Two ways to buy the same inventory
A direct buy begins with a conversation. The advertiser or agency briefs a publisher, the publisher proposes placements, dates and a price, and the agreement is written on an insertion order that commits the publisher to a volume of impressions or a period of exclusivity and commits the advertiser to pay for it. The creative is trafficked by sending the publisher an ad tag from the advertiser's ad server, or by sending the files themselves. Delivery is guaranteed, the placement is known in advance and the relationship carries weight: a homepage takeover, a sponsored section, a newsletter slot or a custom format is sold this way because the publisher is selling something scarce and specific.
A programmatic buy begins with a setting. The DSP is told what to buy, for whom, where and at what price, and then decides impression by impression as bid requests arrive from exchanges. The open auction is the most familiar form, but a programmatic buy can also run through a private marketplaceAn invitation-only programmatic auction a publisher runs for selected buyers through a deal ID., a preferred dealA programmatic deal giving one buyer a first look at eligible impressions at a fixed, agreed price before the auction, with no volume guarantee. or a programmatic guaranteedA programmatic deal with a fixed price and reserved impressions, bought through a DSP. deal, each with its own balance of certainty and price; the deal types guide explains how each is matched and prioritised. The price is set by the auction or the deal, the targeting by data, and the whole plan can be changed while it runs.
The two methods meet inside the publisher's ad server. Google documents in Ad Manager how reserved line items sold directly and demand arriving programmatically are prioritised against each other for the same slot, which is why a page can carry a direct-sold ad in the morning and an auction-won one in the afternoon. The real question is never programmatic or direct; it is which inventory in a plan deserves which method, and how to keep the two in one view.
Compare table
The table sets the two methods against each other on the dimensions that decide the choice. Where the answer is "it depends", the dependency is named.
Programmatic buying vs Direct buying
| Row label | Programmatic buying | Direct buying |
|---|---|---|
| Control over placement | Site, app and placement lists, deals and exclusions shape it; the exact slot is decided per impression | Exact placement, position and dates agreed in advance and guaranteed |
| Price mechanism | Set per impression by auction, or by a negotiated deal price for private and guaranteed deals | Fixed cost per thousand or flat fee on a rate card, negotiated per buy |
| Fees and working media | DSP, exchange, data and verification fees come out of the spend before the publisher is paid | Most of the spend reaches the publisher; agency fees apply either way |
| Targeting and data | Audience, contextual and first-party data applied per impression; the main strength | Limited to the publisher's own audience and whatever segmentation it offers |
| Measurement | Full log-level reporting, viewability, reach and frequency within the DSP; conversions through the ad server | Publisher reports plus the advertiser's ad server tags; viewability depends on the publisher accepting measurement |
| Scale | Reaches many publishers and most of the open web and apps from one setup | Bounded by the publisher's audience; scale means many separate negotiations |
| Workload | Setup and ongoing optimisation; no per-publisher negotiation | Negotiation, insertion orders, tag exchange and reconciliation for every publisher |
| Creative formats | Standard sizes and video; rich formats where the exchange supports them | Custom units, sponsorships, takeovers, native editorial, newsletters and events |
| Speed and flexibility | Live within hours; budget, targeting and creative changeable mid-flight | Weeks to agree; changes need the publisher's cooperation |
| Commitments and minimums | Usually none beyond platform contracts; deals may set floors | Minimum spend and reserved volume committed up front |
| Reach and frequency control | Caps and reach reports across the whole programmatic plan | Per publisher only, unless the advertiser's ad server tags are used |
| Main risk | Quality variance, fees eating working media, reliance on identifiers | Paying a premium for certainty that may not be needed; under-delivery recourse limited to make-goods |
Where direct wins
Direct wins when the thing being bought is scarce. A homepage takeover on launch day, exclusive sponsorship of a section or an event, a presence in a newsletter that people open deliberately, an editorial integration written by the publisher's own team: none of these exist in an auction, because the publisher is selling a position and a moment rather than an impression. The same is true of formats that need the publisher's cooperation to build, and of inventory that publishers choose to keep out of exchanges altogether, which is often their best. Out-of-homeAdvertising in places people pass on the way somewhere else, such as roadsides, stations, shelters and screens, bought by site, panel or screen. is a case in point: classic sites are still bought site by site, while digital screens can go either way, and the guide to buying out-of-home advertising covers choosing between them for a campaign.
Direct also wins when certainty matters more than efficiency. A campaign that must deliver a volume of impressions in a defined context in a defined week cannot wait for an auction to decide whether it does; a reserved buy guarantees it, and the premium paid is the price of the guarantee. Where a publisher holds first-party dataData a business collects directly from its own customers, visitors and users. about its own readers and sells against it only directly, the targeting is available by no other route. And where brand safety is a question of editorial adjacency rather than keyword exclusion, the relationship with a known publisher does what no verification segment can. The trade-off is real: direct costs more per thousand, moves slowly once agreed, measures less uniformly and produces a pile of insertion orders and reconciliations that someone has to manage. The insertion order sign-off checklist keeps each of those orders checkable, with its terms, verification standards and billing source agreed before signature.
Where programmatic wins
Programmatic wins on breadth, data and speed. A single setup can reach people across most of the open web and app inventory, which no number of direct negotiations could assemble, and it can apply the advertiser's own audience data and the contextual signalsChoosing where ads appear from the content of the page, app or video being viewed rather than from data about the viewer. of each page to every impression. That is the core of the case: ReachUnique people who saw an ad. at scale with targeting that direct cannot match, at a cost per thousand (CPMCost per 1,000 impressions.) set by competition rather than a rate card. The CPM calculator turns a budget and an impression count into that unit so the two methods can be set side by side, provided fees and viewability are included on both.
It also wins on control during the flight. Budget moves between line items as results arrive, creative is swapped without asking anyone, and a placement that underperforms is excluded the same day. FrequencyAverage times a person saw an ad. is managed across the whole programmatic plan rather than within one publisher's walls. Testing is cheap, retargeting is only possible this way, and a direct relationship can be extended into a private marketplace so that the publisher's inventory joins the same reporting and frequency management as everything else. The trade-offs are the fees between the advertiser and the publisher, the variance in quality across an open market, and the dependence on identifiers that are less reliable every year.
A product launch buys both
A consumer electronics brand plans a £240,000 launch over three weeks. It books a direct homepage takeover with a technology publisher for launch day at £36,000, a two-week sponsored review section with the same publisher at £44,000, and runs the remaining £160,000 programmatically through DV360 across display, online video and YouTube. All three buys serve through CM360 tags. The direct buys deliver 3,800,000 impressions at an effective CPM of £21.05; the programmatic buy delivers 32,000,000 impressions at an average CPM of £5.00 after fees, with a measured viewable rate of 61%. On a viewable-impression basis the direct buys cost roughly £24 per thousand and the programmatic buy roughly £8, which is the comparison the planner records. Because every impression ran through one ad server, the frequency report shows that 420,000 people saw the takeover and then at least three programmatic exposures, and the Floodlight data attributes 1,150 pre-orders across the plan with 310 of them touching both buys. Next launch, the brand keeps the takeover, drops the sponsored section, and moves that budget into a private marketplace deal with the same publisher.
Running both through one ad server
The decision to use both methods is easy; the discipline is to run them as one plan. That means serving every buy, direct and programmatic, through the advertiser's own ad server, so that there is one count of impressions that the plan is judged against, one frequency ceiling that sees exposures from every source, one set of conversion tags and one creative library with version control. Without it, the direct publisher reports its own numbers, the DSP reports its own, the two cannot be reconciled, and a person exposed by both is counted twice and capped by neither.
In a Google stack the ad server is CM360. Direct placements are trafficked as placements whose tags are sent to the publisher; programmatic line items in DV360 reference creatives hosted or tracked in CM360; Floodlight activities record conversions for both. The ad serving basics guide explains what happens when a page calls one of those tags and why the publisher's count and the ad server's count differ even when everything is working. The arrangement breaks where a publisher refuses third-party tags, inside walled gardens that accept only their own measurement, and in some in-app environments, and those gaps should be listed in the plan rather than discovered at reconciliation.
Within a wider plan, direct and programmatic are two tactics inside the display and video line, and the channel mix guide describes how that line sits with search, social and the rest. The programmatic guide covers the ecosystem the DSP operates in, and the author's media planning and buying expertise page describes how both methods are planned and reconciled in practice. The media planning framework applies the same test to each: what the buy is for, what it would cost to achieve that another way, and how the result will be measured before the money is committed.
Common mistakes
- Comparing a direct rate card with a programmatic CPM without adding fees to one side and viewability to both.
- Buying premium placements programmatically and expecting a takeover; the auction sells impressions, not moments.
- Buying standard inventory direct at a premium for certainty the campaign never needed.
- Running direct buys without ad server tags, so their impressions, frequency and conversions live only in the publisher's report.
- Letting the direct and programmatic teams plan separately, then discovering the overlap in the wrap-up.
- Treating programmatic guaranteed as a reason to stop negotiating. The price is still negotiated; only the delivery changed.
- Judging direct on last-click conversions. A takeover is bought for a moment of attention and should be measured on reach, recall and lift, not on clicks.
- Forgetting the make-good clause. Direct under-delivery is settled by agreement; write the remedy into the insertion order.
