Is Amazon DSP Pricing Actually Fair?

There's been a debate going around about Amazon DSP take rates and whether they're as low as they look. I think the answer depends on what kind of advertiser you are.

If you sell on Amazon

You're already paying Amazon a lot to sell on their platform. Referral fees, FBA fees, storage fees. Then you add DSP on top. The take rates look low, but a lot of what you're buying is Amazon's own inventory, and on that they keep everything. So the cut Amazon is actually taking is higher than it looks.

Then there's the attribution problem. PPC is straightforward. Someone searched, clicked your ad, and bought. DSP is different. It shows ads to people who might buy later. So when DSP claims it drove a sale, you're right to ask: did that ad actually change their mind, or would they have bought anyway?

Amazon does give you the tools to answer that. AMC lets you see overlap between channels, path to conversion, incremental reach. You can compare users who saw DSP ads against users who didn't and measure whether DSP actually changed buying behaviour or just showed up alongside purchases that were already happening. You might find that part of the DSP is retargeting people Sponsored Ads already converted. Or that most of your budget is going to inventory that is not helping sales. Or that Sponsored Brands is doing more for your conversion rate than what it looks like based on ROAS alone. And that custom cleaned audiences can double the ROAS. Once you see that, the take rate question answers itself.

But most brands don't have access because their DSP is run by an agency on a separate account. And a lot of agencies claim they know AMC but don't really know how to use it. AI tools are making AMC much more accessible now. If your agency runs your DSP but you've never seen your own path-to-conversion data, that's worth asking about.

If you can measure conversion properly through AMC, the take rate debate becomes less important. You can see exactly what DSP is doing for you. Invest the time there. The conversation about what cut Amazon is actually taking only really matters when you can't measure outcomes — and with AMC, you can.

If you don't sell on Amazon

For you it's different. You don't have Amazon sales data to measure against, so you're evaluating DSP purely as a programmatic buying tool. The audience data, the reach, the cost. Tracking sales is possible but mostly like any other platform, the only issue is migrate the code. Here the question of what cut Amazon is actually taking matters more. You're comparing Amazon DSP head to head against The Trade Desk, DV360 and others. And you should be honest about what you're really paying across the whole buy, not just the take rate on the media plan.

The bigger question

Some people say this debate is pointless because Meta, YouTube and Spotify have always kept 100% of their ad revenue and nobody complains. That's true. But it doesn't mean brands shouldn't ask what they're actually paying.

For brands selling on Amazon, the take rate debate matters a lot less once you have real measurement in place through AMC. For everyone else, getting clarity on what you're really paying across the whole buy is probably the more useful conversation.

What would it take for you to feel confident your DSP spend is actually incremental?