Programmatic foundations

The shared base for the whole library. This is the single place where terms are defined, auction mechanics are explained, and KPI math lives. Platform skills stay thin by linking here instead of repeating definitions. Use it to settle "what does this term mean", "how is this metric computed", and "which skill do I open for this task".

When to use this skill

Reach for this skill when the question is conceptual, definitional, or about routing:

A term needs a precise definition: DSP, SSP, exchange, RTB, OpenRTB, bid request, floor, deal ID, PMP, Programmatic Guaranteed, Preferred Deal, impression, viewable impression, reach, frequency, pacing, flight, IO, line item, working vs non-working media, brand safety vs brand suitability, IVT, GIVT, SIVT, Floodlight.
A metric needs a formula or interpretation: CPM, CPC, CPA, CTR, CVR, VCR, CPCV, eCPM, ROAS, win rate, viewability rate, measurable rate.
A concept needs explaining: how the auction works, first-price vs second-price, the viewability standard, the marketing funnel and the KPI for each stage.
The user does not know which skill to open. Use the routing table below.

Hand off when the question is about doing something on the platform. This skill defines win rate; dv360-troubleshooting fixes a bad one. This skill defines a bid strategy; choosing and configuring one is dv360-bid-strategy. Define here, act there.

Quick reference

The user wants	Go to
A term defined	`references/glossary.md`
A KPI formula or what it means	`references/metrics.md`
The KPI for a funnel stage	Funnel model below, full table in `references/metrics.md`
To understand the RTB auction	"How the auction works" below
To know which DV360 skill to open	"Which skill do I use" routing table below
The trader / analyst / ops split	"Three hats" below

Core process

When a question lands here:

Classify it. Definition, formula, concept, or routing. This tells you which section or reference answers it.
For a definition, pull the exact wording from references/glossary.md rather than paraphrasing from memory, so the whole library stays consistent.
For a metric, give the formula and the one-line interpretation from references/metrics.md, and compute it if the user supplied numbers. State the unit and flag the per-mille factor on CPM and eCPM.
For routing, name the target skill from the table and hand off. Do not start solving the platform task inside this skill.
Tie the answer back to the user's objective. A number is only right or wrong relative to the funnel stage and goal it serves.

How the auction works

The 100-millisecond path of one impression:

A user opens a page or app. The publisher's SSP packages the available ad slot into a bid request (slot size, position, site or app, allowed user and device signals, floor).
The SSP broadcasts that request to many DSPs over the OpenRTB protocol.
Each DSP evaluates the request against its active line items and decides whether to bid and how much. The trader's targeting, bid strategy, pacing, and frequency caps all resolve in this instant.
DSPs return bid responses (price, creative, any deal ID). The exchange runs the auction.
The highest bid that clears the floor wins. The winning creative serves and renders.

Two facts shape every downstream decision:

The open auction is first-price. The winner pays exactly what they bid, so an overbid overpays on every win. Pacing and bid strategy exist to set the right price, not a loose ceiling. (Second-price, where the winner paid one cent over the runner-up, is largely historical for open auction but still appears in some explanations and legacy setups.)
A deal ID changes the venue, not the mechanics. PMP, Preferred Deal, and Programmatic Guaranteed run over the same plumbing with a negotiated floor, priority, or guarantee attached to the ID. See references/glossary.md for each deal type and dv360-deals-and-inventory to set them up.

Viewability standard

The MRC and IAB standard, implemented in Google Marketing Platform as Active View:

Display: at least 50 percent of the ad's pixels in view for at least 1 continuous second.
In-stream video: at least 50 percent of the ad's pixels in view for at least 2 continuous seconds.
Large display (242,500 pixels or more, for example 970x250): the threshold relaxes to 30 percent of pixels for 1 second.

Served is not seen. Always read viewability rate next to measurable rate (the share of impressions a vendor could measure at all); a strong viewability rate on a thin measurable base means little. Both formulas are in references/metrics.md.

Funnel model

Match the KPI to the objective. The single most common optimization error is judging a line item by a metric from the wrong stage.

Stage	What you are buying	Primary KPIs
Awareness	Reach the right people at scale	Reach, CPM, viewability rate, VCR
Consideration	Earn interest and engagement	CTR, site visits, engagement
Conversion	Drive the action efficiently	CPA, ROAS, CVR
Retention	Keep and re-engage customers	Repeat conversion rate, CPA on retained segments

Read the objective first; it names the primary KPI. Everything else is a guardrail (for example viewability on an awareness buy) or a diagnostic. Full table with "what to ignore at this stage" is in references/metrics.md.

Three hats: trader, analyst, account ops

This library is organized around three jobs. The same person often wears all three, but the mindset and the skills differ. Knowing which hat a task needs tells you which skills apply and when work hands off.

Trader. Builds and optimizes the buy. Owns campaign structure, bid strategy, targeting, deals, pacing, frequency, and brand suitability tuning. Question: "how do I deliver the goal efficiently?" Skills: dv360-campaign-architecture, dv360-bid-strategy, dv360-targeting-and-audiences, dv360-deals-and-inventory, dv360-frequency-and-brand-safety, dv360-pacing-and-optimization.
Analyst. Measures, explains, and recommends. Owns reporting, attribution, advanced analysis, and the data behind custom bidding. Question: "what happened, why, and what should we do next?" Skills: dv360-reporting, dv360-measurement-and-attribution, dv360-advanced-analytics-adh, dv360-custom-bidding.
Account ops. Structures, QAs, governs, and automates. Owns account setup and taxonomy, pre-launch QA, troubleshooting, and API and bulk-file automation. Question: "is it set up correctly, safely, and repeatably?" Skills: dv360-account-setup-and-taxonomy, dv360-launch-qa, dv360-troubleshooting, dv360-api-and-sdf-automation.

How they hand off:

Ops stands up the account and taxonomy, then hands a clean shell to the trader (dv360-account-setup-and-taxonomy to dv360-campaign-architecture).
The trader builds the campaign, then ops runs pre-launch QA and sign-off before anything goes live (dv360-launch-qa).
In flight, the analyst measures and the trader optimizes; a bad number routes to ops for troubleshooting (dv360-reporting and dv360-pacing-and-optimization to dv360-troubleshooting).
Recurring or large-scale changes move from manual trading to ops automation (dv360-api-and-sdf-automation).

Which skill do I use

Route the user's question to the platform skill that owns it. When two could apply, pick by the hat (build vs measure vs govern).

The question is about	Open this skill
Partner, advertiser, campaign, IO, line item structure; when to split or merge	`dv360-campaign-architecture`
Fixed vs automated bidding, Target CPA / CPM / ROAS, maximize conversions or value, learning periods, why a line item is not winning	`dv360-bid-strategy`
First-party and Google audiences, audience combination, geo, device, contextual, viewability targeting	`dv360-targeting-and-audiences`
Open auction vs PMP vs Programmatic Guaranteed vs Preferred Deal, deal IDs, a deal not delivering	`dv360-deals-and-inventory`
Frequency caps, content and publisher exclusions, brand safety vs suitability, verification vendors	`dv360-frequency-and-brand-safety`
Pacing modes and math, under or over-delivery, impression loss diagnosis	`dv360-pacing-and-optimization`
Building reports, report types, the metric and dimension glossary, scheduling	`dv360-reporting`
Floodlight, Campaign Manager 360, attribution models, Brand Lift, reach and frequency studies	`dv360-measurement-and-attribution`
Ads Data Hub, privacy checks, BigQuery transfers, joining first-party data	`dv360-advanced-analytics-adh`
Rule-based, script, or Ads Data Hub custom bidding; scoring and staged rollout	`dv360-custom-bidding`
Account and advertiser setup, naming conventions, roles and permissions, governance	`dv360-account-setup-and-taxonomy`
Pre-flight QA checklist and sign-off before launch	`dv360-launch-qa`
No delivery, pacing, win rate, viewability, creative, or conversion problems in flight	`dv360-troubleshooting`
DV360 API, Structured Data Files, what is safe to automate	`dv360-api-and-sdf-automation`

The table above is DV360. For Google Ads, Amazon DSP, StackAdapt, or The Trade Desk, the same questions map to that platform's skills (the google-ads-*, amazon-dsp-*, stackadapt-*, and ttd-* sets). Cross-platform work has its own skills: dsp-selection to choose a platform, reach-and-frequency-planning for deduplicated reach across platforms, incrementality-and-experimentation to prove lift, and client-deliverable-templates for client-facing plans and reviews.

If the question is conceptual or definitional rather than a platform task, answer it here.

Reference material

references/glossary.md: the full glossary. Read it when a term needs a precise, canonical definition (the marketplace, deals, counting and reach, plan structure, safety and IVT, Floodlight).
references/metrics.md: every KPI and auction formula with interpretation, the full funnel table with "what to ignore at each stage", and worked numeric examples. Read it when you need to compute or explain a metric.

Common pitfalls

Judging a stage by the wrong KPI. CPA on an awareness buy, or reach on a conversion buy. Anchor every metric to the funnel stage in references/metrics.md.
Treating served as seen. Report viewable impressions and viewability rate for visibility goals, never raw impressions, and always alongside measurable rate.
Forgetting first-price. On open auction the bid is the price paid, not a ceiling. An overbid overpays on every win.
Confusing brand safety with brand suitability. Safety is the universal hard floor; suitability is the advertiser-specific layer. Conflating them over-blocks and loses scale.
Reading win rate without bid rate. Win rate is wins over bids submitted, not over requests seen. A high win rate on few bids is not scale.
Comparing line items on different buy models directly. Convert each to eCPM first.

Sources

Media Rating Council, Standards and Guidelines (viewable ad impression measurement guidelines): https://www.mediaratingcouncil.org/standards-and-guidelines (as of June 2026)
Google Ad Manager Help, Active View viewability (display 50 percent for 1 second, video 50 percent for 2 seconds): https://support.google.com/admanager/answer/4524488 (as of June 2026)
Display & Video 360 Help, About Active View: https://support.google.com/displayvideo/answer/3214556 (as of June 2026)
Display & Video 360 Help, Automated bid strategies: https://support.google.com/displayvideo/answer/2997422 (as of June 2026)
Display & Video 360 Help, Overview of deals: https://support.google.com/displayvideo/answer/7243138 (as of June 2026)
Display & Video 360 Help, About Floodlight and Floodlight activities: https://support.google.com/displayvideo/answer/3027419 (as of June 2026)
Display & Video 360 Help home: https://support.google.com/displayvideo (as of June 2026)
IAB Tech Lab, OpenRTB specification: https://iabtechlab.com/standards/openrtb/ (as of June 2026)
IAB, standard terms and definitions for the advertising industry: https://www.iab.com/blog/iab-unites-the-industry-with-new-standard-terms/ (as of June 2026)

programmatic-foundations

SKILL.mdskills/programmatic-foundations/SKILL.md