Every impression is sold in a live auction lasting about a tenth of a second. A study of $14.4 billion in programmatic spend across 86 advertisers found that roughly 47% of it went to low-quality or fraudulent inventory. Cheap CPMs are how that happens.
Programmatic advertising is the automated buying and selling of digital ad space through real-time auctions. Instead of negotiating with individual publishers, an advertiser sets targeting and bidding rules in a demand-side platform, and software decides in milliseconds which individual impressions to buy and how much to pay for each one.
It now accounts for roughly 91% of all US display advertising spend, which means that if you are running display, native, video, audio or connected TV advertising, you are almost certainly buying it programmatically whether or not anyone described it that way.
The mechanic is genuinely remarkable. Somebody loads a web page. Before that page finishes rendering, the publisher sends a bid request describing the available ad slot and what is known about the visitor. A dozen demand-side platforms evaluate it against thousands of active campaigns, decide whether this specific person is worth reaching, and return bids. An auction resolves, a winner is chosen, and the ad loads. All of it happens in roughly a tenth of a second, several billion times a day.
Programmatic advertising companies exist because that automation is what makes it work, and it is also why it goes wrong. No human approves each purchase, so the quality of what you buy is entirely determined by the rules you set beforehand and the inventory you allow yourself to bid on. Set those badly and the system will efficiently, tirelessly, and at enormous scale buy you impressions that no human being ever saw.
This is a single real-time bidding auction, played at whatever speed you can follow. In the real world the whole sequence completes before a page finishes loading, which is exactly why almost nobody in advertising has ever actually watched one happen.
Most bidders pass. Of eight platforms asked, typically only a handful bid at all, because the rest judged this particular visitor not worth reaching. That filtering is the entire value of programmatic and the reason it beats buying a fixed placement.
The winner rarely pays their full bid. Most exchanges settle at or near second price, so the winner pays roughly what it took to beat the runner-up. Bidding higher buys you the right to win, not the obligation to pay more.
Nobody looked at the page. No human approved this purchase, checked what content the ad appeared beside, or confirmed a person was there to see it. Every one of those safeguards has to be configured in advance, which is what separates a managed programmatic campaign from an expensive random number generator.
Open exchange inventory can be bought for a few dollars per thousand impressions. Private marketplace inventory costs several times more. Buyers reliably choose the cheap option, and reliably get worse results, because a CPM only describes what you paid — not what you received.
Two things stand between an impression you bought and a human being who saw it. Viewability: whether the ad was actually on screen long enough to count. Invalid traffic: whether there was a person there at all rather than a bot. Multiply your CPM by both and you get the number that actually matters.
The comparison column models a private marketplace buy at $9 CPM with 78% viewability and 4% invalid traffic, which sits inside published 2026 ranges. Your figures on the left are whatever you set. Viewability and IVT benchmarks are drawn from 2026 reporting including IAS, Comscore and industry compilations; treat them as directional.
A $3 CPM at 55% viewability and 16% invalid traffic and a $9 CPM at 78% viewability and 4% invalid traffic are far closer in real cost than the sticker prices suggest, and on many days the expensive one wins outright. Add the finding that verified, fraud-free inventory converts dramatically better than unverified inventory and the argument stops being close at all.
This is the single most common structural mistake we correct as a programmatic ad agency taking over an inherited account. The account looks efficient on a CPM report and is quietly buying almost nothing of value. It is also why we run conversion optimization alongside media buying — there is no point improving impression quality if the page they land on cannot convert the humans who do arrive.
Programmatic advertising companies offer four ways to buy, and the difference between them is mostly about how much control you trade for how much reach. Most advertisers default to the open exchange because it is the easiest to start on, which is not the same as it being right.
| Method | How it works | Typical CPM |
|---|---|---|
| Open exchange | Anyone can bid on anything. Maximum reach, maximum scale, and the least control over what you end up next to. | $1–$4 |
| Private marketplace | Invitation-only auction on a defined set of publishers. Substantially lower fraud exposure, meaningfully higher viewability. | $5–$15 |
| Preferred deal | Fixed price, first look at inventory before it reaches the open auction, no obligation to buy. | Negotiated |
| Programmatic guaranteed | Fixed price, fixed volume, committed both ways. Closest thing to a traditional insertion order. | Negotiated |
| Curated marketplaces | Newer middle ground — pre-vetted supply pools with transparent pricing and no per-publisher negotiation. Reported to have crossed 11% of programmatic in 2026. | Varies |
CPM ranges compiled from 2026 programmatic benchmark reporting. Actual pricing depends on format, geography, targeting depth and how many intermediaries sit in the supply path.
| Metric | What it means | 2026 figure |
|---|---|---|
| Display viewability | Half the pixels on screen for at least one second, per the IAB and MRC standard. Above 70% is good, above 80% excellent. | 72.4% |
| Video viewability | Consistently ahead of display, and the gap has been widening. | ~79% |
| Desktop banner | The weakest common format. Leaderboards are worse still, reported near 56%. | 64% |
| Native inventory | Notably stronger, which is part of why native spend is growing faster than any other display format. | 81% |
| CTV display | Effectively unskippable on a screen nobody scrolls past. See CTV advertising. | 96% |
| Open exchange IVT | Invalid traffic. Bots, device spoofing and fabricated inventory. | 14–18% |
| Private marketplace IVT | The single clearest argument for paying more per thousand. | 3–5% |
| Display conversion rate | Low by design. Display is a reach and assist channel, not a last-click one. | 0.71% |
| Retargeting conversion | Double the display average, which is why retargeting usually earns the first dollar. | 1.42% |
Compiled from 2026 reporting including the IAS Media Quality Report, Comscore benchmarks, Google Active View data and industry statistical compilations. Methodologies differ between sources, so use these as neighborhoods rather than targets.
Format choice moves viewability more than bidding does. A leaderboard at 56% and a half-page unit at 82% are the same auction, the same targeting and the same money, producing radically different amounts of actual exposure. Excluding the weakest formats is free.
Supply path optimization is not optional any more. Reporting on advertisers who combined supply path optimization with real-time fraud detection describes invalid traffic falling from around 21% of impressions to under 5%. That is the difference between one in five impressions being worthless and one in twenty.
Ad blocking still removes about a third of desktop reach. Roughly 31% of users globally block ads, close to 38% on desktop web, and no bid strategy recovers them. Plan reach accordingly rather than being surprised by it, and note that CTV is effectively unblockable, which is part of its appeal.
Running a demand-side platform is the easy part, and it is roughly where most display advertising agency engagements stop. Everything below is what determines whether the money buys anything.
Cutting the number of intermediaries between your budget and the publisher. Fewer hops means more working media and less fraud exposure.
Naming the sites you will buy on rather than blocking the worst offenders after the fact. Slower to build, dramatically cleaner.
Blocking fraudulent and unviewable inventory before you bid on it rather than reporting it afterwards.
Verifying that whoever is selling the impression is authorised to. Basic hygiene that a surprising number of accounts skip.
Choosing units by measured viewability rather than by what the creative team already produced.
A programmatic media buyer builds segments large enough to deliver, then caps frequency so you are not paying to annoy the same person forty times.
Full sets sized to the placements that actually perform, produced through video and creative rather than resized once.
Third-party measurement of viewability and invalid traffic, reported alongside spend rather than buried.
Programmatic runs alongside search, CTV and retargeting so budget moves to whatever is working.
Send us a placement report from your current programmatic ad agency or display advertising agency. We will tell you what share of your impressions were viewable, where your supply path is leaking, and which sites you are funding that you would not choose.
Get my free auditProgrammatic is a reach and assist channel. It suits businesses with long consideration cycles and audiences worth reaching before they start searching.
Account-based targeting and long cycles where search alone cannot reach the committee.
🏥HealthcareBrand safety and category restrictions make private marketplace buying close to mandatory.
⚖Law FirmsHigh case value absorbs the CPM. Geographic and in-market targeting do the work.
🛒EcommerceDynamic creative and product-level retargeting. Margin decides how far it can scale.
🏠Real EstateHousehold and geographic targeting map directly onto how the business operates.
💻SaaSWorks for category leaders building awareness. Hard to justify before product-market fit.
🦷DentalTight radius targeting only. Broad programmatic rarely pays back at this ticket size.
🍽RestaurantsUsually the wrong channel. Margins are too thin unless catering or events are the goal.
Programmatic advertising is the automated buying and selling of digital ad space through real-time auctions. Rather than negotiating with individual publishers, an advertiser sets targeting and bidding rules in a demand-side platform and software decides in milliseconds which individual impressions to buy and what to pay. It now accounts for roughly 91% of US display spend, so if you run display, native, video, audio or connected TV advertising you are almost certainly buying programmatically already. The visualiser on this page shows a single auction play out across its full 100 milliseconds.
Open exchange display typically runs $1 to $4 CPM and private marketplace inventory $5 to $15, with connected TV considerably higher. But CPM alone is misleading, because it describes what you paid rather than what you received. A $3 CPM at 55% viewability with 16% invalid traffic can cost more per viewable human impression than a $9 CPM at 78% viewability with 4% invalid traffic. The true cost calculator on this page runs that comparison with your own numbers. Management fees are separate and we price them to the work rather than as a fixed percentage of spend.
The open exchange is a public auction where anyone can bid on almost anything. It offers maximum scale and the least control over what your ad appears beside. A private marketplace is an invitation-only auction across a defined set of publishers, so you know where the inventory comes from. The practical difference shows up in quality: reported invalid traffic on open exchanges runs around 14 to 18%, against roughly 3 to 5% on private marketplace deals. Private inventory costs several times more per thousand impressions and frequently costs less per impression a human actually saw.
Ad fraud covers bot traffic, device spoofing and fabricated inventory — impressions billed to you that no person ever saw. Industry estimates put global losses in the tens of billions of dollars annually, and one study of $14.4 billion in programmatic spend across 86 advertisers concluded that roughly 47% went to low-quality or fraudulent inventory. Reported invalid traffic runs 14 to 18% on open exchanges and 3 to 5% on private deals. The defences are unglamorous and effective: ads.txt and sellers.json verification, pre-bid filtering, third-party verification and inclusion lists rather than blocklists.
Under the IAB and MRC standard, a display ad counts as viewable when at least half its pixels are on screen for at least one second, and two seconds for video. The 2026 cross-network average sits near 72% for display and around 79% for video. Above 70% is generally considered good and above 80% excellent. Format matters enormously: desktop leaderboards report near 56% while half-page units reach 82%, on the same inventory at the same price. Dropping structurally weak formats is the cheapest viewability improvement available.
Supply path optimization means reducing the number of intermediaries between your budget and the publisher. The same impression can often be bought through several routes, each with different fees and different fraud exposure, and the shortest path is usually both cheaper and cleaner. Reporting on advertisers who combined supply path optimization with real-time fraud detection describes invalid traffic falling from around 21% of impressions to under 5%. It is one of the highest-return technical changes available and one that few programmatic advertising companies raise unprompted.
Because display is not a last-click channel and judging it as one will always disappoint. Average display conversion sits near 0.71% against roughly 4.4% for paid search, which reflects intent rather than execution: search reaches people already looking, display reaches people who were not. Retargeting converts around twice the display average because those people have already shown interest. Display earns its place through assisted conversions, branded search lift and reach, and should be measured that way.
No, but you should plan around it. Global ad block penetration sits near 31%, close to 38% on desktop web and around 24% on mobile web, and no bidding strategy recovers those users. That is a reason to model reach honestly rather than abandon the channel, and a strong argument for weighting budget toward environments where blocking is impractical — in-app inventory, and connected TV, where blocking is effectively zero.
The Google Display Network is one source of programmatic inventory bought through Google's own system, and it is genuinely convenient if you already run Google Ads. An independent demand-side platform reaches inventory well beyond Google's network, gives considerably more control over supply path and verification, and lets you buy private marketplace deals Google will not offer. For smaller budgets the display network is often the sensible starting point; past a certain scale the control is worth the added complexity.
Demand-side platform minimums vary, and several impose monthly floors that put independent programmatic out of reach below roughly $5,000 to $10,000 a month. Below that, the honest answer is usually that your money works harder in Google Ads or Meta, where the same budget buys higher-intent traffic without the supply chain complexity. We will tell you that rather than sell you a channel your budget cannot support properly. Call +1 720-712-8615 and we will be straight about which side of that line you are on.
Send a placement report and we will tell you what you are actually buying. We respond within one business day.
Your information is never sold or shared. We respond within one business day.