By Ted Gibson — RippleEffect
Amazon DSP is worth it when you already have more demand than sponsored ads can catch: enough people looking at your products to be worth retargeting, and enough margin to fund prospecting that will not pay for itself this month. It is not worth it as a rescue. If the listing does not convert, or the sponsored account has never been split between defending your own name and buying new demand, DSP will spend faster and teach you less. One clarification before the numbers, because the search results make it necessary: this is Amazon's demand-side platform, the programmatic advertising product. Amazon also runs a Delivery Service Partner programme for van fleets, which shares the initials and owns most of page one for this question.
Nobody agrees what the minimum is, and Amazon's own figure is the one nobody quotes
Amazon's advertising site says the managed-service option typically requires a minimum spend of USD $50,000, and that the minimum may vary per country. That is the only primary source in this whole conversation, and the agency pages answering the question rarely cite it. Skale Strategy gives Amazon-managed as $35,000 to $50,000 a month, says some agencies require $10,000 to $15,000, and starts its own service as low as $3,000. Darkroom gives the same $35,000 to $50,000 for Amazon-managed and $10,000 to $15,000 agency-managed. Canopy Management says $50,000, and that Amazon removed the self-service floor at unBoxed in November 2025, leaving practical starting budgets of $5,000 to $10,000. Seller Labs says around $50,000 and tells you to confirm current minimums with Amazon, which is the most useful sentence any of them writes.
Published minimums for the same product therefore run from $3,000 to $50,000 a month, a spread of more than sixteen times, and they are not describing the same thing: some mean what Amazon requires, some mean what that agency requires, and the pages rarely say which until well down the article. Two things follow for a UK brand. Amazon's "may vary per country" means the dollar figure is not automatically yours, and we could not find a sterling minimum published anywhere. And Canopy's November 2025 removal of the self-service floor could not be confirmed on any Amazon-owned page we could reach, so it is worth asking your account team directly rather than budgeting against a blog post, ours included.
What it costs on top of the media
Management is charged separately from spend. Skale puts the typical agency fee at 8% to 15% of ad spend and its own at 8%; Darkroom gives 10% to 15%, or a flat $3,000 to $8,000 a month. On a £10,000 monthly DSP budget that is £800 to £1,500 of fee; on a $10,000 budget, $800 to $1,500. The media itself is bought on impressions, and the format decides the bill far more than the audience does. Skale's ranges run $3 to $8 per thousand impressions for display on Amazon, $5 to $12 off Amazon, $4 to $10 for retargeting, $12 to $25 for video and $20 to $40 for connected TV; Darkroom compresses the same picture into $3.00 to $15.00. On those ranges connected TV costs roughly five to seven times display on Amazon, so adding it is a budget decision before it is a creative one. What we charge for the work sits in our published rate card, and DSP is inside it where it is warranted rather than sold as an upgrade.
Two jobs, and only one of them brings new money
DSP does retargeting and it does prospecting, and blending them into one return figure hides everything worth knowing. Seller Labs treats 3x to 6x as a reasonable initial target for retargeting and says prospecting will often land at 1x to 3x, with new-to-brand running at 60% to 80% of orders. Darkroom reports 1.5x to 4x measured last-touch and argues the effect on total Amazon revenue is often 20% to 40% higher than that once the halo is counted. Retargeting flatters the blend, because it is advertising at people who already found you. On the collagen brand Ancient + Brave, DSP is where the step change came from, with custom audiences bringing in new customers at a positive return rather than showing the same buyers the same product more often. That distinction is the whole question, and a DSP report that cannot separate the two is not telling you anything you can act on.
When it is not worth it
Three cases, in order of how often we see them. First, the listing is the constraint: sending programmatic traffic to a page that does not convert buys expensive proof of a problem you could have found for nothing, and what actually moves conversion is a cheaper place to start. Second, sponsored advertising is still undivided. If you cannot yet say what share of your spend defends your own brand name, that split is measurable in a fortnight and usually moves more money than DSP will. Third, the timescale does not fit. Darkroom puts stabilisation at six to eight weeks; Seller Labs describes a 14-day unified attribution window covering both clicks and view-throughs, with Amazon moving to shopping-signal enhanced last-touch attribution from 1 January 2026. A four-week trial read on last-click is not a trial, it is a way of proving whatever you already believed. On readiness, Seller Labs puts the economics at $1m or more of annual Amazon revenue and Canopy puts the entry point at $40,000 to $75,000 of monthly total ad spend; both are agency judgements rather than Amazon's, and both are dollar figures with no published sterling equivalent.
The question that settles it is not whether DSP works. It does, at the right size. Ask instead which of the two jobs your budget is doing, what the new-to-brand share is expected to be, and which number will tell you in eight weeks whether it worked. If the answer is a blended return on ad spend, the budget is buying reach and calling it performance. Working out whether an account is at that point is one of the first things a free account audit answers, and the honest answer is often not yet.