Thirty-six cents of every dollar entering a DSP effectively reaches a consumer. That is what the ANA found when it traced log-level data across US open-web programmatic, a market where roughly a quarter of spend leaks out before it works: about $22B wasted on $88B. Much of that leak sits in open internet advertising, the inventory bought outside closed platforms like Google and Meta, through exchanges anyone can access. The number gets read as an argument to retreat into walled gardens. But waste at that scale is evidence of a buying problem, not a channel problem, and the brands paying for it are the ones treating the open internet as simply overflow for whatever search and social leave behind.
The brands getting paid back run the open internet as its own budget category, with its own controls and proof.
The open internet is a budget category
Open internet advertising is the digital inventory bought outside closed platforms, through exchanges any buyer can access. It spans CTV, streaming audio, digital out-of-home, native, display, and online video. Treat it as a budget category with its own targets, not as a bucket you fill after search and social are set.
Attention has already moved. Roughly two-thirds of consumer attention now lives on the open internet, while only about one-fifth of ad spend follows it out of the walled gardens. Despite changing consumer behavior, budgets continue to follow habit.
Why does the open internet get remnant treatment?
Planning order decides it. Most teams lock the Meta and Google line items first, then buy whatever reach is left. That sequence hands the open internet remnant budget, remnant creative, and remnant scrutiny, which is the exact combination that produces waste.
The scale of the walled gardens sets the reference point. Google, Meta, and Amazon captured 61% of US domestic digital ad spend in 2026, compared with 52% in Europe. Three companies appear to define what normal looks like across the entire digital ad market.
What you get back by planning it first
One buyer, one audience, one frequency plan across every screen. CTV, audio, DOOH, native, display, and OLV all live in the same buyable pool, so a single team controls how often a household sees you and in what order. No walled garden does that. Each one counts only its own impressions.
That control shows up in cost. Agility's work with one luxury retailer delivered a $0.29 cost per site visit, with 90% of budget reaching net-new prospects. Planning first makes those numbers repeatable and defensible when your CFO asks. Our guide to measuring brand advertising covers the proof side.
What buying the open internet badly actually costs
Bad open internet buying costs roughly a quarter of every dollar you put into it. The ANA's Programmatic Media Supply Chain Transparency Study traced log-level data across a slice of the US open-web programmatic market, and found that only 36 cents of every dollar entering a DSP effectively reaches the consumer. The leak pools in specific places.
Most of it collects in made-for-advertising sites: pages built to carry ads, not readers. Made-for-advertising sites took about 21% of impressions and 15% of ad spend in the ANA's sample. Those pages carry the report's cheapest CPMs. Selling those impressions at the lowest possible price is the business model.
Why does cheap inventory keep winning the auction?
Walled gardens trained a generation of buyers to chase volume. Inside a closed platform, the algorithm handles quality, so the buyer's job narrows to lowering cost per impression and raising delivery. Carry that habit into the open internet, and you get what you optimized for: more impressions, on worse pages, seen by fewer people.
More budget doesn't correct it. A different objective does. Buy toward a named audience and a quality floor, and the CPM stops being the scoreboard.
Buyers already know this. 84% of buyers say they would pay a premium for quality inventory, which makes the cheap-CPM equilibrium an artifact of how teams buy, not what they want.
The cost of doing nothing is not zero
Staying inside the gardens has its own price tag, and it rises every year. Auction competition continues to push platform costs up: average Google Ads cost per click rose 12.9% year over year to $5.26, with cost per click increasing in 87% of industries. You pay the inflation either way.
The difference is what you get back. Disciplined open internet advertising produces reach you can price against that curve. For one client, Agility delivered 4M+ display-driven site visits at a $0.09 CPSV, the same media, bought with a floor. Our breakdown of supply path optimization covers where the leaks sit.
The channels that make the open internet worth buying
Four channels carry a ton of the value: connected TV, online video, streaming audio, and digital out-of-home. Each reaches people on a screen the walled gardens don't own. CTV anchors the plan because it holds the largest share of attention and the widest gap between where people watch and where the money goes.
Why is CTV the anchor channel?
The audience already moved and the budget has not. Streaming accounted for 43.8% of all US TV time, up ten percentage points in two years. Spend is chasing it slowly. CTV ad spend grew 16% year over year in 2024, from $20.3B to $23.6B, as digital video passed linear TV for the first time.
By 2026, CTV will account for 20% of US adults' daily media time but only 8.1% of total ad spend, a roughly 12-point gap between attention and money. Early buyers get cheaper reach. Our guide to CTV advertising measurement covers how to prove what those impressions do.
What the other three channels add
Audio buys attention during commutes, workouts, and chores, hours no screen reaches. Podcasts are the fastest-growing slice: podcast advertising grew 17.6% in 2025 to $2.9B. It sits inside a digital audio market worth $8.4B. DOOH buys physical context, and it moves people. Nearly two-thirds of DOOH viewers took at least one measured action after seeing a digital billboard, and 57% of those who noticed a directional digital billboard immediately visited the business.
Channel | What it buys you | Where it sits in the plan |
CTV/OTT | Full-screen attention, household reach | Anchor: sets frequency for everything else |
Online video | Mid-funnel reach at lower CPM | Extends CTV audiences off the TV |
Streaming audio | Screenless time, high completion | Fills the hours video cannot |
DOOH | Physical context, measured store visits | Local weight and store proximity |
Display/native | Cheap repetition against a named audience | $0.09 CPSV |
The payoff is the halo. Open internet exposure lifts the platform channels you already buy: one campaign produced a 138% paid search conversion lift, an 88% Bing conversion lift, and a 2.6x Meta ROAS. Buy the channels together and search gets cheaper.
The buying discipline: Five controls that separate reach from waste
Five controls determine whether open internet advertising delivers reach or waste: audience definition, a supply-quality floor, cross-channel frequency governance, hygiene rules, and attention as an allocation input. Set all five before you set a CPM target. Every one of them is a buying decision, not a reporting decision.
1. Define the audience before you pick the channel
Buy a named persona across every channel. Buying each channel's native audience product separately gives you five different definitions of the same customer, and no way to reconcile them. Agility builds personas from 1,000+ third-party data sources across behavioral, demographic, purchase intent, and geo categories, so the same person can be targeted on CTV, audio, and display. Most audience tools stop well short of that natively, which is why so many plans default to the channel's stock segment. Our guide to persona-based advertising covers how that definition gets built.
2. Set a supply floor before you optimize price
Exclude made-for-advertising sites and low-attention placements first. Then negotiate CPM. Reverse that order and the auction hands you the cheapest inventory on the exchange, which is the same inventory the ANA traced to roughly 15% of programmatic spend. A floor costs you a higher CPM and buys back working media.
3. Govern frequency across channels, not inside them
Siloed buys overdeliver. A CTV buy, an audio buy, and a display buy each capped at four impressions per week hit the same household twelve times. Cap at the household level across the whole open internet pool, and the waste disappears without cutting reach.
4. Ask vendors the hygiene questions they rarely answer
Three specifics separate real targeting from a slide: geofencing to one-meter accuracy, a 12-month look-back window for past-visitor targeting, and a 36-hour audience scrub that strips bots before the buy. Most vendors publish none of this. Ask anyway.
5. Treat attention as an input, not a report
Attention data belongs in the allocation model. Across 141 econometric models covering £1.8B in ad spend, attention data explained 97.9% of the variation in profit. Our breakdown of attention metrics covers where it fits.
What changes when the open internet leads the plan
Planning open internet advertising first changes the question your team argues about. Brand versus performance gives way to a buying decision: which environment reaches the buyers a platform buy structurally can't reach. Reallocation then turns on reach math, not philosophy, so it survives contact with finance.
Why does consistent presence compound?
Presence accumulates where time already sits. In 2023, US consumers spent 61% of their online time on the open internet, compared with 39% on walled gardens, a reversal from 2014, when the gardens held 62%. Most of that inventory sells through open exchanges, not a closed buying tool. Show up there every week and reach builds on itself. Go dark for a quarter, and you buy the same audience again at a higher price: brands that go dark lose about 2% of future revenue per dark quarter.
The first move for a $10M+ budget
Consolidate before you add a dollar. Most brands already run CTV, audio, display, and DOOH as separate line items, with separate vendors, separate audience definitions, and no shared cap. Pull them into a single cross-channel buy with precision brand advertising governed by personas and frequency limits.
That single step usually recovers reach you already paid for. It costs nothing in new budget. It also gives you a clean baseline, so any incremental dollar you add later gets measured against something real. Our guide to unified marketing measurement covers how to read the combined number.
How Agility buys the open internet as a named line item
The five controls above are the job. Agility runs them as one buy, not five. Persona targeting comes first: we build audiences from 1,000+ third-party data sources spanning behavioral, demographic, purchase intent, and geo categories, so the same household is addressable on CTV, streaming audio, DOOH, and display. That makes a single frequency cap possible across the entire pool.
Media buying sets the quality floor before it negotiates price. Made-for-advertising inventory gets excluded up front, the only reliable answer to the ANA's roughly 25% waste finding. Precision creative keeps the impressions working once they land. We measure six levers per ad: value proposition, CTA, emotional theme, messaging, people and talent, and art and imagery.
Measurement science closes it. Every campaign runs against a holdout, so the number you report is incremental rather than attributed. In a recent customer campaign, we delivered a 4x ROAS with 90% of budget reaching net-new prospects. Prospecting economics, on inventory most buyers treat as remnant.
See what precision brand advertising looks like for your brand at agilityads.com/test-precision-advertising.
Share in...





