Programmatic Direct vs Open Exchange: Which Should Publishers Use

Programmatic Direct: Guaranteed Deals That Just Happen To Run Through An API

Programmatic direct is really just an old insertion order wearing new plumbing. An advertiser and a publisher agree on a price, a volume, and placement terms before a single impression runs, and then the delivery mechanics — pacing, targeting, reporting — get handled through the ad server instead of a manual trafficking sheet. The two flavors that matter are programmatic guaranteed (PG), where the publisher commits to a fixed number of impressions at a fixed CPM, and preferred deals, where the advertiser gets first look at inventory at an agreed price but volume isn't locked in. Both skip the open auction entirely.

The CPMs here are the highest you'll see anywhere in your stack. I've negotiated PG deals for finance and travel clients running $9 to $18 CPM on desktop homepage placements, against a blended open exchange rate on the same pages of $1.40. That gap is the whole reason sales-driven publishers bother with the extra overhead — a 500,000 impression PG deal at $12 CPM is $6,000 that would've cleared for under $1,000 through RTB. For a foundational rundown of how the layers connect, this programmatic advertising explainer is worth reading alongside this one.

What buyers get in return is certainty. A brand running a product launch doesn't want to gamble on whether they'll win enough auction impressions in the right geography during the right week — they want guaranteed delivery against a guaranteed audience, with brand safety terms spelled out in a contract, not inferred from a sellers.json file. That certainty is worth a premium to them, and it should be worth negotiating hard for on your end too, especially around make-good terms if your own traffic underperforms during the flight.

One distinction that trips people up: programmatic direct is not the same thing as a manual, non-programmatic IO where a trafficker builds creative tags by hand and emails a screenshot for approval. Some publishers still run that old-school version alongside programmatic direct, and it's worth killing off if you can — you get none of the automated reporting, pacing alerts, or reconciliation that programmatic direct gives you for free, and it eats far more of your ad ops time per dollar of revenue than either programmatic direct or the open exchange.

The Open Exchange Is Still Doing Most Of The Work

The open exchange is real-time bidding in its purest form — any DSP with an active seat can bid on any impression you make available, the highest bid wins (or clears at a hybrid first/second-price mechanic depending on the exchange), and the whole thing resolves in under 100 milliseconds. No relationship required, no minimum spend, no forecast call. That's why it clears close to 100% fill on most sites, even ones nobody in a media buying seat has ever heard of.

The auction mechanics matter more than most guides admit. Since header bidding became standard, publishers stopped waking up 15-20% of demand partners in a sequential waterfall and started letting all of them bid simultaneously against the same impression, which is a structurally different — and better — way to sell inventory. If you haven't compared the two setups directly, how waterfall and header bidding auctions actually differ explains why that shift alone tends to lift open exchange yield 10-25% on its own.

But liquidity comes at the cost of average price. Blended open exchange CPMs I see across mid-size content sites run $0.60 to $2.20 depending on vertical, geography, and viewability — finance and B2B sites at the top, general entertainment and international traffic at the bottom. Volume is never the problem here. Ceiling is, and no amount of demand partner tinkering changes that ceiling much once you've already connected a reasonably deep bidder stack.

The Tradeoff Nobody States Plainly: Predictability vs Effort

Here's the part most monetization guides gloss over: open exchange revenue isn't just lower on average, it's less predictable month to month, even at constant traffic. I've watched sites swing 18-30% in blended RPM between a strong November and a soft February with zero change in pageviews, purely because advertiser demand and budget cycles moved. Direct deals don't do that. A signed PG contract pays out the agreed rate whether the broader market is soft or hot that quarter.

The cost side is what gets ignored. Programmatic direct isn't a switch you flip in your ad server — it requires someone to build relationships with agencies or brands, forecast available inventory months out, negotiate rates, and manage renewals. That's a sales function, not an ad ops function, and it doesn't scale down well. A single account manager can realistically manage $150,000-$300,000 in annual direct bookings before service quality drops; below that volume, the overhead-to-revenue ratio gets ugly fast.

So the real decision isn't "which is better" — it's "what's my forecast confidence worth, and can I staff for it." A publisher with volatile, low-value traffic gains almost nothing from direct sales effort. A publisher with a defined, valuable audience — enterprise software buyers, affluent travel intenders, parents shopping for a specific category — is leaving obvious money on the table by staying open-exchange-only past a certain point in its growth.

There's also a reporting cost to predictability that nobody budgets for. A PG deal means committing to a delivery number in a contract, which means you need reasonably accurate traffic forecasting — not just "last month plus 5%," but a real model that accounts for seasonality, algorithm updates, and the possibility that a chunk of your traffic simply doesn't show up one week. Publishers who sign PG deals without that forecasting discipline are the ones who end up paying make-goods out of their own open exchange inventory later.

Private Marketplaces: The Tier Everyone Mentions And Nobody Explains

A private marketplace is an invite-only auction. You set up a deal ID, decide which buyers get access — usually a curated list of DSPs or trading desks — set a price floor, and only those invited parties can bid. It still runs through real-time auction mechanics, so there's no guaranteed volume like PG, but the floor and the curated buyer list mean the average price clears meaningfully higher than open exchange, typically $2 to $5 CPM depending on the vertical and how tightly you've curated the buyer list.

PMPs solve a specific problem for both sides. Buyers get access to inventory they trust — lower fraud risk, better viewability, brand-safe context — without the commitment of a signed IO. You get a higher average price than the open exchange without needing a salesperson to negotiate every single deal; a lot of PMP setups get initiated by the demand side reaching out to your ad ops team or your SSP rep, not the other way around.

• Data-driven PMPs: the buyer supplies audience segments and pays a premium to reach them on your inventory
• First-look deals: a single buyer gets priority bidding before the open auction runs on that impression
• Category-exclusivity PMPs: a buyer pays for the only ad slot in a given content vertical on your site
• Curated marketplace PMPs: run through an SSP-managed marketplace that pre-vets buyers for you before they ever reach out

How You Actually Get Into Programmatic Direct Deals

This is the part that surprises publishers who've only ever dealt with self-serve platforms: there's no dashboard where you flip on programmatic direct and deals start appearing. You get into PG and preferred deals one of three ways — you build an internal or outsourced sales function that pitches agencies directly, you work with a monetization partner or rep firm that already has agency relationships and negotiates on your behalf for a revenue share, or you get listed in a platform-level deal marketplace where buyers discover and initiate deals with you.

Google Ad Manager's Marketplace and similar tools from Xandr, PubMatic, and Index Exchange fall into that third category — they surface your inventory to a pool of agency buyers who can propose a deal directly, which lowers the barrier but still requires you to meet baseline qualification: consistent traffic volume, clean ads.txt and sellers.json records, viewability above roughly 55-60% on the placements you're offering, and no history of policy violations or invalid traffic flags.

Working with a partner is the fastest path for most mid-size publishers, because the partner already has the agency relationships and the credibility to get a deal reviewed on short notice. If you're evaluating that route, it's worth understanding what a full-stack monetization partnership actually covers versus what you'd still need to manage in-house — reporting transparency and deal exclusivity terms vary a lot between partners, and the fine print matters more than the pitch deck.

• Minimum scale most buyers expect: roughly 500,000-1M+ monthly pageviews before a direct sales conversation is worth an agency's time
• Clean supply chain documentation — ads.txt, sellers.json, and app-ads.txt if applicable, fully accurate and current
• Viewability and brand safety metrics you can actually show a buyer, not just claim in a sales call
• A defined audience story — demographic, intent, or vertical — that a media planner can pitch internally to their client

What A Realistic Revenue Mix Looks Like At Scale

The publishers I work with who've built out all three tiers land somewhere around 55-65% of revenue from the open exchange, 20-30% from PMPs, and 10-20% from PG and preferred deals — and that mix shifts hard by season. A lifestyle site running a strong direct relationship with a home goods brand might see PG jump to 30% of revenue in November and December, then drop back to 8-10% of revenue in the January-February trough once direct budgets dry up industry-wide.

Blended RPM tells the real story better than any single CPM figure. A site running open-exchange-only might sit at a $4.50 blended RPM. Add two solid PMP relationships and that can move to $6-$7. Layer in even one meaningful PG contract and you're looking at $8-$10+, not because the PG volume is huge, but because it pulls the weighted average up disproportionately — 50,000 monthly impressions at $14 CPM adds $700 that the open exchange portion would've paid maybe $90 for on the same pages.

None of this replaces the need for demand diversity within each tier, by the way. I still see publishers who built a beautiful three-tier structure but only plugged in three or four demand partners at the open exchange layer, capping their own liquidity before the auction even starts. The tiers and the breadth of demand sources you connect are separate levers — you need both pulled, and neglecting one undermines the other.

• Open exchange only: roughly $3.50-$5.00 blended RPM on a typical content site with a solid header bidding setup
• Open exchange plus two to three PMPs: roughly $6.00-$7.50 blended RPM
• Open exchange, PMPs, and one active PG relationship: roughly $8.00-$11.00+ blended RPM, with heavier seasonal swing

When Chasing Direct Deals Is A Waste Of Your Time

I'll say something most monetization consultants won't: if you're under roughly 1-2 million monthly pageviews, or your traffic is heavily international with low commercial intent, building out a programmatic direct sales motion probably isn't worth it yet. The math doesn't work — the time spent building agency relationships, responding to RFPs, and managing a handful of small insertion orders costs more in opportunity cost than the incremental CPM lift delivers.

At that scale, your time is better spent on the layers that don't require a sales function: tightening your header bidding setup, adding PMP deals through your SSP rep (who does that outreach for you), and making sure your open exchange demand is as deep as it can be. Direct sales overhead is largely fixed cost — a rep costs roughly the same whether they're managing $50,000 or $500,000 in annual bookings — so it only pays off once volume justifies the salary.

A common bad habit is publishers pursuing a single vanity direct deal — a $3,000 sponsorship from a local advertiser — purely because it feels more "premium" than programmatic revenue, while it ties up a homepage placement that would've cleared $4,500 through PMPs and open exchange combined over the same period. Direct isn't automatically better. It's only better when the rate and terms actually beat what the auction would've delivered on that exact inventory.

There's a middle path worth mentioning here too: some SSPs and monetization partners will run light-touch curated deals on your behalf even at moderate scale, essentially acting as your PMP sales function for a revenue share, without asking you to hire anyone or build a pitch deck. That's usually the right on-ramp — you get access to some of the direct-adjacent pricing without the fixed overhead, and you can revisit a real in-house sales motion once traffic and revenue justify it.

Mistakes That Quietly Cap Revenue Across All Three Tiers

The most expensive mistake is setting PG or PMP price floors that block open exchange demand from ever getting a shot. If your ad server prioritizes a PMP line item ahead of the open auction but the PMP only fills 40% of the time, you're leaving the other 60% of impressions sitting idle waiting on a deal that isn't going to clear, instead of letting them fall through to RTB where they'd sell in milliseconds at a real, if lower, price.

The second is treating a signed direct deal as passive income. I've seen PG contracts underdeliver for months because nobody was monitoring pacing, and the publisher only noticed when the advertiser flagged a make-good clause at the end of the flight. Direct deals need the same weekly attention as any other revenue line — pacing, viewability compliance, creative approval turnaround — or you end up owing free inventory to fix a shortfall you should've caught in week two.

• Not reviewing deal priority order in the ad server quarterly as demand shifts
• Letting PMP floors sit static instead of testing them against real clearing prices every few weeks
• Signing PG terms without a make-good clause that protects you if your own traffic dips mid-flight
• Ignoring the latency added by direct or PMP tags, which slows the whole page and hurts open exchange viewability too

Making The Three Tiers Coexist In Your Ad Server Without Cannibalizing Each Other

None of this works if your ad server doesn't understand how to arbitrate between tiers on the same impression. Google Ad Manager's unified pricing rules and line item priority let you set PG at the top (it's contractual, it has to deliver), PMPs next with a price floor that has to beat the open exchange's real-time bid to win, and open exchange demand competing for whatever's left. Get the floor wrong on the PMP tier and you'll either starve it of volume or let it steal impressions that would've cleared higher through RTB.

I run a simple test with new clients: pull the average clearing price of a PMP deal over two weeks, then compare it against the open exchange's average clearing price for the same inventory during the same window, segmented by hour of day if you can. If the PMP isn't clearing meaningfully above the open exchange average — at least 20-30% higher — the floor is either mispriced or the buyer isn't as valuable as the deal implied, and it's time to renegotiate or kill it.

The same discipline applies to PG. Just because it's guaranteed doesn't mean it should run unconditionally ahead of everything else all day — most ad servers let you cap PG delivery by hour or by page section so it doesn't crowd out higher-value open exchange bids during your peak traffic windows, when RTB demand is often paying more per impression than your contracted PG rate anyway. Reviewing that pacing weekly, not just at the end of the flight, is what separates a stack that's tuned from one that's just assembled.

Frequently Asked Questions