The Glossary

Smart Bidding is the name for four Google Ads bid strategies: Target CPA, Target ROAS, Maximize Conversions and Maximize Conversion Value. You stop setting « 2 € max per click ». You give a goal (conversions or value), and Google's AI sets the bid at the exact moment each search happens. To decide, it combines the device, the city (even if your targeting covers a whole region), the day and time, the language, the browser, your remarketing lists and the actual search query, not just the keyword that matched it.

A ready buyer and a casual browser sometimes type the same search. You can adjust device and location by hand; signal combinations are reserved for Smart Bidding, and they let it bid more on the first than on the second. The algo also learns across all your campaigns, so a new campaign with no history can still benefit. To judge results, Google recommends a period with at least 30 conversions, a month or longer (50 for Target ROAS).

You set the average amount you are willing to pay for a conversion, and Google sets each bid based on how likely the ad is to convert. With a target CPA of 50 €, some conversions cost you 30 €, others 80 €, but the average trends toward your target. If your campaign has history, Google suggests a target: your average CPA over the last 30 days, adjusted for conversions that come in late. Since June 2026 the interface calls it simply « Target CPA » instead of « Maximize conversions with a Target CPA », and it works the same way.

You steer profitability through cost: every conversion counts the same, and the algo holds their average price. It fits when you know what a lead or a sale is worth to you, and it can start with no conversion history at all. Just plan for some days where the campaign spends up to 2 times your average daily budget, without going over the monthly limit. To evaluate, compare your real CPA with the « average target CPA » shown, which includes your target changes over time, over 30 days and at least 30 conversions.

You set the revenue you want back for each euro spent. Google's own math: 5 € in sales for 1 € in ad spend gives 5 ÷ 1 × 100%, a target ROAS of 500%. The algo predicts the value of each possible conversion from the values you report, then bids high on searches likely to bring a big basket and low on the others. Some sales return more than your target, some less: it holds the average. On Search and Shopping, you need at least 15 conversions in the last 30 days. Since June 2026 the interface calls it « Target ROAS » instead of « Maximize conversion value with a Target ROAS ».

Profitability is read in value: one 500 € sale weighs more than five 20 € sales. Google recommends this approach when your conversions don't all have the same value for your business, and when you have a precise return target. You can even weight value with conversion value rules (a customer type, a device or a location that is worth more to you), so the algo goes after what really counts.

You give a daily budget, and Google aims to spend all of it to bring back as many conversions as possible, setting the bid on every search. No cost target: every conversion counts the same, whatever it costs. Your bid adjustments for location or schedule are no longer used, with one exception: you can still exclude a device with a -100% adjustment.

This is the volume strategy: filling a calendar, clearing stock, when every conversion is worth as much as the next and you don't have a target acquisition cost yet. Since it is built to spend the whole budget, it shows as « limited by budget » by design. To know whether more budget would bring more, Google advises the budget simulator rather than the « Lost IS (budget) » column, which doesn't account for how this strategy works.

Same logic as Maximize Conversions, but the algo optimizes value rather than count: it bids higher on searches likely to bring a high-value conversion. It prefers one 300 € sale over three at 40 €. You decide which value to maximize when you set up conversion tracking: the sale revenue, or your margin.

When your sales don't all have the same value, counting conversions is not enough: for Maximize Conversions, a 300 € order and a 40 € one weigh the same. Here total value guides the algo, and if you send your margin as the value, it goes straight for what makes you money. As soon as you add a target ROAS, the strategy behaves like Target ROAS: that is the next step, once you know your profitability goal.

Performance Max is a goal-based campaign type: from a single campaign, it reaches all Google Ads inventory, meaning YouTube, Display, Search, Discover, Gmail and Maps. You provide the assets (headlines, descriptions, images, videos, product feeds) and your bidding goals, and Google's AI combines them to serve where it expects the best conversions. Assets sit in asset groups: Google recommends adding as many as possible, up to 15 headlines and 5 descriptions, with images and videos in several orientations. Without a video, Google may generate one from your assets.

One campaign covers channels you used to manage separately, and the algo optimizes across them based on your conversion goals. That is why Google recommends judging a Performance Max campaign at campaign level, and using placement reports for brand safety rather than for performance evaluation. It lives alongside your Search campaigns under a clear rule: when a query matches an exact match keyword, the Search campaign takes priority. Performance Max search themes, on the other hand, carry the same priority as phrase and broad match keywords.

Quality Score is a diagnostic tool: a score from 1 to 10, available for each keyword. It combines three components: expected clickthrough rate, ad relevance and landing page experience. Each one gets a status of « Above average », « Average » or « Below average », compared with other advertisers whose ads showed for the exact same search over the last 90 days. There is no ad group, campaign or account Quality Score.

Think of it as a thermometer: Google states that the score itself is not an input in the ad auction. What counts at auction time is ad quality, and the score sums it up. The link with your budget is real: according to Google, higher quality ads typically cost less per click. So the score tells you where to dig, and each component below average points to a specific job: the ad's hook, its match with the keyword, or the page after the click.

Customer value × margin × closing rate. 2,000 € customer, 40 % margin, 25 % closing: 200 € max per lead. Above that, you lose money on every signature.

Without this ceiling, no CPL is "good" or "bad": 60 € per lead is excellent for a 15,000 € project and unsustainable for a 500 € offer.

Customer acquisition is one of the customer lifecycle goals in Google Ads. It offers three settings: "New Customer Value" (bid higher for new customers than existing ones), "High Value New Customer" and "New Customer Only". To recognize an existing customer, Google relies on your first-party data, a customer list and your site tag. You can also tell it on every purchase with the new_customer parameter of the conversion tag: true for a buyer with no purchase over the chosen period (540 days by default), false for a known customer.

Without this setting, a first order and a repeat purchase weigh the same in your bids. With "New Customer Value", you add extra value to every new customer, and the algorithm goes after it: your bids work to grow your customer base, not only to sell again to people who already know you. This setting needs value-based bidding (Target ROAS or Maximize conversion value) and a Purchase conversion goal.

The Conversions API creates a connection between your marketing data (website or app events, messaging events, offline conversions) and Meta's systems, straight from your server, website platform, app or CRM. Meta processes these server events like the ones sent by the Meta Pixel. The recommended setup is redundant: the same events go through both the Pixel and the API, and Meta deduplicates them. For that, the event ID and the event name must match on both sides.

The Pixel depends on the browser, and Meta acknowledges it can lose events due to network connectivity issues or page loading errors. The API sends the same ones from your server, like a second line that takes over when the first one drops. Meta presents it as a way to optimize ad targeting, decrease cost per result and measure outcomes. When both versions of an event arrive within 5 minutes of each other, Meta keeps the browser one.

It is your Monday morning checklist, except it runs by itself. A Google Ads script is a JavaScript program written in the account's editor: you paste it in the Scripts section, authorize it, schedule it (once, daily, weekly or monthly, at the hour you pick), and it watches your budgets, spots disapproved ads, suggests negative keywords or builds a report. An account script stops after 30 minutes. At manager account (MCC) level, one script can access all the managed accounts, and process up to 50 of them in parallel.

A check you do by hand always ends up skipping a week. The script shows up every time, and it warns you the same day a budget drifts or an ad gets disapproved. An account can hold up to 250 authorized scripts: each recurring check can have its own, and on a manager account, a single script covers your whole portfolio.

Server-side tagging adds a second Google Tag Manager container, hosted in a cloud environment, next to the web container installed on your site. The browser sends events to this server container, which you own, and it then forwards them to Google, Meta or your other tools. Until you choose to send the data elsewhere, you are the only one with access to it. Set up in a first-party context, under your own domain, it keeps site data and cookies within that domain.

Picture an airlock between your site and the ad platforms: every piece of data goes through it before leaving, and you hold the door. In the server container you can remove personal information before passing it to marketing partners, and you decide the shape and destination of each piece of data. Cookies set from your server can also carry the HttpOnly flag, which, according to Google, makes them more durable and more secure.

Consent mode passes each visitor's cookie choice to Google, and Google tags adjust their behavior to that choice. It doesn't provide the banner: your consent management platform shows it and passes the information along. It comes in two versions. In basic mode, nothing is sent to Google until the visitor accepts, not even the refusal. In advanced mode, tags send cookieless pings when consent is denied, and Google uses them to model the conversions it can't observe.

In the European Economic Area, Google requires you to collect consent and pass these signals on to keep measuring, personalizing ads and running remarketing. Consent mode is the channel that carries them. The two versions don't model the same way: basic mode relies on a general model, advanced mode on a model specific to your account. According to Google, users who consent are typically 2 to 5 times more likely to convert than those who don't, which is why Google models the relationship between the two groups instead of extrapolating from the first one.

Google defines first-party data as information a customer shares directly with you: on your website, in your app, in store, through offline conversions, wherever they interact with your business. You can load it into ad platforms. At Google, Customer Match uses it to reach your customers across Search, the Shopping tab, Gmail, YouTube and Display. At Meta, a customer file builds a Custom Audience, and the data must be shared hashed.

This is data you own, and it serves measurement as much as targeting: enhanced conversions, for example, send first-party conversion data hashed with the SHA256 algorithm for more accurate measurement. Google describes it as essential to fuel AI and accurate measurement. And Chrome has not removed third-party cookies: in 2024, Google dropped that plan and left the choice to users. Its value lies in where it comes from: your real customers.

BigQuery is Google Cloud's fully managed data platform: you store tables there and query them in SQL, with no server to run. GA4 can export all its raw events to it, and the Data Transfer Service copies your Google Ads reports every day, with a 7-day refresh window by default. Your CRM arrives as a file. On pricing, the first 1 TiB of queries and the first 10 GiB of storage are free each month, per billing account.

This is where ad spend and the sale end up on the same line. A query can link a campaign's cost to the revenue signed in your CRM, something no interface does on its own. Without it, you compare an ad cost to a form, never to a signed customer. And since Google states that data exported from GA4 can't be exported again, every day that lands in BigQuery becomes history you keep.

Enhanced conversions add to the conversion you already measure. When a customer converts, the first-party data entered on your site, such as their email, name, address or phone number, is hashed with the SHA256 algorithm, sent to Google in that form, then matched with the Google accounts that were signed in when they engaged with your ad. It exists for web and for leads: starting in June 2026, Google combines the two into a single feature with a simple on/off switch.

Some conversions slip past the tag alone. By matching hashed data with Google accounts, enhanced conversions help recover them: Google presents the feature as a way to improve the accuracy of your conversion measurement and unlock more powerful bidding. The data stays protected by one-way hashing, so Google receives a fingerprint and never the address in plain text. Allow about 30 days before you see its impact in your reports.

Google Tag Manager is a tag management system: you add two code snippets to each page of your site, once, then manage all your tags from the interface. The set of tags, triggers and variables installed on a site is called a container. A tag sends information to a tool (Google Ads, GA4, Meta), a trigger is the rule that decides when it fires, and the data layer is a JavaScript object that temporarily stores what the visitor does. Nothing changes on the site until you publish your changes.

It is the dashboard of your tracking: instead of asking a developer to add each tracking code, you add, edit and test them yourself. Preview and debug mode lets you browse your site as if the draft were already live, so you check before you publish. And each publication creates a version, a snapshot of the container that lets you roll back if a change breaks something.

GA4 brings together your website and app data, and measures it through events rather than sessions. An event measures a specific interaction: a page view, a click, a submitted form, a purchase. The ones that matter to your business are called key events, and the word « conversion » now means an action you use to measure your campaigns and optimize your bidding. Once processed, the data is stored in a database where it can't be changed.

This is where you understand what visitors do after the click, and where your buyers come from. The event model follows a customer's journey from site to app. And since data can't be changed once processed, what you set up today decides what you will be able to read tomorrow: a forgotten event can't be recovered afterwards. Custom events don't show up in most standard reports: you need custom reports or explorations to analyze them.

Firebase brings together the services a team needs to run an app: user sign-in, databases, storage, push notifications, crash reports, and Google Analytics. A Firebase project is a Google Cloud project with extra Firebase settings, and the iOS and Android versions of the same app each become a data stream in the same GA4 property. For an advertiser, the key piece is the Google Analytics for Firebase SDK: it collects events inside the app, like the first open or a purchase made through the stores, which you can then import into Google Ads as conversions.

On Android, the Google Play link already tracks downloads, Google Play purchases and pre-registrations with no code. Everything else, the first open, the return to the app, a sign-up, goes through Firebase or a third-party app analytics tool. That is the difference between buying installs and buying users: an App campaign can only bid on what you send back to it. On iOS, Google's on-device conversion measurement also runs through this SDK.

The GCLID (Google Click Identifier) is a URL parameter passed with clicks on your ads: it identifies the campaign and the other attributes of the click, for tracking and attribution. It appears when auto-tagging is turned on, and it is required for website conversion tracking. Sometimes it is created at the impression: if the person clicks the same ad again, the same GCLID is reused. To measure without identifying people, Google also uses two other parameters: GBRAID for clicks tied to app conversions, WBRAID for web conversions.

It is the thread that ties a sale to a click, even when the sale closes far from the site. To import offline conversions, a signed quote or a call, Google requires auto-tagging: you keep the GCLID from the form all the way to your CRM, then send the sale back with it. The original click has to stay within the conversion window, which you can set from 1 to 90 days depending on the source.

A custom variable gives a conversion event a dimension specific to your business: the project type of a loan request, the product range, the budget bracket of a quote request. There are two ways to create one. Either you add a parameter to your conversion tracking tag: Google Ads picks it up on its own, and you then need to "activate" it in the account for the data to be recorded. Or you add it as a column to the offline conversions you import. Once activated, it segments your reports within a few hours.

Google Ads tells you how many forms a campaign brought in, not which ones. With a variable, you see which keywords bring the big projects and which bring the small ones, while the Conversions column shows the same number for both. Google's own example is a hotel chain that splits its bookings by hotel, room price category and loyalty status, to learn which generic keywords lead to the most expensive rooms.

Your purchase tag sends, with every order, the list of items sold: their ID, price and quantity. Google links each item to your Merchant Center feed and produces reports the conversion alone cannot give you: number of orders, average cart size, revenue, and even gross profit if your feed carries the cost of goods sold. Only sales of items from a Merchant Center account linked to the Google Ads account are processed.

The standard report tells you a purchase happened; cart data tells you what the customer bought. You see which campaigns sell the most and at what margin, whether location or device changes cart size, and whether the customer bought the advertised product or something else from your store. That is what lets you, in Google's words, raise bids on the products that drive higher-value purchases.

When someone clicks your ad, the landing page URL carries the details of that click. The conversion linker tag detects them and stores them in first-party cookies on your domain, the _gcl_* ones such as _gcl_aw, and in the browser's local storage. When the visitor converts further along, the conversion tag finds that click and links the conversion to it. Google recommends firing it on "All Pages". It can also carry the click across several of your domains, by adding a parameter to the links that lead there.

The click lands on one page, and the conversion often happens on another, sometimes much later in the visit. With no trace of the click in between, the conversion tag has nothing left to link to. One point Google's documentation makes clear: if your container already loads a Google tag on every page, it does not also need a conversion linker tag, and a container with Google Ads tags loads one automatically before sending its events.

An ad does not always lead to an online sale: often it starts a journey that ends by phone, in a meeting or with a signed contract. You keep the click ID, the GCLID, with the lead in your CRM, then send Google Ads that GCLID, the conversion type and its date when the lead converts. On the API side, these imports go through the Data Manager API since June 15, 2026, and Google advises anyone starting out to choose enhanced conversions for leads instead, its upgraded version, which relies on user-provided data such as the email address.

Without import, Google Ads optimizes on the form that was filled, and a curious visitor's form counts as much as a future customer's. With import, every stage of your CRM can become a conversion, from qualified lead to signed contract, and the algorithm learns from what actually pays. Google reports a median 10% increase in conversions for advertisers who add first-party data to their imported GCLIDs, compared to standard offline imports.

ROAS (return on ad spend) divides the value of your conversions by what you spent. You put 1,000 € into Google Ads and it brings in 4,000 € of revenue: your ROAS is 4, or 400%. In Google Ads, the column that shows it is called « Conv. value / cost ». The number depends entirely on the value you report through conversion tracking: the sale amount, or your margin if you choose to send that instead.

It is the metric that tells you at a glance what each euro of advertising brings back, and it compares easily from one campaign to another. It is also what Target ROAS steers: the strategy aims for an average value per euro spent. Keep it apart from ROI, which Google defines as the ratio of your net profit to your costs: two campaigns with the same ROAS can leave very different profits.

100 € sale, 60 € product cost, 20 € ads: ROAS = 5, POAS = 2. Same campaign, two stories: ROAS looks at revenue, POAS looks at what you keep.

On a catalog with mixed margins, two campaigns at the same ROAS can have opposite profitability. POAS aligns your bidding with real profit.

Break-even ROAS = 1 / gross margin. 40 % margin: floor at 2.5. 25 % margin: floor at 4. The first number to set before judging any campaign.

Without a floor, a "good ROAS" does not exist: a ROAS of 3 is profitable at 40 % margin and losing money at 25 %. This threshold turns the ROAS column into a decision.

100,000 € of revenue, 20,000 € of ads across all channels: MER = 5. Every ad euro generates 5 € of revenue, no matter whether Google or Meta claims the sale.

Platforms often claim the same conversions. If your platform ROAS climbs while your MER declines, they are claiming sales that would have happened anyway.

Acquisition spend / new customers. Store with an 80 € average order and 40 % margin: above a 32 € CAC, the first order does not pay you back; customer lifetime value decides whether you can accept it.

A stable CPA can hide a degrading CAC when the share of repeat purchases grows: the account looks healthy while acquiring new customers gets more and more expensive.

600 € LTV, 150 € CAC: ratio of 4. Benchmarks: ≥ 3 profitable, = 1 break-even, < 1 every acquired customer loses you money.

It is the global ROAS of the customer relationship, not of one campaign. It allows (or forbids) acquiring customers at a loss on the first order to recover over time.

« Save 50 € » converts less than « Do not lose your 50 € discount ». A cart countdown, « only 2 left in stock », a free trial ending: all of it triggers the fear of losing, which drives more than the desire to gain.

Understanding this bias changes how you write an ad or an offer: you frame the message around what the prospect risks missing, not only what they gain.

A crossed-out price of 199 € before the real 99 € makes 99 € look cheap, because 199 € became your anchor. Without the anchor, 99 € might have looked expensive.

Anchoring shapes a pricing page, a quote, a negotiation. Showing the premium offer first anchors high and makes the standard offer look reasonable.

« 12,000 customers », star ratings, known brand logos, « 8 people are viewing this product »: all signals that say « others trusted this, you can too ». The brain takes a shortcut instead of judging alone.

One of the strongest conversion levers on a page. Placed well, next to the action button, social proof removes the last hesitation.

Google defines it as a program or system that trains a model from data, to make useful predictions on new data. Where a developer codes a function by hand, a model learns its parameters during training. You don't program « if the user is 25-34, bid higher »: you show the algo thousands of past conversions, and it figures out on its own which profiles convert, often combinations a human wouldn't have guessed.

It is what powers Smart Bidding: according to Google, its algorithms train on data at a vast scale to predict how each bid amount affects your conversions. It also builds GA4's predictive audiences, provided at least 1,000 returning users met the target condition and 1,000 others didn't. The cleaner the data it gets, the better its predictions hold, which is why Google recommends judging Smart Bidding on at least 30 conversions.

Trained on billions of texts, it learns the regularities of language. You write a prompt, it generates the most likely continuation, word after word. From this mechanism come summary, translation, writing, code.

For a marketer: writing ads at scale, analyzing customer verbatims, sorting search terms, brainstorming angles. A productivity lever, with your judgment still in charge.

The algo serves, sees who converts, adjusts, serves better, watches again. Every reported conversion feeds the loop. Broken tracking or false conversions, and the loop learns the wrong pattern.

This is why reliable tracking is a direct competitive edge: a competitor with better data trains a better algo, and the gap widens on its own over time.

Five groups, always in the same order: innovators (2.5%), early adopters (13.5%), early majority (34%), late majority (34%) and laggards (16%). To place a topic, listen to the word people use. When the word was « Large Language Model », only innovators talked about it. Then it became « generative AI », then « AI », and today « ask the AI »: the topic has reached the majority.

In advertising, a new placement has few advertisers, so the entry cost is often low. It rises as the majority arrives. Knowing where a channel sits on the curve helps you pick the moment to test it, with a test budget and clean measurement.

OpenAI's Advertiser API manages campaigns, ad groups, ads, product feeds, conversion tracking and reporting. In each ad group, you describe in sentences the products, use cases or needs your offer covers: these are context hints, up to 2,000 per ad group. OpenAI states that they are not exact-match keywords, and that they don't replace geographic, platform or audience targeting. The ad title fits in 50 characters, the body in 100.

It is a new placement where intent is expressed in conversation. Context hints supply context instead of fixing a list of queries: you describe your customer's situation, and that information adds to what your ad and landing page already say. Platform targeting lets you choose the ChatGPT surfaces where the campaign can deliver.