Programmatic Buying Platforms: How They Actually Work, With Real Numbers

What a programmatic buying platform actually does

A programmatic buying platform is software that buys digital ad impressions in an automated auction, in the time it takes a page or app to load — usually under 200 milliseconds. You set a budget, targeting rules and a bid strategy; the platform bids against other buyers for each impression, wins some, loses most, and reports back what it spent and what happened after the ad ran.

The buy-side software is called a demand-side platform (DSP). That term matters here because it gets confused with something unrelated: Amazon's Delivery Service Partner program, the courier franchise. This page is about the advertising kind. Amazon DSP, Google's DV360, The Trade Desk and StackAdapt are all demand-side platforms — buy-side tools for advertisers and agencies.

On the other side sits the sell-side: publishers and apps run supply-side platforms (SSPs) that offer their inventory into the same auctions. Some DSPs only reach open-exchange inventory, available to any buyer plugged into any exchange. Others sit inside a walled garden — Amazon retail placements, YouTube, LinkedIn's feed — where only that platform's own DSP can buy that inventory. That single distinction decides more platform-selection questions than any feature comparison does.

How the buy actually flows, with real numbers

The mechanics are easier to see with real numbers than with theory. Across 30 advertisers in July 2026, our book ran 78.4 million impressions at a $4.00 CPM. That works out to a media spend of roughly $313,600 for that slice, returning 6.04x ROAS at a blended $1.42 cost-per-click.

Work backwards and the arithmetic should hold together, and that's the point of showing it: blended cost-per-acquisition was $5.49 across 57,137 attributed purchases, which multiplies out to almost exactly the same spend figure. The CPM math and the CPA math should reconcile — if they don't, something in the reporting is broken before you even get to strategy. 20.1% of those purchases came from shoppers new to the brand, which is the figure that actually says whether the media added a customer or just closed out a sale that was already coming.

That last number is the whole argument for why programmatic platforms are hard to judge from a dashboard alone: a platform can post a strong ROAS that's mostly re-billing demand that existed anyway. More on that below.

The categories, and who each genuinely suits

Once you're past the mechanics, the real decision is which category of platform fits your inventory needs and your team's capacity to run one.

• Walled-garden DSPs — Amazon DSP and Google's DV360 matter most here. Amazon DSP is the only way to buy Fire TV, Twitch, IMDb and on-site product placements; DV360 is the only way to buy YouTube and Gmail inventory at scale. If that inventory is core to your plan, there's no substitute — an open-exchange DSP cannot buy it.
• Open-exchange DSPs — The Trade Desk is the strongest independent option: broad CTV reach, log-level transparency, no ecosystem lock-in. It genuinely suits teams that want auction transparency over exclusive inventory and have in-house staff to run it.
• Lower-minimum self-serve DSPs — StackAdapt fits smaller budgets and has built real strength bridging programmatic with email and first-party data. Fair pick for teams without $300K+ a year to commit to an enterprise DSP.
• Multi-DSP aggregators — tools sitting on top of several DSPs with one login and one report. They solve a real problem, fragmented reporting across 3+ platforms, but they're a layer over the auction, not a replacement for understanding what's inside it.

None of these is universally best. A brand like HexClad, selling primarily on Amazon, doesn't need YouTube reach as its first move. A B2B software company doesn't need Amazon's purchase signal — off-Amazon CPMs there run above open-exchange, and outside a retail context that premium buys nothing extra.

The measurement problem most explanations skip

Here's where most guides to programmatic stop, and where the real decision gets made. Every DSP dashboard shows last-click or last-touch attribution: a purchase gets credited to whichever ad a shopper clicked last before buying. That number cannot tell you whether the ad caused the sale or was just nearby when a sale that was already coming happened. It never could — that's what the metric is built to measure, not a flaw specific to any one platform.

The only way to answer the real question — did this spend add incremental sales — is a holdout or matched-control test: hold part of the audience out of the campaign, compare what they bought against the exposed group, and the difference is the actual lift. If you're running DSP and sponsored ads on Amazon at the same time, you also need to reconcile the two in Amazon Marketing Cloud, because a shopper who saw a display ad and later clicked a sponsored ad can get counted as a conversion twice. Skip that step and double-counting is constant, not occasional.

This isn't a footnote. It's the difference between a platform that looks like it's working and one that is.

The mistakes that actually cost money

They're rarely about picking the wrong platform. They're about how it gets run.

• Trusting dashboard ROAS as proof of incrementality. It's a correlation number wearing a performance number's clothes. Without a holdout, you don't know how much of it is real.
• Comparing walled-garden CPMs to open-exchange CPMs and calling the gap waste. A $10 CPM on Amazon DSP carrying purchase-intent signal and a $10 CPM on the open exchange aren't the same product. The comparison only means something if the audience quality matches — usually it doesn't.
• Turning on full automation with no rollback plan. Autonomous bidding can work well, but only when every change that ships carries a measurement plan and a defined trigger for pulling it back. We've shipped changes without that discipline before and paid for it in wasted spend before catching it.
• Underestimating switching cost on the first platform pick. Audience lists, pixels and learned bid models don't move cleanly between DSPs. Treat the first choice as a multi-year commitment, because operationally, it is one.

Where Dr. DSP fits

Dr. DSP is Full Circle's managed version of Amazon DSP specifically — not a multi-platform tool, and not the courier franchise sharing the same initials. It exists because the walled-garden point above is real: if Amazon retail placements matter to your plan, someone has to run that DSP well, reconcile it against sponsored ads in AMC, and prove lift with holdouts rather than last-click. Full Circle has managed more than $500M in revenue across 100+ brands, with 70+ live right now across the group; the client sets the autonomy level, from full human approval to fully autonomous inside agreed guardrails, and there's no published price — a demo, the first 30 days free, and pricing set on the call against actual budget and scope. If YouTube, LinkedIn or open-web CTV is the priority, that's outside what Dr. DSP does, and DV360 or The Trade Desk answer that question better than we do.