Amazon DSP Self-Service vs Managed: Real Costs, Real Control (2026)

What "self-service" actually means on Amazon DSP

The phrase suggests a sign-up form and a credit card, the way Sponsored Products works. Amazon DSP self-service is not that. It is a seat on a professional demand-side platform — the same console agencies use — with your team's hands on the controls: audience construction, inventory selection, bid strategy, frequency management, brand-safety exclusions, creative rotation and reporting all belong to you.

Amazon's own framing is that self-service customers are in full control of their campaigns, while the managed-service option adds Amazon's advisory team and is described as typically requiring a minimum investment of USD 50,000, a figure Amazon notes varies by country. What Amazon does not publish is a universal self-service entry bar. Access is arranged through Amazon Ads, and the expectations — spend commitment, operating capability, sometimes category — are agreed case by case and differ by region. That is why most self-service seats in practice are held by agencies and larger brands, and why smaller advertisers usually reach DSP inventory through an agency's seat rather than their own.

One disambiguation before going further, because the search results mix them: Amazon DSP here is the demand-side platform for programmatic display, video and audio. It is not the Delivery Service Partner courier programme. Same abbreviation, unrelated business.

Who a self-service seat actually qualifies: a checklist, not a form

There is no public qualification page to quote, so here is the practical version — the conditions under which a seat makes sense and tends to be granted, based on operating them ourselves.

• Spend that justifies the access. A seat carrying a few thousand dollars a month of media is not worth the setup on either side. Sustained five-figure monthly programmatic budgets are where the conversation usually starts; below that, an agency seat pools you with others.
• A trader, on payroll, permanently. Not a media generalist who also runs paid social on Tuesdays. Someone who has operated a DSP — this one or another — and can read a log-level anomaly, structure a frequency cap, and know when a deal ID is underdelivering. If this person does not exist yet, the seat is a hiring project wearing a media project's clothes.
• Creative production on tap. DSP inventory spans display sizes, online video and streaming TV. A seat with one banner size and no video runs a fraction of the platform you are paying to access.
• Cover. One trader is a single point of failure. Campaigns run through holidays, illness and resignations; media keeps spending either way.

What managed actually buys — and the two very different sellers of it

"Managed" bundles two distinct offers that get compared as if they were one.

Amazon's managed service. Amazon's own team plans and runs the campaigns, with the stated typical minimum above. You get Amazon's people and Amazon's roadmap access; you accept a service designed to scale across thousands of advertisers, which means process over bespoke attention at the margin. For advertisers clearing the minimum comfortably who want the platform owner's own hands, it is a legitimate choice we would not argue anyone out of.

An agency holding seats. The common mid-market route. The agency owns the console access, the trader and the process; you own the strategy conversation and the results review. Floors vary widely by agency, and so does what sits inside the fee — which is exactly where managed arrangements go wrong. The failure mode is not the existence of a fee; it is a retainer that bills whether or not work happened, creative charges bolted on after signature, and accounts left on autopilot behind a monthly deck of charts. Those are the things to interrogate, and they are all discoverable with direct questions before you sign. Our fuller scoring method for that conversation is in the Amazon DSP agency comparison.

The hidden costs of self-service, priced honestly

The seat itself is the cheap part. The costs that decide the comparison sit around it.

• The salary you are not counting. A capable programmatic trader is specialist headcount, employed whether campaigns run well or badly, plus the recruitment cost of finding one in a market where they are scarce.
• The ramp. A trader new to Amazon DSP specifically — even one fluent in other DSPs — spends their first months learning Amazon's audience taxonomy, its retail signals and its measurement quirks on your media budget. Tuition is real money; it is just invoiced as underperformance instead of a line item.
• Mistakes at media prices. An unbounded conquesting audience that quietly buys your existing customers back, a frequency cap left at default, a viewability floor never set — each is invisible in a topline dashboard and each burns spend for months. The wasted-spend figures at the top of this page are search-side, but the lesson transfers: waste hides inside averages until someone goes looking term by term, placement by placement.
• Measurement you now own. Clean-room analysis in Amazon Marketing Cloud, holdout design, deduplicating DSP against sponsored ads — on a self-service seat there is no one else to do this, and skipping it means grading your own homework with the answer key missing.

The hidden costs of managed, priced just as honestly

Symmetry matters, so here is the other column.

• The fee you can see — and the ones you sometimes cannot. The stated rate is rarely the problem; the additions are. Ask what creative production costs, what clean-room work costs, and whether either is inside the number you were quoted. A retainer with unpriced extras is the most expensive "simple" arrangement in this market.
• Autopilot risk. A managed account can coast: same audiences, same exclusion list, same deck with new dates. You are paying for judgement, and judgement leaves evidence — changed campaign structures, killed placements, test results. No evidence, no judgement.
• Dependency and exit. The audiences, structures and query logic live in someone else's account unless your contract says otherwise. Write the export terms before signing, not during the breakup.
• Skill you never build. Every managed year is a year your team does not learn the platform. That is a fine trade if programmatic is not core to you, and a bad one if you intend to hold the seat eventually — in which case say so, and structure the engagement as a handover from day one.

Check your fit: six questions, one recommendation

Answer the six below and the checker scores which route your situation actually supports. It runs on your answers only — nothing is sent anywhere.

When each route genuinely fits

Self-service fits when the trader already exists, the media budget is sustained rather than seasonal, creative production is solved, and someone owns measurement. Under those conditions it is the cheapest route per dollar of media and the only one where platform skill compounds inside your own walls.

Amazon's managed service fits when you clear its stated minimum comfortably, want the platform owner's own team, and can work within a standardised service model.

An agency seat fits most brands in between: real budgets, no in-house trader, and a need for the audience architecture, inventory discipline and measurement design to exist on day one rather than after a hiring cycle. The route decides most of the cost before any company is chosen — the fuller map of all four routes in, including packaged products, is in our Amazon DSP partner comparison.

A hybrid deserves more attention than it gets. Start managed, contract for portability, and treat the engagement as a two-year handover. Brands that say this out loud at signature get better-structured accounts than brands that spring it at renewal.

Where reMKTR sits on this map

We are the managed side of this comparison — an agency holding its own seats, focused on Amazon DSP and nothing else. reMKTR is the DSP arm of Full Circle, a full-service Amazon management company with $500M+ in managed spend across 100+ brands.

Across 30 advertisers in July 2026 — the scope stated because a figure without one is not evidence — the book returned 6.04x on ad spend, ran 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost per click, and delivered $5.49 cost per acquisition across 57,137 attributed purchases, 20.1% of them from shoppers new to the brand.

And the disclosure that keeps this page honest: if the checker above told you to buy the seat, buy the seat. A brand with a strong trader in-house is not our client, and pretending otherwise would be the kind of selling this comparison exists to help you see through. Where we earn the fee is everywhere the seat alone does not help — audience architecture, inventory discipline, and measurement design in Amazon Marketing Cloud, where holdouts and matched controls replace last-click guesswork.