Programmatic: buying the open internet by the impression

Programmatic is the automated buying and selling of ad space across the open web — news sites, blogs, apps, connected TV — through software auctions rather than manual insertion orders. Each impression is valued and sold individually, in the fraction of a second while the page loads.

Demand sources and publisher inventory connected through a programmatic marketplace
A bid request moves from publisher supply to eligible buyers; the selected creative returns through the same transaction path.

The machinery: DSP, SSP, exchange

Three kinds of platform make the market:

The flow: a person opens a page → the publisher's SSP announces an impression with context (page topic, device, rough location, consent status) → DSPs evaluate it against their campaigns and bid → the selected creative is served. This is real-time bidding (RTB), and the transaction must complete inside the marketplace's short response timeout while the page or app loads.

Open auction vs private deals

Deal typeHow it worksUsed for
Open auctionAny eligible buyer bids on any impressionCheap reach, prospecting at scale
PMP (private marketplace)Invitation-only auction on a publisher's inventoryQuality environments with auction pricing
Preferred dealFixed price, first look before the open auctionSecuring specific placements without commitment
Programmatic guaranteedFixed price and fixed volume, automated deliveryThe classic direct buy, executed by software

Formats beyond the banner

Programmatic pipes carry standard display banners, native ads styled to match the surrounding content, online video (pre-roll, mid-roll, outstream), connected TV, and digital audio. The same buying logic applies to all of them; what changes is the creative and how viewability is measured.

Brand safety, fraud and viewability

Buying the open internet means your ad can, in principle, appear next to anything — or be "seen" by no one. Three disciplines keep spend honest:

Rule of thumb: judge programmatic on viewable CPM and downstream conversions, not raw CPM. The cheapest impressions are usually cheap for a reason.

Where programmatic fits

It is rarely a first channel. It shines when you need reach beyond the walled gardens, retargeting across many sites at once, or formats like online video and connected TV — and you have the measurement in place to tell good inventory from bad.

Map the supply path before judging price

The same publisher impression can reach a buyer through several exchanges and resellers. Each extra hop can add fees, duplicate auction opportunities and make accountability harder. Supply-path optimization means comparing routes to the same inventory, preferring direct or well-documented relationships and removing paths that add cost without adding unique reach.

SignalWhat it establishesQuestion to askAction
ads.txt / app-ads.txtWhich seller accounts a publisher authorizesDoes the exchange and seller ID match the publisher file?Exclude undeclared or unexplained sellers
sellers.jsonThe identity or role behind a seller accountIs the seller direct, an intermediary or confidential?Investigate opaque chains before scaling
SupplyChain objectThe sequence of parties in a bid requestHow many hops connect the publisher to the buyer?Compare cost and quality by route
Domain or app bundleThe declared environment for the impressionDoes the delivered placement match the report?Block spoofed or unsuitable inventory

These signals reduce uncertainty; they do not prove that an impression was human, visible or effective. An authorized seller can still deliver poor placements. Pair supply transparency with invalid-traffic controls, placement reporting, viewability and downstream outcomes.

A preflight framework for an inventory test

  1. Define eligible environments. Write the countries, devices, formats, content categories and inventory types the campaign may buy. Put prohibited environments in an explicit exclusion list.
  2. Choose the economic unit. Use CPM for delivery, viewable CPM for exposure, and CPA or conversion value for the business result. Do not let a cheap raw CPM become the success criterion.
  3. Verify creative and landing measurement. Confirm impression and click tracking, conversion events, value fields and destination URLs before spending.
  4. Start with inspectable inventory. Require domain, app, placement and supply-path reporting. A test that cannot identify where ads appeared cannot teach you which supply to keep.
  5. Apply exclusions from evidence. Remove placements for invalid traffic, unsuitable content or repeatedly poor outcomes. Avoid deleting an entire category because of one low-volume observation.

Read viewability and fraud as separate dimensions

Viewability measures whether an ad had an opportunity to be seen. It does not measure attention, human identity or persuasion. The Media Rating Council's baseline for many display impressions is at least 50% of pixels in view for one continuous second; video uses a two-second time requirement. A viewable impression can still come from invalid traffic, and a human impression can still fail viewability because it appeared below the fold.

Use a layered diagnostic: authorization describes who may sell; invalid-traffic filtering estimates whether delivery is legitimate; viewability describes opportunity to see; attention or engagement describes response; conversion measurement connects response to an outcome. Collapsing those layers into one “quality score” hides the reason a placement succeeded or failed.

Further reading on this site

Every acronym above is defined in the glossary. For the metrics used to judge programmatic buys — vCPM, viewability, frequency — see the metrics page.

Sources and standards