What is an ad server, and what do its reports tell a buyer?
Part of Programmatic media planning
The short answer
An ad server is the system that decides which ad fills a slot, delivers the creative and records what happened. Publishers use one to choose among direct campaigns and programmatic demand for each slot. Advertisers and agencies use their own to host creatives, rotate them across sites and count impressions and clicks independently of the seller.
Key takeaways
- A publisher ad server decides which campaign or programmatic buyer gets each ad slot on the publisher's own pages and apps.
- An advertiser ad server hosts the creative, decides which version to show once the slot is bought, and counts delivery for the buyer.
- Publisher ad servers rank demand by line item type, priority and price, and one server's documented defaults rank guaranteed campaigns ahead of non-guaranteed demand.
- Ad server and DSP counts differ because they record different events at different points in the chain, so a plan names which one is the system of record.
What is an ad server?
Definition
- Ad server
- Software that receives ad requests, decides which ad or creative to return, delivers it or redirects to the system that will, and logs the events that follow, such as impressions and clicks.
The industry's measurement rules treat the ad server as a counting system as well as a delivery one. The MRC and IAB Tech Lab Desktop Display Impression Measurement Guidelines (opens in a new tab) are written for media sellers and "ad serving organizations (including third-party ad servers and organizations that serve their own ads)." For client-side ad serving, the guideline lets "the ad server itself" count the impression, as long as counting waits until the ad has loaded and begins to render.
Video makes the role explicit. VAST 4.3 (opens in a new tab), IAB Tech Lab's video ad serving template (December 2022), says VAST "transfers important metadata about an ad from the ad server to a media player." The player's request "is sent to the primary ad server, which may be the publisher's ad server or a supply-side platform (SSP)", and that server may answer with the ad or redirect the player to a secondary ad server.
How is a publisher ad server different from an advertiser ad server?
Both are called ad servers, and both log impressions, but they sit on opposite sides of the transaction and answer different questions. The publisher's server decides who gets the slot. The advertiser's server decides which creative fills it once it is bought, and keeps the buyer's own count.
| Question | Publisher ad server | Advertiser ad server |
|---|---|---|
| Who runs it | The site or app owner, or its sales house | The advertiser or its agency, sometimes called a third-party ad server |
| What it decides | Which direct campaign, deal or programmatic buyer fills each slot | Which creative, size or version the bought slot shows |
| What it is set up with | Ad units, line items, priorities, prices, inventory rules | Campaigns, placements by site, creatives, rotation and click tracking |
| Where its code sits | An ad tag or tagging library on the publisher's page or in the app | A creative tag or VAST response inside the bought ad slot |
| What its report shows | Fill, revenue and delivery by ad unit and buyer | Impressions, clicks and creative delivery by site and placement, counted for the buyer |
| Why buyers care | It is where a direct deal or sponsorship actually gets priority | It is the buyer's independent count across every site and DSP |
On a programmatic buy, both run on the same impression. The publisher's ad server passes the slot to an SSP, the DSP wins it, and the winning markup calls the advertiser's ad server, which returns the creative and logs the impression. In VAST terms, the DSP's response can be a wrapper: VAST says wrappers "provide a way for one ad server to redirect a media player to another, secondary ad server."
How does an ad server pick which ad to show?
A publisher ad server filters, then ranks. It first keeps only the line items that can run on this request: the right ad unit and size, the targeted geography and device, the flight dates and any frequency or competitive rules. Then it orders what is left by line item type and priority, and within remnant demand by price.
| Line item type | Default priority | Typical use |
|---|---|---|
| Sponsorship | 4 | A guaranteed share of a section's or a site's impressions |
| Standard | 6, 8 or 10 | Guaranteed impression goals sold direct |
| Network, Bulk, Price priority, exchange demand | 12 | Non-guaranteed demand competing on price |
| House | 16 | The publisher's own promotions when nothing else fills |
The table comes from that server's line item types and priorities (opens in a new tab) page, which warns that "line item type and priority are only one factor in ad selection." It describes dynamic allocation letting non-guaranteed demand compete with guaranteed line items in real time, and remnant line items being ranked by value CPM so the system "chooses the highest value remnant line item." Other ad servers use their own names and rules, so ask each publisher how its server ranks guaranteed and price based demand.
How one ad request is decided on the publisher side
- Read the request. The tag reports the ad unit, sizes and any custom targeting values for the slot.
- Filter eligible line items. Drop anything that does not match the unit, size, geography, device, dates or rules.
- Rank by priority. Sponsorships and guaranteed goals that need delivery go first.
- Let remnant demand compete. Non-guaranteed line items and programmatic bids compete on value.
- Return and log. Send the winning ad or redirect, then record the impression when the ad renders.
The advertiser ad server's choice is narrower. The slot is already bought; it picks which creative to send, by rotation rules, sequence, the placement's size or the targeting set on the placement, and logs it.
Why do ad server and DSP counts differ?
Because they count at different points. A DSP sees the auction: it knows when it won, and it may report impressions on a billing event the exchange defines. OpenRTB 2.6 (opens in a new tab) says the win notice is "not necessarily indicative of a delivered, viewed, or billable ad", and the billing notice fires "based on exchange-specific business policy." The advertiser ad server counts only when its own tag is called from the device, after the creative has loaded.
- Chain losses. Every redirect between the DSP's win and the advertiser's tag is a request that can fail or time out; the desktop guideline notes, for rich media, that redirects "can impact the accuracy of ad counting due to latency."
- Timing of tracking calls. VAST 4.3 asks that all Impression URIs in a response and its wrappers fire "at the same time" because delays cause "discrepancies."
- Filtering. Each system removes invalid traffic its own way, and on its own schedule.
- Reporting setup. Time zones, date cutoffs and how placements map to DSP line items are set separately in each system and need aligning.
What does an ad tag do?
An ad tag is the snippet of code, or the URL, that asks an ad server for an ad and tells it where the ad will go. On a publisher's page it marks a slot: one publisher tagging library (opens in a new tab) describes itself as an ad tagging library that takes "key details like the ad unit code, ad size, and custom targeting" and builds the request. In OpenRTB, the tagid field carries the "identifier for specific ad placement or ad tag that was used to initiate the auction."
On the buyer's side, a creative tag or a VAST tag sits inside the bought slot and calls the advertiser's ad server, which returns the creative and the tracking URLs. A buyer never touches the publisher's tags, but the publisher's setup decides whether a slot exists, what size it is and whether it refreshes. Those are page facts worth checking before a site joins a plan.
How can a planner tell which sites run a real ad stack?
What should a buyer ask about ad server reports?
Ad server questions before a campaign starts
- Which ad server is the system of record for impressions and clicks, and who owns the account.
- Whether that server counts on render, and which MRC guideline it follows for each format.
- How placements in the advertiser ad server map to DSP line items and deal IDs.
- Whether video is served as a VAST wrapper, and how many redirects sit between the DSP and the creative.
- Which time zone and date cutoff both reports use.
- How each system filters invalid traffic, and when its numbers become final.
- For direct buys, which line item type and priority the publisher booked the campaign at.
Fictional example
A car dealership group in Newfoundland and Labrador
A fictional dealership group in St. John's and Corner Brook runs a spring display and video campaign across Newfoundland and Labrador, in English, beside local news, car reviews and home and family sections. It buys a sponsorship of a local news site's driving section directly and runs the rest programmatically through a DSP on an approved site list.
The agency hosts every creative in its own ad server, with one placement per site and one for the sponsorship. The planner asks the news publisher to confirm the sponsorship was booked as a sponsorship line item, so it serves ahead of exchange demand in that section, and sets both reports to Newfoundland time.
When the first weekly DSP report runs ahead of the ad server, the planner finds one video placement wrapped twice and asks the DSP team to call the agency's VAST tag directly. The client report uses the ad server as the count. No delivery or performance result is implied.
For where the ad server sits in the wider plan, see the programmatic media planning hub and the what is an SSP guide.
Frequently asked questions
Is a third-party ad server the same as an advertiser ad server?
In the MRC guideline's usage, yes. Third-party ad server is the MRC guideline's term for ad serving organizations other than the site that sells the space, and an advertiser or agency ad server fits that description. That separation is why its count can serve as an independent check on the seller's.
Does a buyer need its own ad server for programmatic buying?
Not strictly. A DSP can host creatives and report delivery on its own. An advertiser ad server adds one count across several DSPs, direct buys and sites, central creative control and click tracking that does not depend on any seller. The case for one grows with each buying route a plan adds.
Is an ad server the same as an SSP?
No. A publisher ad server decides how each slot is filled, from direct campaigns to programmatic demand. An SSP sells the slots that go to programmatic buyers and runs or connects to the auctions. VAST notes that the primary ad server a player calls may be either one, which is why the terms get mixed up.
Can an ad server see which website an ad ran on?
It can, at the domain or page level, from the request and the placement setup. Advertiser ad servers are set up with placements by site, so their reports can show delivery by site or placement. On apps, connected TV and some private deals the reported value may be an app ID or a placement name instead, so check what the site column actually contains.
Sources
- Desktop Display Impression Measurement Guidelines, version 1.1, October 2017 (opens in a new tab)IAB Tech Lab and Media Rating Council
- VAST 4.3, December 2022 (opens in a new tab)IAB Tech Lab
- OpenRTB 2.6 specification (opens in a new tab)IAB Tech Lab
- Line item types and priorities (opens in a new tab)Google Ad Manager Help
- Get started with the Google Publisher Tag (opens in a new tab)Google for Developers
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.