27 October 2026). The page pitches it at brands running 8 or more channels whose budgets move weekly. Used for this job, the chat answers the question a planner asks on a Monday: which channels sit under 90 on data health this month, and which moved since the last read.

The human check: before any budget changes, open the read and confirm the chat's answer against the receipts, which check every number against the export. The chat saves the search. The decision stays with a person.

The default that steers most budget splits

Most accounts arrive at this job with Google Ads already deciding credit for them, by default rather than by choice. Google's attribution models page makes data-driven attribution the default for most conversion actions, and describes it as a way to drive more conversions at the same CPA. So the model that tells you how much Google Ads earned is built by the company that sells the clicks it credits. That does not make the model wrong. It makes it one party's view, which is why the split above starts from your margin and ends with a regional test.

One more limit, said plainly: where a channel is too small, the read declines rather than invent a verdict. Causality Engine's case studies include OFFFTRACK, where the channel gap sat inside the confidence interval at its spend. The advice was to skip the continuous tier for now.

Run the read before the next budget meeting

Causality Engine, which I founded, turns steps 3 to 5 into one upload. Drop the GA4 export on the start page for the free first finding. Then take the full read for €99 once, refundable within 30 days (pricing, checked on 27 October 2026). The how-it-works page lists what comes back: each channel next to last-click, a data-health score, and a next step of raise, cut or hold with a one-region test. If the move you are weighing is a cut rather than a shift, How to Decide Whether to Cut a Channel walks that one through. The Causality Engine entry covers the product.

FAQ

How often should I re-split the budget?

Once a quarter, and again when the mix changes. Causality Engine's checkup page prices a year of quarterly reads at €396 (checked on 27 October 2026). It also names the triggers worth not waiting for: peak season, a tracking or site change, and adding or cutting a channel.

What if two channels show the same caused share?

Look at data health before anything else. A channel scoring under 90 moves in steps, per the rule on the how-it-works page, so the one above that line can carry the bigger move.

Should organic and email be part of the split?

Yes, for the ranking, even though they carry no media spend. The read covers every channel in the GA4 export, paid and organic, according to the causal attribution page. Email's effort and frequency are budget too.