Amazon DSP Has No Self-Serve Minimum Now: That Is Not The Same As Being Ready

For years the answer to "should we try DSP" was decided for you by a spend minimum you could not meet. That gate is gone, and the question is suddenly yours to answer. Which is harder, because access and readiness are not the same thing, and Amazon has no reason to tell you the difference.

We are seeing this land in inboxes as an opportunity, and for some accounts it genuinely is. For others it is a way to spend a retainer's worth of budget on an audience that was going to buy anyway. The useful work is telling those two situations apart before the money moves.

What actually changed

Amazon removed the spend minimum on self-serve DSP, announced at unBoxed in late 2025. Velocity Sellers, Marketplace Ad Pros and Canopy Management all report the same change in their 2026 guidance, which is enough agreement for us to treat it as settled.

Practically, that means an advertiser can open self-serve DSP through the Amazon Ads console and run display and streaming placements without committing to a floor first. You pay media cost plus a technology fee charged as a percentage of spend.

Managed service is a different product

The figure people still quote, around fifty thousand dollars a month, was never the price of DSP. It was the entry point for Amazon's managed service, where an Amazon Ads team runs the campaigns for you, and by every account we have seen that tier still works the way it always did.

Worth separating the two in your head before any conversation about DSP, because they are sold in the same breath and they are not remotely the same commitment. Self-serve means you or your agency operate the console. Managed means Amazon does, at a budget level most mid-size brands were never going to reach.

What DSP is for, and what it is not for

Sponsored Products meets demand that already exists. Someone types a query, and you compete for that moment. DSP works earlier and wider: it reaches people based on shopping behaviour rather than a search they just performed, and it can follow them off Amazon.

The strongest and least controversial use is retargeting, showing your product to people who viewed it and did not buy. The most common expensive mistake is treating DSP as a way to get more of what Sponsored Products already gives you. If search is working and you want more of it, the cheaper answer is usually more search, not a new channel with a technology fee attached.

The signal problem nobody mentions in the announcement

DSP optimises with a predictive model, and a predictive model needs conversions to learn from. This is the part that decides whether a small budget works, and it is rarely in the pitch.

Below a certain volume of conversions the model simply cannot distinguish a good audience from a bad one, so spend spreads across placements without ever concentrating on what works. The campaign is not broken. It has nothing to learn from. We are not going to print a threshold figure here, because the only source we found for one was a single agency blog and the number will vary by category and price point anyway. The principle holds regardless: too few conversions means the targeting stays generic, and generic display advertising is an expensive way to reach people.

What has to be true before DSP is worth it

These are the conditions we look for. They are not a scorecard, and an account can be ready without every one of them, but a brand that fails most of these will usually be disappointed.

1. The detail pages convert. DSP sends traffic to a page. If that page converts poorly, DSP buys you more evidence of that at a higher cost per visit. Fix the page first.
2. Search is already profitable and near its ceiling. If Sponsored Products still has profitable headroom, that headroom is cheaper than a new channel.
3. Stock will hold. Running awareness spend into a product that goes out of stock mid-flight wastes the whole flight.
4. There is enough conversion volume to teach the model. See above.
5. You can wait. DSP is measured over weeks. If the budget needs to prove itself in seven days, this is the wrong instrument.
6. Somebody owns it. Self-serve means self-operated. An unattended DSP account spends smoothly and reports politely while doing very little.

The cheaper checks we run first

Before recommending DSP to anyone, we look at three things that cost nothing to examine and frequently remove the need for it entirely.

First, the search term report, because most accounts are funding queries that will never convert. That is money already inside the budget, and reclaiming it is faster than opening a channel. Our guide on clicks that produce no sales covers the triage, and finding the bleeders covers what to do with them.

Second, conversion rate on the pages the ads point at, because ad spend multiplies whatever the page already does. Nine things to check before you touch PPC is the sequence we use.

Third, total advertising cost of sales rather than ACOS on individual campaigns, because a channel can look efficient while the account overall gets worse. TACOS is the number that tells you whether advertising is growing the business or just moving sales from organic to paid.

How to read DSP reporting without fooling yourself

The reporting question matters more here than in search, because display attribution is generous by nature. An impression someone scrolled past can be credited with a sale they were always going to make, and retargeting is especially prone to this: you are advertising to people who already showed intent, so some of what it claims would have happened anyway.

The honest way to read it is incrementally. Look at what total sales and TACOS do across the account while DSP runs, not just at what DSP reports for itself. If DSP shows strong returns and account level TACOS is flat or worse, the channel is probably claiming credit rather than creating demand.

What we would do first if this were our account

1. Pull the search term report and cut the queries that have spent real money with no sales. Do this whether or not DSP is on the table.
2. Check conversion rate on the products you would advertise. If it is weak, that is the project, not DSP.
3. Write down the current account level TACOS. Without a baseline you cannot tell later whether DSP helped.
4. If you still want to test, start with retargeting only. It is the use case with the clearest logic and the smallest surface for waste.
5. Give it a defined budget and a defined window, and agree in advance what result would make you stop.
6. Judge it on what the account did, not on what the channel reported about itself.

If you want a second opinion on whether the account is at that ceiling yet, that is exactly what our DSP management service starts with, and the audit costs nothing.

Related guides

• The only ad metric that shows if you are actually growing
• Ads getting clicks but no sales, how to triage
• How to find the bleeders in your campaigns

Common questions about Amazon DSP