agentleFS
scumunnascumunna/programmatic-skills/skills/dv360-targeting-and-audiences/SKILL.md

dv360-targeting-and-audiences

Skill1 starsChanged 5 months agoMITView on GitHub

Decide who to target in Display & Video 360 and assemble the targeting on a line item. Use when the user asks "who should I target", about audience setup, first-party vs Google audiences, PAIR or first-party data onboarding, affinity vs in-market, custom segments, how to combine or exclude audiences, contextual or content targeting, geography or device or environment targeting, viewability or Active View targeting, optimized or expanded targeting, or why layered targeting is choking delivery.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this when the task is about WHO and IN WHAT CONTEXT a line item should serve, including:

    line 15
  2. Very specific:

    If targeting looks correct but the line item still will not deliver, hand off to dv360-troubleshooting.

    line 32
  3. Distinctive:

    Narrow deliberately with AND, one category at a time: Add a second category (a demographic floor, a custom segment) only when it earns its …

    line 74
  4. Distinctive:

    Decide on optimized targeting last: If you need scale, enable it knowing it treats your first-party lists as a soft signal and may move …

    line 87
  5. Distinctive:

    AND across categories, OR within a category: Use OR (multiple lists in one group) to widen.

    line 106
  6. Distinctive:

    Use AND (a new group) to qualify.

    line 107
  7. Distinctive:

    Target predicted viewability when the KPI is viewable: If you are measured on viewable CPM or viewable impressions, set the Active View …

    line 115

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

DV360 targeting and audiences

Pick the right audiences and contextual signals, then layer them so the line item reaches the right people at enough scale to spend. This skill is the targeting brain. For where the inventory itself comes from (open auction, deals, packages) hand off to dv360-deals-and-inventory. For KPI and reach math definitions, see programmatic-foundations.

When to use this skill

Use this when the task is about WHO and IN WHAT CONTEXT a line item should serve, including:

  • Choosing between your first-party data, Google audiences, and custom segments.
  • Affinity vs in-market vs demographics vs life events.
  • Building combined audiences and deciding include vs exclude, AND vs OR.
  • Adding contextual layers: geography, language, device, environment, day and time, content categories or keywords, and content exclusions.
  • Targeting predicted viewability with Active View.
  • Deciding whether to turn on optimized targeting and what it costs you.
  • Debugging "my line item barely spends" when the cause is over-layered targeting.

Boundaries with sibling skills:

  • Inventory sourcing, deal IDs, and inventory-source targeting live in dv360-deals-and-inventory.
  • Frequency caps, brand-safety verification vendors, and content-label exclusions as a brand-safety program live in dv360-frequency-and-brand-safety. This skill covers the targeting controls; that skill covers the safety policy and caps.
  • Bid pricing against the audience you select lives in dv360-bid-strategy.
  • If targeting looks correct but the line item still will not deliver, hand off to dv360-troubleshooting.

Quick reference

Audience taxonomy and when to reach for each:

Audience typeWhat it isReach for it when
Your first-party dataFloodlight or activity-based remarketing lists, customer match lists, and PAIR audiences reconciled with a publisherYou have a site, app, CRM, or pixel signal and want to re-engage or seed, and you want a path that survives without third-party cookies
Google audiences: affinityBroad, durable interest groupsUpper funnel awareness and consideration at scale
Google audiences: in-marketActive purchase intent for a categoryLower funnel, capture people about to buy
Google audiences: demographicsAge, gender, parental status, household incomeCoarse qualification, usually layered not standalone
Google audiences: life eventsMoving, graduating, marrying, and similarTime-bound relevance windows
Custom segmentsYou define by search terms or phrases, URLs, and appsNo off-the-shelf list fits the niche you want
Combined audiencesBoolean assembly of the above into one personaYou need AND across categories or curated OR groups
Optimized targetingModel finds new converters beyond your manual listsYou want scale and will accept it reallocating spend

Combination logic in one line: within a category, lists are OR (any match); across categories, groups are AND (must match all); exclusions remove users regardless of includes; each contextual layer you add (geo, content, device) is an additional AND that shrinks reach fast.

Targeting dimensions checklist for a line item: audience, geography, language, device and OS and browser, environment (web vs app, instream vs outstream, CTV), day and time, content (categories, keywords, and exclusions), inventory source (see dv360-deals-and-inventory), and Active View predicted viewability.

Core process

  1. Start from the goal, not the lists. Map the funnel stage to the audience class first. Awareness leans on affinity plus broad context. Consideration leans on in-market and custom intent. Performance leans on first-party remarketing plus optimized targeting. This keeps you from stacking lists that fight each other.
  2. Lay the base audience. Choose the primary include layer (for example, an in-market list or a first-party remarketing list). Within that category you may add several lists; they combine as OR, so adding more widens reach. When the durable, privacy-safe activation path matters, put first-party data at the base: onboard CRM and site or app signal through Floodlight, customer match (directly or via a customer match uploader partner), or PAIR audiences reconciled with a publisher inside a data clean room. PAIR is the way to activate first-party data against a specific publisher's audience without relying on third-party cookies or a shared user identifier; the clean room returns only an aggregate match rate, and no user-level match is exposed to either side or to the exchange. Onboard those lists first, then extend outward.
  3. Narrow deliberately with AND, one category at a time. Add a second category (a demographic floor, a custom segment) only when it earns its keep. Each new category is an intersection that can cut reach by an order of magnitude, so add, check the reach estimate, then decide.
  4. Add exclusions before more includes. Exclude converters, existing customers, employees, and audiences that overlap or waste spend. Exclusion is the cheapest precision you have and prevents serving to people you do not want.
  5. Layer the contextual dimensions. Set geography, language, device and environment, day and time, and content controls. Treat each as an AND against the audience. Watch the cumulative effect: audience plus tight geo plus narrow content can starve a line item.
  6. Add Active View viewability prediction. Set a predicted viewability floor (for example, bid only on impressions at least 50 percent likely to be viewable) so you pay for impressions a human can actually see. Know this trims inventory, and that CTV is excluded from video viewability targeting by default.
  7. Decide on optimized targeting last. If you need scale, enable it knowing it treats your first-party lists as a soft signal and may move budget away from them when it finds equal-ROI traffic elsewhere. Keep geo and brand-safety controls on; those stay hard constraints.
  8. Sanity-check scale. Pull the reach or forecast estimate after layering. If it collapses, peel back the most recent AND layer rather than loosening the core audience.

Decision rules and thresholds

  • First-party first, Google audiences for scale. Spend your owned signal on the people most likely to convert, then extend with Google audiences and optimized targeting. The durable, privacy-safe path leans on first-party data and publisher-provided signals: Floodlight, customer match, PAIR audiences, and deal-level publisher signals rather than third-party lists. PAIR is the primary way to reconcile your first-party data with a publisher's without a third-party cookie or a shared user id, so a strategy built on onboarded first-party data plus PAIR survives when cookie-based and browser-signal targeting is not available. Onboard the owned data, then reach for Google audiences only to scale beyond it.
  • Affinity vs in-market is funnel position. Affinity equals durable interest for the top of the funnel. In-market equals active intent for the bottom. If the KPI is a conversion or a strong action, prefer in-market and custom intent. If the KPI is reach or awareness, prefer affinity.
  • AND across categories, OR within a category. Use OR (multiple lists in one group) to widen. Use AND (a new group) to qualify. Never express an intent as five intersected groups when one curated combined audience would do; deep intersection shrinks reach and starves learning.
  • Cap intersection depth. Combined audiences allow many includes but a limited number of intersects. Treat 2 to 3 AND groups as the practical ceiling for most line items; beyond that, reach and bid-strategy learning both suffer.
  • Recency is a lever, not a default. For remarketing lists, tighter recency raises intent and cuts reach. You can include and exclude the same list at different recency windows to isolate a band (for example, include 0 to 30 days, exclude 0 to 7 days to skip very recent visitors).
  • Target predicted viewability when the KPI is viewable. If you are measured on viewable CPM or viewable impressions, set the Active View floor rather than hoping placement quality covers it. Expect a smaller, pricier pool.
  • Mind audience size. Very small audiences (tight first-party lists, deep intersections) cannot spend a meaningful budget and starve automated bidding of data. If the audience cannot support the budget and flight, widen the audience or move budget to a line item that can.

Reference material

  • See programmatic-foundations for definitions of CPM, reach, frequency, viewability, and funnel stages. This skill does not redefine them.
  • For the structured way to express any targeting on a line item programmatically, the DV360 API models every dimension as an assigned targeting option on the line item; see the Sources for the API resource and the assign-targeting guide. Read those when automating targeting via the API or Structured Data Files; dv360-api-and-sdf-automation covers the automation workflow.

Templates and examples

Performance line item, retail SUV launch:

  • Base audience (OR within category): in-market for "SUVs" plus in-market for "Auto financing".
  • Narrow (AND, one group): custom segment built from competitor model search terms and review URLs.
  • Exclude: Floodlight converters 0 to 90 days, existing-owner customer match list.
  • Geography: target sale-region DMAs only. Language: English plus Spanish.
  • Environment: web and app display, exclude very small in-app sizes that hurt the creative.
  • Viewability: bid only on impressions at least 50 percent likely to be viewable.
  • Optimized targeting: on, because the seed lists are tight and the KPI is a configured conversion.
  • Reach check: confirm the forecast supports the daily budget before launch; if it does not, drop the custom-segment AND first.

Awareness line item, new beverage brand:

  • Base audience: affinity "Food and Dining" plus "Health and Wellness" enthusiasts (OR).
  • Demographic floor (AND): age 21 and over for an alcohol product.
  • Exclude: recent purchasers if a first-party list exists.
  • Content: target relevant content categories, exclude sensitive categories per the brand-safety policy in dv360-frequency-and-brand-safety.
  • Viewability: set a viewability floor since the KPI is viewable reach.
  • Optimized targeting: off or cautious; awareness wants predictable reach, not conversion-chasing.

Common pitfalls

  • Stacking too many AND layers. The single most common cause of a healthy budget barely spending. Audience intersected with tight geo intersected with narrow content can cut reach by one hundred times. Add layers one at a time and watch the estimate.
  • Forgetting exclusions. Without converter and customer exclusions you pay to re-serve people who already acted, and you let overlapping includes double-count. Exclusions are mandatory hygiene.
  • Treating optimized targeting as free scale. It can quietly pull spend off your best first-party audience when it finds equal-ROI traffic elsewhere. That is fine for performance, bad if you needed those exact users reached. Decide intentionally.
  • Ignoring recency. Targeting a remarketing list with no recency window mixes red-hot and cold users and dilutes intent. Set the window.
  • Setting an unreachable viewability floor. A very high predicted-viewability floor on scarce inventory can leave nothing to bid on. Tune it against delivery, and remember CTV video is excluded from viewability targeting by default.
  • Building tiny audiences for big budgets. A deep intersection of first-party lists may be too small to spend the plan or to teach the bid strategy. Match audience size to budget and flight.

Sources

Official Display & Video 360 Help and Display & Video 360 API documentation. PAIR, first-party onboarding, audiences, optimized targeting, and viewability lines verified as of July 2026; the remaining lines verified as of June 2026.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.