Programmatic and RTB: What's the Actual Difference?

Programmatic vs RTB: what's actually different

Programmatic advertising is the umbrella term for buying media through software instead of a person calling a publisher and negotiating a rate by hand. A demand-side platform (DSP) sends a bid request, a supply-side platform (SSP) collects it from publishers, and an ad exchange matches the two. That's the whole category.

Real-time bidding (RTB) is one buying method inside that category, not a synonym for it. RTB is the live auction: every individual impression is priced and won in under 100 milliseconds, based on what each advertiser is willing to pay for that specific viewer at that specific moment. It's the mechanism most people picture when they hear "programmatic," because it's the most visible and most common one.

But programmatic also includes buying methods that never touch a live auction: private marketplace (PMP) deals, where a publisher reserves inventory for a shortlist of buyers at a negotiated floor price, and programmatic guaranteed (PG), where the price and volume are fixed in advance with no bidding at all. Same DSP/SSP infrastructure, same automation — just a fixed price instead of an auction.

How an RTB auction actually works, with real numbers

A user loads a page. That triggers a bid request containing whatever the publisher passes along — page context, rough location, device type, sometimes an audience segment. The request goes to an ad exchange, which fans it out to every DSP plugged into it. Each DSP decides, on its own logic, what that impression is worth to its advertisers and returns a bid. The highest bid above the publisher's floor price wins, the ad serves, and the buyer pays what the auction settled at. Total elapsed time: under 100 milliseconds, before the page has finished loading.

The number that matters to an advertiser isn't the auction mechanics — it's what a thousand of those auctions cost, and what they returned. Across 30 advertisers we managed in July 2026, the book delivered 6.04x return on ad spend, 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost-per-click. Run the arithmetic and it lands somewhere real: at a $4.00 CPM, 78.4 million impressions is roughly $313,600 in media before fees — a useful sanity check any time a vendor quotes you a CPM without volume, or volume without CPM. Either number alone tells you almost nothing.

The three ways programmatic inventory actually gets bought

RTB isn't the only door into programmatic inventory, and mixing the three up is where most confusion about "programmatic vs RTB" actually comes from.

• RTB — open auction, every impression priced live, no reservation.
• Private marketplace (PMP) — a publisher invites a specific list of buyers to bid on set-aside inventory, still an auction, but with a floor and a smaller pool.
• Programmatic guaranteed (PG) — fixed price, fixed volume, negotiated ahead of time, no bidding.

The mistake that makes RTB results look better than they are

The most common error in programmatic isn't in the buying — it's in reading the results afterward. Last-click attribution credits whichever touchpoint happened right before a purchase, and display is almost never the last touch. That doesn't mean display did nothing; it means last-click was never built to answer the question "did this impression cause this sale." It can't prove incrementality and never could — someone has to run the same offer against a holdout group who saw nothing, or a matched control, to find out what actually happened without the ad.

We've made this mistake too: crediting a view-through conversion to a shopper who already had the product in their cart from a search click twenty minutes earlier. The display impression didn't do anything there — the attribution system just handed it credit because it happened to sit in the path. The fix isn't more dashboards. It's reconciling DSP and sponsored ads in the same measurement layer — for Amazon, that's Amazon Marketing Cloud — so the two channels stop double-counting the same shopper and you can see whether display added incremental reach or just rode along on demand that search would have captured anyway.

What to check when the numbers look wrong

Bad numbers usually trace back to one of a handful of places. Before assuming the campaign failed, check these:

• CPM spiked suddenly — check whether you're now competing in a smaller pool (a PMP deal with fewer bidders drives price up fast) or whether a floor price changed on the publisher side.
• Win rate collapsed — check the bid cap first; it's the most common self-inflicted cause, followed by a pacing setting that's throttling delivery on purpose.
• Attribution looks too good — check for overlap with sponsored ads or search before crediting display. If the same purchase shows up in two channels' reporting, something is double-counted, not doubly effective.
• Frequency is high but conversions are flat — check the cap. An impression a shopper didn't convert on the third time rarely converts on the eighth either; that spend is usually better redirected.

None of these are exotic. They're the same four checks every time, in roughly that order.

Where a managed DSP fits into this

Everything above holds whether you run RTB yourself, through an agency, or through a managed product. Dr. DSP is Amazon DSP — Demand-Side Platform, Amazon's programmatic display, video and audio buying, not Delivery Service Partner, the courier franchise — run by Fable 5, from Full Circle, which has managed more than $500M in revenue across 100+ brands. The point of running it as a managed product isn't to hide the auction mechanics described above; it's to reconcile DSP against sponsored ads in Amazon Marketing Cloud on every campaign, so the return you see isn't inflated by the same double-counting problem described earlier on this page. There's no published price — a demo, the first 30 days free, and pricing set on the call — because that reconciliation work varies by how tangled the existing attribution already is.