Contextual PMP: keeping a site list's context intact inside a private deal
Part of Curated deals and PMPs
The short answer
A contextual PMP is a private marketplace deal whose inventory was chosen because the pages match a topic, not because of who is reading. The buyer brings a reviewed site list, the seller maps it to sites, sections or ad units in the deal, and the buyer checks after launch that delivery stayed on matching pages.
Key takeaways
- A contextual PMP has to carry the topic evidence through setup, not shrink it to a domain list.
- Section and subdirectory choices decide whether a general site belongs in the deal at all.
- OpenRTB carries section and page categories only when the seller sends them, and cattax names the category list.
- A served site report proves the list, not the page, so page level proof is agreed before launch.
What is a contextual PMP, and what makes it different from other private deals?
Definition
- Contextual PMP
- A private marketplace deal whose sites, sections or ad units were selected because their content matches the brief's topics, with the topic evidence kept alongside the deal terms.
Any private deal limits supply. What makes a deal contextual is the reason for the limit: it picks pages, not publishers by size or users by segment, and the claim behind it is that the ad will sit next to reading about a particular subject. That claim is easy to lose. Once the list becomes a deal, the exchange sees domains and ad units, and a domain says nothing about whether the page on screen matched.
In OpenRTB 2.6 (opens in a new tab), the Deal object holds the terms: ID, floor, floor currency, allowed seats and advertiser domains. It has no topic field. Context travels in the Site object, where sectioncat and pagecat hold IAB Tech Lab content categories for the current section and page. A contextual PMP depends on the deal's inventory definition and those page signals lining up.
| Layer | What it holds | Who controls it | What it cannot show |
|---|---|---|---|
| Site list | Domains chosen from reading their articles | Buyer or planner | Which pages a given impression comes from |
| Deal inventory definition | Sites, sections, ad units or key-values the seller targets | Seller or curator | Whether a section's content still matches next month |
| Bid request signals | Site.page URL, sectioncat, pagecat and the cattax value | Seller's ad stack and exchange | Anything the seller chose not to send |
| Buyer side targeting | Keyword or URL rules in the buying platform | Buyer | Pages the platform could not read |
How do you turn a contextual site list into a deal?
The site list is the start, not the deal. In Google Ad Manager's Private Auctions overview (opens in a new tab), the publisher invites specific buyers, sets a minimum CPM and must "Define the portion of inventory you want to offer to buyers in the Private Auction." That step is where a contextual list survives or flattens: if the seller targets the whole domain, the whole site is in the deal.
Ad Manager's targeting types (opens in a new tab) page says publishers "target specific inventory (ad units and placements)" and can include or exclude ad units, and that custom targeting "allows you to include key-values, audience segments, or content metadata." A section such as a site's food or travel channel can be a deal on its own only when the publisher's ad units or key-values already follow that section. Ask how the site is built before asking for a section level deal.
From reviewed list to contextual deal
- Mark each site's scope. Note whether the whole site matches or only named sections, with two or three example article URLs per section.
- Ask each seller how sections map. Confirm the matching sections have their own ad units, placements or key-values; drop sections that cannot be separated.
- Split specialist sites from general ones. Specialist sites go in whole; general sites go in by section, or not at all.
- Agree the page signals. Ask whether bid requests carry the page URL and section or page categories, and which taxonomy cattax names.
- Write the scope into the deal sheet. List each site with its sections, example URLs and promised signals, next to floor, currency and seats.
- Match the confirmation to the sheet. Before activation, the seller's confirmation lists the same sites and sections, at the same version.
Should contextual signals be layered on the deal in the buying platform?
Display & Video 360's keyword targeting (opens in a new tab) checks "webpage text, app store descriptions, or video metadata for keywords" on Ad Manager and third-party exchange inventory. Its website targeting (opens in a new tab) page supports "two levels of subdirectory targeting" and a combined 10,000 individual apps and URLs per line item. Both can narrow a deal to matching pages after the seller has done its part.
The cost is scale. Each layer applies on top of the deal, so a deal already limited to three sections can be cut to a handful of pages by a strict keyword rule. Display & Video 360's inventory source targeting (opens in a new tab) page also says "Deals are only targeted when you add the deal itself or deal group to targeting," so the layered line item must include the deal. The deal targeting settings guide covers double targeting on geography and audience; topic layers follow the same logic.
| Method | Strength | Weakness | Fits when |
|---|---|---|---|
| None, rely on the deal's inventory | Keeps the scale the seller set up | Context rests entirely on the seller's mapping | The deal holds specialist sites only |
| Subdirectory targeting | Matches the site's own structure, easy to audit | Fails where articles sit outside their section path | Sections have stable URL paths |
| Keyword targeting | Reads page text the platform crawls | Can block matching pages that use other words, and shrinks delivery | A general site's sections are mixed |
| Contextual segment from a provider | Can classify pages the buyer has never read | Its method may not match the list's own reasoning | The buyer can test what it reaches first |
How do you check a contextual deal ran on matching pages?
Start with the evidence each report can carry. Display & Video 360's report dimensions (opens in a new tab) define App/URL as "The name of a site, or mobile app" and Inventory Source as "The name of the deal that supplies buyable impressions to your line items." Together they show which sites a deal ran on. They do not show which article was on screen, so a site level match is proof of the list, not proof of the context.
Page level proof has to be agreed before launch. The Site object's page field is the "URL of the page where the impression will be shown," so a seller that logs it can share served page URLs for the deal; write the log, its time window and format into the terms. If the deal carries a keyword layer, Display & Video 360's keyword report (opens in a new tab) groups impressions "by each keyword," but only for keyword targeted line items, and it leaves excluded keywords out.
Contextual delivery check for the first two weeks
- The served site report for the deal shows only domains on the approved list version.
- Impressions on general sites come from the included sections, per the seller's page URL log.
- A person has read at least 20 served page URLs per general site, marking each match, partial or off topic.
- Categories in the seller's log use the cattax value written on the deal sheet.
- Any site with a visible share of off topic pages has a written follow up with the seller.
- Keyword layer results, if used, are read beside the page sample, not as proof on their own.
How often should a contextual deal list be refreshed?
The reason a site is on a contextual list can vanish while the domain stays. For list upkeep, the Trustworthy Accountability Group's Brand Safety Certified Guidelines (opens in a new tab), version 5.0 of July 2026, section 4.11, require certified direct buyers and intermediaries to document how inclusion and exclusion lists are maintained and to execute those procedures "as part of their internal quarterly review process." That rule covers companies seeking the seal, not all buyers. A quarter is a sensible outer limit for a contextual deal too.
| Trigger | What to recheck | Who asks |
|---|---|---|
| Quarterly review date | Read fresh articles from each site and section; confirm scope still holds | Planner |
| Publisher redesign or new section paths | Whether section ad units or URL paths still follow the content | Trader, with the seller |
| Change of site owner | Topic mix, ad load and who sells the inventory | Planner |
| Off topic pages found in the delivery sample | The section's recent articles and whether to drop it | Planner |
Removing a section mid-flight is a change on the seller's side, so agree in advance whether the seller edits the existing deal or creates a new deal ID. The site list version control guide shows how to keep each list version traceable to the deal it fed.
What does a contextual PMP look like on a real brief?
Fictional example
A western wear retailer's contextual deal in Wyoming
A fictional western wear and boot retailer with stores in Cheyenne, Casper and Laramie wants display ads beside rodeo coverage, ranch life writing and country music news for Wyoming readers, in English, over a six week spring flight.
The planner's reviewed list holds 31 sites. Twenty two specialist rodeo, ranching and western lifestyle publishers go into the deal whole. Nine general regional news and entertainment sites match only in one or two sections; the seller confirms seven have section ad units, and the other two are dropped.
The deal sheet lists each site's scope with example article URLs, a USD floor, the agency's seat, Wyoming as the geography, and page URLs and section categories in the seller's log, with cattax stated. In week one the served site report matches the list, but a reading of 20 served page URLs per general site finds one entertainment site serving gossip pages outside its music section. The seller confirms the section ad unit also runs on a shared template, and the site leaves the deal before week three.
How does PlacementGraph prepare a site list for a contextual deal?
Frequently asked questions
Can a contextual PMP include a large general news site?
Only through the sections that match. A whole general news site brings in all of its pages, and the deal's topic claim no longer holds. Ask the publisher whether the matching sections have their own ad units or key-values. If not, keep the site out of the deal or test it on a separate line item.
Does a contextual PMP use cookies or audience data?
It does not need to. The inventory is chosen from page content, and OpenRTB carries context in the Site and Content objects, apart from user data. A seller may still offer audience segments on the same deal; decide on the deal sheet whether to accept them, since they change what the deal sells.
What should happen when a seller cannot share page URLs?
Treat the deal's context as unverified at page level. Keep only specialist sites whose whole output matches, and move general sites to a line item you can check with buyer side tools. Record on the deal sheet that page logs were not available, so the client knows what the reports can show.
Can one contextual PMP mix English and French sections?
Yes, if both languages match the brief and the seller can separate them. OpenRTB 2.6 lets a seller state content language in the Content object with language or langb. Ask which field the seller fills, and use separate deals when the creative differs by language.
Sources
- OpenRTB 2.6 specification (opens in a new tab)IAB Tech Lab
- Private Auctions overview (opens in a new tab)Google Ad Manager Help
- Targeting types (opens in a new tab)Google Ad Manager Help
- Keyword targeting (opens in a new tab)Google Display & Video 360 Help
- App, channel, and website targeting (opens in a new tab)Google Display & Video 360 Help
- Inventory source targeting (opens in a new tab)Google Display & Video 360 Help
- Dimensions in reports (opens in a new tab)Google Display & Video 360 Help
- Keyword report (opens in a new tab)Google Display & Video 360 Help
- Brand Safety Certified Guidelines, version 5.0 (July 2026) (opens in a new tab)Trustworthy Accountability Group
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.