Glossary

How campaigns, ad groups and ads are organized inside an ad account, by objective, market, budget and audience. Fewer, larger groups give the platform more data to learn from.

The check a platform runs on an ad against its policies before and while it runs. An ad that fails is rejected or limited, and can be fixed and sent back for review.

The checks a platform runs on the business or person behind the ads, with documents that must match the name on the billing profile. Ads can be paused until it is complete.

A way of buying in which the price is set in real time, impression by impression or click by click, among the buyers bidding at that moment. In self-serve auctions the platform weighs each bid against how likely the ad is to get the result asked for and how people respond to it.

A company a publisher lists publicly as allowed to sell its inventory, so a buyer can check that an impression comes from a legitimate seller.

A bid strategy that sets the most the platform may bid in any single auction. It controls the price and can limit volume.

The rule a platform's model follows when it bids, such as getting the most results for the budget, holding an average cost per result, staying under a maximum bid or aiming at a return on spend.

Brand safety keeps ads away from content no brand should appear next to. Suitability goes further and sets which contexts fit a given brand.

Searches that include the advertiser's own name. Ads on them catch demand that often exists already, so what they add is tested.

Ads a platform builds from a product feed, showing each person the products most relevant to them.

A bid strategy that asks the platform to keep the average cost per result near a set amount. Delivery slows when results cost more.

The price paid for each click on an ad, the unit most search and some social auctions charge on.

The price paid per thousand ad impressions, the unit most display, video, audio and connected TV inventory is priced in.

A block of hours in the broadcast day, such as morning drive or prime time, by which TV and radio time is sold and priced.

An insertion order placed directly with a publisher or media owner, with delivery terms agreed in writing.

A channel whose formats, access or measurement are still settling. It is tested with a dedicated budget, apart from the core plan, and reported on separately.

Scheduling spend in bursts with gaps, as opposed to continuous presence.

The period after a campaign starts or changes significantly, while the platform's model is still learning whom to show the ad to. Results vary more, and a large edit can start it again.

Replacement delivery a seller provides when a guaranteed buy falls short of what was agreed.

The conversion a campaign asks the platform to find more of, such as a purchase, a qualified lead or an install. It needs enough volume for the model to learn, and should sit as close to the sale as that volume allows.

Paying only for an agreed outcome, such as a sale, a lead or a call, through cost per action, cost per lead, revenue share or pay per call.

Where an ad appears inside a platform or its network, such as a feed, short video, search results or the sites and apps the platform sells.

Invitation-only auctions or fixed-price access to chosen publishers, without guaranteed volume.

Price and volume agreed in advance with a seller and delivered through automated buying.

Inventory a seller has not sold as air date approaches, offered close to air at a lower price and often used for direct response.

Reach and frequency booked in advance at a fixed cost, so delivery is planned before the campaign runs.

A product or service, such as financial services, health products or alcohol, that a platform allows only under extra rules or with its own prior authorization.

Showing ads to people who already visited the site or app or interacted with the brand. Many of them would buy anyway, so what it adds is tested.

Television inventory bought closer to air date, outside the upfront commitments.

An ad platform where campaigns are set up, funded and run directly, and bought in the platform's own auction.

A flat fee for association with a program, team, event or venue.

A TV or streaming format that shrinks the program into part of the screen to show an ad beside it.

Where a channel works on the way to a purchase. Discover reaches people who don't know the brand yet. Consider helps them compare. Convert turns intent into a sale, a lead or an install. Retain brings customers back. The catalog can be filtered by stage.

The chain of companies an impression passes through between the publisher and the buyer.

A bid strategy that asks the platform to bid for conversion value, aiming at a set ratio of value to spend. It works only as well as the value it receives.

Commitments made before a television season, in exchange for priority, pricing and audience terms.

The way a mix model represents carryover, letting an ad's effect build and then fade over time.

Assigning credit for an outcome to the media that contributed to it. Each method answers a different question, and none is complete on its own.

The period after an ad interaction during which a conversion can be credited to it. It is agreed before launch and disclosed with the results.

Total revenue in a period divided by total ad spend across every channel. It credits no single platform, so it is read next to each platform's own numbers as a check.

The change in awareness, consideration or preference that a campaign causes, measured by surveying people who saw it and a control group who didn't.

Adjusting a model so it agrees with experiment results.

The part of an ad's impact that shows up days or weeks after it runs.

Sales matched to ad exposure inside the same retailer or network.

The share of video or audio ads played to the end.

An experiment in which a platform holds back a random share of an audience from the ads and compares its conversions with those of the people who saw them.

Ad spend divided by the number of chosen actions, such as purchases, leads or installs. It is read with the attribution window it was counted in.

The drop in results when the same people have seen an ad too often. It shows in a rising cost per result and falling response.

Comparing ads, or the ideas behind them, under controlled conditions and on the result the campaign is bought for.

A model that splits credit using statistical estimates of each interaction's contribution. It is still observational, so it shows patterns in the path rather than proof of cause.

Promo codes, web addresses and phone numbers unique to a channel or placement, so responses can be traced back to it.

The average number of times each reached person was exposed.

A test where whole regions are assigned to see or not see a campaign, then compared to measure lift.

Total delivery of a schedule against a target audience, the sum of the rating points of every spot. A rating point is a spot's audience as a share of the target population, and it is the currency of TV and radio.

People or customers deliberately not exposed to a campaign, used as a comparison.

In digital out of home, the estimated number of people who see one play of an ad, applied to every play to count impressions.

People reached by adding a channel who were not reached by the others.

The additional outcomes caused by marketing, compared with what would have happened without it.

A model that gives all credit to the last ad clicked before a conversion.

The difference in outcomes between exposed and control groups. A lift study sets up that comparison to measure the change an ad caused.

An estimate of how many of the people with an opportunity to see an out-of-home ad are likely to have noticed it, adjusted for the size, position and visibility of the site.

The outcome expected from the next dollar spent, as opposed to the average dollar spent so far.

A statistical model that estimates what each channel, online and offline, contributes to results, using spend and outcomes over time and no individual tracking.

Test and control markets chosen because they behave alike.

The smallest change a test is designed to detect reliably. It is set before spend, together with the size of the control group and the duration.

Customers who buy a brand for the first time within a set look-back period, as reported by a retailer or commerce network.

The number of people who pass an out-of-home site or screen and could see it, before any adjustment for whether they noticed it.

A study that links exposure to an ad, often on TV or streaming, to later outcomes such as site visits, store visits or sales, by matching exposed households with outcome data. Its results are estimates.

Reporting results in aggregate, with added statistical noise where needed, so campaigns can be measured without exposing any individual. On mobile devices, app installs and events are reported this way where a person hasn't agreed to tracking, and through measurement SDKs with consent where they have.

A record that the ad ran on the screen at the time and place agreed.

The number of different people or households exposed at least once.

The estimated relationship between spend in a channel and the outcome it produces, including the point where each extra dollar returns less.

Conversion value divided by ad spend, as a platform or a model reports it. It shows what was credited. What the media caused takes an experiment.

Diminishing returns as spend in a channel increases.

A modeled comparison built from other markets or periods to estimate what would have happened without the campaign.

Combining attribution, experiments and mix modeling to reach one decision.

A conversion credited to an ad that was seen but not clicked. It is reported apart from conversions after a click.

An ad impression that met a minimum on-screen time and visibility. Viewability is the share of impressions that meet it.

Tags added to the link of an ad that record its source, campaign and creative, so visits and sales can be read campaign by campaign.

A code an ad platform adds to a link so a later conversion can be tied back to the click that started it.

The record of what a person has agreed to, attached to each event so every downstream system knows what it may do with it.

Machine-readable information attached to an image, video or audio file that records how it was made and edited.

The value attached to a conversion, such as order value or margin, sent with the event so bids can aim at value as well as volume.

A list of the advertiser's own customers, hashed before it is shared and used only with the consent it needs, that a platform matches to its users to reach them, leave them out or find people who behave like them.

A controlled environment where two companies match and measure combined data without handing raw customer records to each other.

Collecting and keeping only the data a stated purpose needs, for only as long as it needs it.

A central store where events, spend and outcomes are kept in a form that can be queried and audited.

Placing ads into a podcast or a stream at the moment it is played, instead of recording them into the file, so each play can carry a different ad and be counted.

Making sure an action reported by both the browser and the server is counted once, usually by sharing one event ID across both.

The agreed structure of every event, with its names, fields, formats and required values. It lets errors be caught before they reach a report.

The address on the advertiser's own domain that receives events, so data is collected in the advertiser's context and not a third party's.

Information a company collects directly from its own customers and visitors, through its own sites, apps, stores and service channels.

Turning an identifier such as an email address into a fixed code with a one-way function before it is shared. It protects the value in transit. The same input always gives the same code, so hashed data still points to one person and is treated as personal data.

Linking records that belong to the same person or household. Deterministic matching uses shared identifiers. Probabilistic matching uses statistical likelihood and is reported as an estimate.

Impressions or clicks that do not come from a real person with real interest, from automated bots to deliberate fraud. Industry practice is to detect it and keep it out of reporting and billing.

The share of events a platform can connect to its own users. Only matched events can be used to measure and optimize.

A sale, signed contract or store visit that happens away from the website and is sent back to the ad platforms from the CRM or point of sale.

A browser or device setting that tells websites a person does not want their data sold or shared for advertising. Systems should read and apply it automatically.

A file of the advertiser's products, with titles, prices, availability and images, from which shopping listings and catalog ads are built. Its quality decides where and how the ads show.

Stitching ads into a streaming program on the server before it reaches the screen. Measurement then relies on server logs rather than the device.

Measurement in which events go first to a server on the advertiser's own domain, which checks, cleans and routes them, instead of going from the browser straight to each vendor.

A channel offered by ad platforms to receive conversion events directly from an advertiser's server, including sales and leads that close offline or days later.

The gap between what happened and what measurement can see, caused by browser limits, app privacy rules, consent choices, ad blockers and technical failures.

The range around the plan, set with the client, inside which pacing and bids can move without a person's approval. Any move beyond it waits for one.

Software that uses a language model to plan and carry out a sequence of steps with tools, such as pulling data, drafting a plan or checking delivery, within limits set by people.

Automated checks that flag spend, delivery or results moving outside their expected range, so a person can act.

A time-stamped record of every automated and human action, with the data and the person or system behind it.

Bidding in which an ad platform's model sets the bid for each auction toward a goal the advertiser defines. It works only as well as the conversion data it receives.

The interval within which a forecast outcome is expected to fall. A planning number without a range hides its uncertainty.

The limits an automated system cannot cross, such as budgets, bid ranges, brand rules, audiences and approval thresholds.

A design in which defined decisions wait for a person's approval before they take effect.

Coordinating agents, tools and people across channels so each task runs in the right order, on shared data, with one record of what happened.

A rule that makes an agent stop and hand the case to a person, for example when a check fails, a limit is reached or the data looks wrong.

The currency an ad account is billed in, set when the account is created.

The place on an ad platform where campaigns run, billing is set and spend is recorded. Each one is billed in a single currency, set when it is created.

One invoice that covers spend across several platforms and accounts.

The average a campaign may spend per day. Some platforms spend above it on a busy day and below it on a quiet one, within a limit they publish for the week or the month.

A total set for the whole flight, which the platform paces across the days the campaign runs.

Controlling how fast a budget is spent across a flight, so it is neither used up early nor left unspent.

An amount a platform returns or offsets, for example for invalid traffic, matched back to the line it belongs to.

Funds deposited in advance that pay for campaigns as they run.

Matching what was planned, delivered, billed and paid, and resolving the differences.

The line-by-line view of a plan, where each line shows what was planned, delivered, billed and paid, and is closed when the four agree.

One company you contract with and pay for media across several platforms.

The most a campaign, account or balance can spend in a period, set before it runs.

A share of a balance assigned to one brand, team, client or market, with its own limits.

Checking the order or plan, the proof of delivery and the invoice against each other before a payment is made.