The media, marketing & AI glossary – explained the way we’d explain it in a meeting.

A CRM tracks known relationships, contacts, deals and interaction history, usually for sales and service teams. A CDP unifies data from many more sources, including anonymous and behavioural data, into a single customer profile for marketing activation. Many businesses run both, feeding the CRM’s known-customer data into the CDP alongside everything else.

Zero-party data is deliberately shared by the customer (a stated preference). First-party data is what you collect yourself from your own properties (site behaviour, purchase history). Second-party data is another company’s first-party data, shared directly with you by agreement. Third-party data is aggregated and sold by data brokers with no direct relationship to the source. Trust and accuracy generally decrease in that order.

Data Governance covers the rules for collecting, storing and sharing data generally. AI Governance is a more specific layer on top, evaluating how that data (and the models built on it) get used in AI systems specifically, covering bias, consent for automated decisions, and brand and compliance risk before a use case reaches production.

RTB is one method within programmatic, an auction that happens in the milliseconds before a page loads. Programmatic is the broader category, covering RTB as well as programmatic guaranteed and private marketplace deals that skip the open auction entirely.

An ad network aggregates inventory from multiple publishers and resells it as a bundle, often with the network taking on commercial risk. An SSP is the underlying technology that lets a publisher offer its own inventory into programmatic auctions directly. Many ad networks use an SSP under the hood; the two aren’t mutually exclusive.

The underlying concepts are global, CPM is CPM everywhere. Where regulation or market structure matters, privacy, retail media, data governance, entries are written from an Australian context specifically, since that’s where we operate. Treat the mechanics as universal and the regulatory detail as AU-specific unless noted otherwise.

MTA works at the individual level, tracking specific touchpoints for a specific customer, which makes it precise but increasingly limited as cookie-based tracking declines. MMM works statistically at the aggregate level, measuring channel impact without needing individual tracking at all, more resilient to a cookieless environment but less granular.

Retail media is the broader category, any advertising that runs on or is informed by a retailer’s own audience and data. A Retail Media Network (RMN) is the specific platform a retailer builds to sell that advertising, the infrastructure retail media actually runs on.

Segmentation groups customers into shared buckets based on common characteristics. Personalisation goes further, tailoring content or offers to the individual, sometimes using segments as an input, sometimes going more granular still. Segmentation is usually a step on the way to personalisation, not the same thing.

A DSP is what advertisers and agencies use to buy inventory. An SSP is what publishers use to sell it. Same programmatic transaction, opposite sides of it.

No, an LLM (Large Language Model) is a specific type of AI architecture trained on text. Generative AI is the broader category, covering LLMs as well as image, video and audio generation models that work differently under the hood. Every LLM is generative AI; not all generative AI is an LLM.

Deterministic identifiers come from confirmed data, a login or email match, so they’re accurate but require the customer to have actively identified themselves. Probabilistic identifiers are inferred from signals like device or browser fingerprints, covering more of your audience but with a margin of error. As third-party cookies fade, the industry is leaning harder on deterministic approaches.

Yes, particularly in fast-moving areas like AI, data privacy and retail media, where the underlying mechanics can shift within a year. If something here looks out of date, let us know and we’ll review it.

Yes, that’s a genuinely common use for it, giving your team and prospective vendors a shared, plain-English definition of the terms in scope before you get into vendor-specific claims. Link directly to the relevant entry rather than copying the text, so it stays current if we update it.

A MarTech stack is the full set of tools an organisation runs, CMS, email platform, analytics, personalisation engine and more, including how well they talk to each other. A CDP is one specific piece of that stack, usually the layer that unifies customer data so the other tools can act on it consistently.

Churn rate is the input, the share of customers you lose over a period. LTV is the output, the total profit a customer represents once that churn rate is combined with margin and revenue into one number. Run your own churn rate through the LTV & Churn Impact calculator to see it turned into a dollar figure.

Post-Merger Integration is the specific, time-boxed programme of combining two organisations after a deal closes, systems, teams, brand and process. Operating Model is the ongoing structure, centralised, decentralised, hub-and-spoke or federated, that the combined business runs under once integration is done.

Data Governance is the DMBOK2 hub discipline, the rules, roles and decision rights for who can do what with data. Data Quality is one of the outcomes governance is meant to protect, how accurate, complete and consistent the data actually is. Poor data quality is usually a symptom of weak governance further upstream.

A genuine requirement traces back to a specific, documented regulatory clause. Process debt is delay that’s accumulated around that requirement over time, extra sign-offs, informal escalation, or “we’ve always done it this way” caution, without a clause actually mandating it. The Compliance Friction Diagnostic is built to help separate the two.

MDM creates one trusted version of core entities, customers, products, locations, across every system in the business, including ones with nothing to do with marketing. A CDP is narrower, focused specifically on unifying customer data for marketing activation. A strong MDM foundation makes a CDP implementation faster and more reliable, not the other way around.