The vocabulary of modern measurement, explained in terms that inform the business decision. In 2026, tracking is no longer a snippet on the page — it is a system of first-party data, server-side collection and orchestrated consent. Each term has a short technical definition and, where it matters, a note on what it means in practice.
The current generation of Google Analytics, built on an event-based data model rather than the sessions and pageviews of Universal Analytics. Everything — a click, a scroll, a purchase — is an event with parameters. Universal Analytics was shut down for good in 2024.
If you still think in “sessions and bounce rate”, you think in the old paradigm. GA4 measures behavior at the event level — a different way to read the data.
The tag management system: a single interface to add, edit and fire tracking scripts (GA4, pixels, conversions) without touching the site code each time. The logic of “what fires, when, and with what data” lives in a container, not scattered across pages.
The single Google tag that sends data to GA4, Google Ads and other Google products from one snippet. Replaces the old separate libraries (analytics.js, conversion.js) with a single codebase.
A structured JavaScript object that sits between the site and GTM, carrying clean data (order value, product ID, login status) to tags. It is the source of truth tracking relies on — a well-built data layer means reliable data; a bad one means fragile measurement.
90% of tracking problems start from an incomplete or inconsistent data layer. This is where data quality is won or lost.
The fundamental unit of data in GA4: a measured action (page_view, purchase, generate_lead, add_to_cart) with attached parameters. Some are collected automatically, some recommended by Google, others custom, defined by you.
The unique identifier of a GA4 property (format G-XXXXXXX) that ties collected data to the right property. The GA4 equivalent of the old “UA-” from Universal Analytics.
The package that holds all of a site's or app's tags, triggers and variables in GTM. A web container runs in the browser; a server container runs in your infrastructure — the key distinction for server-side tracking.
The rule that tells a tag when to fire: on page load, on a click, on a form submission, on a data layer event. A misconfigured trigger is the typical cause of conversions counted twice or not at all.
A reusable container for a value used by tags and triggers — from an order value read from the data layer to the current page URL. Variables make the setup modular instead of hardcoding values everywhere.
The testing mode that lets you see, before publishing, exactly which tags fire, with what data and in what order. Skipping this step is the most frequent cause of an implementation that “looks” fine but sends wrong data.
Moving data collection and distribution out of the browser onto a server you control. Instead of dozens of third-party scripts running in the user's browser, the browser sends one request to your server, which then distributes the data to GA4, Google Ads, Meta and so on. The 2026 standard for serious advertisers.
Recovers the data browsers, ad blockers and ITP lose — implementations report ~37% more tracked conversions and up to +46% on Google Ads with Enhanced Conversions.
Data you collect directly from your customers, with their consent, through your own channels: orders, accounts, sign-ups, on-site behavior, CRM. Unlike third-party data, it belongs to you, is more accurate and doesn't depend on third-party cookies. The central measurement strategy in 2026.
With third-party cookies declining and Privacy Sandbox abandoned, first-party data is no longer an edge, it's the foundation. Reverse ETL on it reports CAC −15-30% and ROAS +25-40%.
A cookie set by the domain the user is visiting (yours). Unlike third-party cookies, it is less affected by browser restrictions and has a longer lifespan — especially when set server-side, on your own domain.
A cookie set by a domain other than the one being visited, historically used for cross-site tracking and retargeting. Google abandoned its plan to remove them from Chrome in 2024 (confirmed in 2025), but their effectiveness is declining anyway: Safari and Firefox block them, ITP shortens them, and opt-outs are rising.
The myth “third-party cookies are dead” is false — but “everything's fine” isn't true either. They're alive but unreliable. That's why strategy shifts to first-party, not because they vanished.
Measurement methods that don't rely on third-party cookies: first-party data, your own identifiers, statistical conversion modeling and server-side collection. It doesn't mean “no cookies at all”, but “no dependence on vulnerable third-party cookies”.
Connecting GA4 to BigQuery, Google's data warehouse, to export raw, event-level data without sampling. It gives full control: you can query, join with business data and build reports the GA4 interface doesn't allow.
The reverse of classic ETL: instead of bringing data into the warehouse, you send data from the warehouse back into operational tools (Google Ads, Meta, CRM, email). This is how you activate first-party data — for example syncing high-LTV customer lists to ad platforms.
A subdomain you own (e.g. metrics.site.com) through which server-side collection runs, so requests and cookies are treated as first-party. It extends cookie lifespan and reduces browser blocking.
The central store where data from all sources (web, CRM, ERP, ads) is gathered in one queryable place — BigQuery, Snowflake, Redshift. For tracking, it is where behavioral data meets business data and becomes actionable.
Google's mechanism by which tags adjust their behavior based on the user's consent. Mandatory since March 6, 2024 for EEA/UK traffic using Google Ads, GA4 or Floodlight. It passes four boolean consent signals to Google.
Without it, you lose remarketing and part of measurement in the EEA. Implemented correctly, with modeling, it recovers 30-50% of conversions lost to consent refusal.
One of the four Consent Mode v2 signals: it controls storage of advertising cookies and identifiers (e.g. for remarketing). “Denied” stops these cookies from being set.
The Consent Mode v2 signal that controls storage of analytics cookies and identifiers (GA4). On “denied”, GA4 switches to cookieless measurement and modeling.
The signal introduced in v2 that governs sending user data to Google for advertising purposes. Adding it (alongside ad_personalization) is the reason v2 became mandatory, not optional.
The Consent Mode v2 signal that governs using data for personalized advertising and remarketing. On “denied”, remarketing audiences no longer populate with that user.
The EU data protection regulation. For tracking, it requires a legal basis (usually consent) before setting non-essential cookies and processing personal data. It is the framework that makes Consent Mode and CMPs necessary, not optional.
The platform that collects and stores users' consent choices (the cookie banner) and passes them on to tags and Consent Mode. OneTrust, Cookiebot, Usercentrics are common examples.
Transparency & Consent Framework — the industry standard (IAB Europe) for communicating consent between sites, CMPs and ad vendors in a common format. Google requires it for certain ad integrations in the EEA.
A signal sent by the browser expressing the user's preference not to be tracked / have data sold. Relevant mainly for US regulations (CCPA). In 2026, a correct setup orchestrates GPC together with Consent Mode v2 and TCF.
The set of APIs Google proposed (Topics, Protected Audience/FLEDGE, Attribution Reporting) as a replacement for third-party cookies. All 10 remaining APIs were deprecated in October 2025. Worth knowing precisely because it is no longer the future — strategy shifts to first-party data.
Many old articles still present Topics and Protected Audience as the “cookieless future”. They aren't anymore. Any strategy relying on them is outdated.
Transforming personal data (email, phone) into an irreversible string before sending it to platforms, so matching happens without exposing the raw data. It underpins Enhanced Conversions and compliant customer-list uploads.
The valuable action you measure as an outcome: a purchase, a lead, a booking, a download. The entire tracking system exists to count conversions correctly and attribute them to the right source.
Sending conversions directly from your server to the platform (Meta, TikTok, Google), server-to-server, instead of relying only on the browser pixel. It recovers conversions lost to ad blockers and browser restrictions and improves matching.
The browser pixel alone loses 20-40% of signal. CAPI plus server-side collection closes that gap — directly into the quality Smart Bidding sees.
Google's feature that uses hashed first-party data (email, phone) provided at conversion to improve matching and recover conversions that would otherwise be lost. It combines naturally with server-side tracking.
Importing into ad platforms conversions that happen offline or later (a lead that becomes a customer after a call, a showroom sale). It ties the online click to the real business outcome — essential for bidding on value, not raw leads.
The rule by which credit for a conversion is split across the touchpoints along the customer journey. From simple models (last-click) to data-driven ones — the choice radically changes which channels appear effective.
The attribution model that uses your actual data and algorithms to distribute credit by each touchpoint's real contribution, not a fixed rule. The default model in GA4 and Google Ads since 2023.
A conversion attributed to an ad the user saw but didn't click before converting. Useful for assessing the impact of visual ads (Display, video), but easy to overvalue if not read with caution.
The mechanism by which a conversion sent on two paths (browser pixel + server CAPI) is counted once, via a shared event ID. Without deduplication, hybrid tracking inflates numbers and breaks optimization.
The Google tag that reads click information (e.g. GCLID) and stores it in a first-party cookie so a later conversion can be tied to the click. Its absence is a frequent, easily-missed cause of unrecorded conversions.
Conversions statistically estimated by Google for users who can't be measured directly (refused consent or cookies). With Consent Mode v2 plus server-side, Google can model and recover 60-70% of conversions lost to consent refusal.
The degree to which your tracking data is complete, correct, consistent and current. A bidding or decision system is only as good as the signals feeding it — bad data produces bad optimizations, with high confidence.
Information that identifies a person: name, email, phone, address. Sending raw PII to GA4 violates Google's terms and GDPR — which is why data is hashed first and handled with controlled access.
The period for which you keep tracking data before it is automatically deleted. GA4 has its own settings (2 or 14 months at user level), and GDPR requires retention to be justified and limited to purpose.
The technique by which a report is computed on a subset of data, not all of it, for speed — at high volumes. It introduces imprecision. Exporting to BigQuery removes sampling by working on the raw, complete data.
Excluding automated traffic (bots, crawlers, referral spam) from data so reports reflect real people. Unfiltered bot traffic inflates sessions and dilutes conversion rates.
Tags added to URLs (utm_source, utm_medium, utm_campaign) to identify where traffic comes from. A disciplined UTM convention is the difference between clean channel reports and a mess of un-matchable sources.
The setup by which the same user is recognized when moving between your domains (e.g. from the site to a checkout on another domain), so they don't appear as two different sessions from different sources.
The error where your own domain appears as a traffic source (referral), usually from a misconfigured cross-domain setup or payment gateways. It distorts attribution if not excluded.
The process of merging the same customer's touchpoints (devices, sessions, channels) into a single coherent profile. With a first-party identifier (e.g. user ID on login), attribution and personalization become far more accurate.
The terms are not islands. They work as a system: first-party data feeds server-side collection, orchestrated consent decides what is sent, and recovered conversions improve the signal quality AI bidding relies on. Weakness in one pillar breaks performance in the others. That is what we do at WebSEM: we don't just install a tag and call it done; we build the measurement infrastructure in which your data produces correct decisions — measured on profit, not vanity metrics.
Definitions reflect the state of the measurement ecosystem in mid-2026: Google abandoning third-party cookie removal (2024–2025), Privacy Sandbox deprecation (October 2025) and Consent Mode v2 mandatory in the EEA/UK since March 2024. Terms evolve; we update periodically.
Want a measurement infrastructure that doesn't lose data?