Ad Budget Allocator: Meta vs. Amazon Ads

Budget allocation is deciding how much of your ad money goes to each channel. For an Amazon brand, the usual choice is between Meta ads, which reach people who aren't searching yet, and Amazon Ads, which reach shoppers already on Amazon.

The common approach is to give more to whichever channel shows the best ROAS. That goes wrong in two ways. Each platform's ROAS counts sales that would have happened anyway, and they don't over-count by the same amount. And ROAS is an average: it says nothing about what the next dollar will do. A good split uses incremental ROAS and accounts for diminishing returns.

The best split is the one where the next dollar returns the same on both channels. If Amazon Ads returned more for the next dollar, you'd move a dollar there, and keep moving until the two are equal.

incremental sales on a channel = k × spend^b
k   = measured iROAS × current spend^(1 − b)
next-dollar return = b × k × spend^(b − 1)
best split: each channel's share ∝ k^(1 ÷ (1 − b))

• iROAS: sales a channel causes per dollar, at today's spend, measured with a holdout test. The incremental ROAS calculator works it out from test results.
• b: how fast returns fall as spend grows. At 0.6, doubling spend raises sales by about 52%, not 100%. The marginal ROAS calculator estimates b from your own months.
• Cap: no channel gets more than 80%. The curve is least reliable far from what you spend today, so big moves should happen in steps.

A made-up brand spends $5,000 a month on Meta and $5,000 on Amazon Ads. Holdout tests measured an iROAS of 1.60 on Meta and 2.40 on Amazon Ads. It plans to keep a $10,000 total budget, with b = 0.6. These are the calculator's example values.

1. Today's split brings in $8,000 + $12,000 = $20,000 of incremental sales.
2. The best split is $2,663 on Meta (26.6%) and $7,337 on Amazon Ads (73.4%).
3. At that split, the next dollar returns 1.24 on both channels, so there's nothing left to gain by moving more.
4. Expected incremental sales rise to $20,587, about $587 more a month for the same budget.

Meta doesn't go to zero, even with the lower iROAS. Its first dollars still return more than Amazon Ads' last ones. That's diminishing returns at work.

There's no standard ratio between Meta and Amazon Ads. The right split is whatever makes the next dollar return the same on both, and that depends on your measured numbers:

• Next dollar returns about the same on both: your split is close to the best one already.
• One channel returns clearly more: move budget toward it in steps, then re-measure.
• Both return less than your break-even ROAS: the split matters less than the total. Spend less overall. The break-even ROAS calculator shows your line.

• Splitting by platform-reported ROAS. Amazon Ads reports often credit sales from shoppers already searching for your brand. Meta can't see Amazon sales at all. Neither report shows what the ads cause.
• Moving everything to the "winner". With diminishing returns, the best channel's next dollars return less and less. A split almost always beats all-in.
• Jumping straight to the new split. Move part of the way, then measure again. The curve is an estimate.
• Forgetting Meta's effect on Amazon. Meta ads drive Amazon searches and sales that Meta's report never shows. Leave them out and Meta looks worse than it is.

This allocator is only as good as the iROAS you put in. Measure each channel with a holdout test: pause Meta ads in some regions to measure Meta, and use a pause or a lower-spend test for Amazon Ads. Plan the test with the holdout test calculator.

Spillmark runs holdout tests for Meta ads that drive Amazon sales, records each one, and scores its budget recommendations against what happens next. See how lift is calculated.