Meta → Amazon Halo Effect Calculator

In marketing, the halo effect is when an ad in one place drives sales somewhere else. A TV ad lifts store sales. A Facebook ad sends people to search for you on Amazon. The ad gets no credit for those sales, because nothing links the shopper back to it.

For Amazon brands running Meta ads, the halo is the gap between two numbers: the Amazon sales your ads caused, measured with a holdout test, and the Amazon sales Amazon Attribution credits to them. If the ads caused more than Attribution sees, the difference is halo. It's real money your reports leave out, and it can make good ads look like they lose money.

Amazon Attribution counts purchases within 14 days of a click on a tagged link. Halo sales come from every other path:

• Branded search. Someone sees your ad, doesn't click, and later searches your brand name on Amazon.
• Direct navigation. They open the Amazon app, find you in their history or cart, and buy.
• Another device. They see the ad on their phone and buy on a laptop, where the tag never followed them.
• After the window. They buy on day 20, after Attribution has stopped counting.
• Organic rank. More sales can help your listing rank in Amazon search, which brings more sales from shoppers who never saw an ad. Amazon doesn't publish exactly how ranking works, so this part is the hardest to pin down.

halo sales      = incremental Amazon sales − Amazon Attribution sales
halo multiplier = incremental Amazon sales ÷ Amazon Attribution sales
true ROAS       = incremental Amazon sales ÷ Meta spend
attributed ROAS = Amazon Attribution sales ÷ Meta spend

• Incremental Amazon sales: what your Meta ads caused on Amazon, from a holdout test. The incremental ROAS calculator works it out from test results.
• Amazon Attribution sales: product sales credited to the same campaigns' tagged links, for the same dates.
• Meta spend: what you spent on those campaigns over the same dates.

A made-up brand spends $5,000 on Meta in a month. These are the calculator's example values.

1. Amazon Attribution credits $3,500 of Amazon sales: an attributed ROAS of 0.70.
2. A holdout test measured $8,000 of Amazon sales the ads caused: a true ROAS of 1.60.
3. Halo sales = $8,000 − $3,500 = $4,500. That's 56.3% of what the ads caused.
4. Halo multiplier = $8,000 ÷ $3,500 = 2.29×.

On Attribution alone, the brand would see a ROAS of 0.70 and probably cut the ads. Counting the halo, they return 1.60. Once a brand has measured its multiplier, it can use it to read future Attribution reports: credited sales × 2.29 is a rough estimate of caused sales, until the next test.

There's no target, because the multiplier describes your reports, not your ads. Here's how to read it:

• Above 1: Attribution under-credits your ads. The bigger the number, the more you'd undervalue them on Attribution alone.
• About 1: Attribution is close to what your ads cause, at least for now.
• Below 1: Attribution credits sales the ads didn't cause. Some tagged shoppers would have bought anyway, often people who already knew your brand.

Multipliers vary by brand, category, price and how much your ads build awareness rather than drive clicks. They also drift as your ads and season change, so one measured last spring may not hold this winter.

• Guessing the halo. A halo multiplier from someone else's brand tells you little about yours. Measure it.
• Mixing dates. Use the same dates for the test and the Attribution report, or the gap measures the calendar.
• Counting all Amazon sales as halo. Sales rise for many reasons. Only a holdout separates what the ads caused from everything else.
• Forgetting untagged ads. If some Meta ads link to Amazon without Attribution tags, their sales all land in the halo.

The halo only exists once you've measured incremental sales. Plan a test that can see it with the holdout test calculator, and compare all your reports at once with the attribution gap calculator. For setup, tagging and the rest of the picture, read Facebook ads for Amazon.

Spillmark runs the holdout test, measures the halo with an uncertainty range, and keeps your multiplier up to date as each new test finishes. See how lift is calculated.