Plain English, no buzzwords
Every term you'll hear in a 2026 board meeting — defined for marketers, not engineers. Skim it before your next vendor call. Bookmark it.
An AI system that doesn't just answer prompts but takes multi-step actions — planning, executing, evaluating, refining. The defining shift between a chatbot (you ask, it answers) and an agent (you brief it, it works). VIMDRIVE's four agents are agentic in this strict sense.
The rule for crediting which marketing touchpoint deserves credit for a conversion. Last-click is the simple default; data-driven and blended models are more accurate but harder to explain. VIMD01 builds blended models that survive privacy changes.
A group of users defined by shared behaviour or attribute (e.g. "people who viewed a luxury collection in the last 14 days"). Cohorts are the unit of personalisation — most modern targeting is cohort-based.
Customer acquisition cost calculated using all marketing spend (paid + content + agency fees) divided by all new customers — paid and organic. The most honest version of CAC. Most agencies report only paid CAC because it looks better.
An AI system constrained to produce only outputs that fit a brand's voice, regulatory rules, and tonal guardrails. Done well, it sounds like the brand at scale; done poorly, it produces generic AI slop.
Revenue minus variable costs (including paid media). The metric that actually tells you whether a campaign makes the business money. Stop measuring ROAS; start measuring contribution margin.
A server-side feed sending conversion events directly to ad platforms (Meta, Google), bypassing the browser. Critical for accurate attribution after iOS privacy changes. VIMD01 sets these up as part of any engagement.
The systematic process of running variants of an ad against each other to find what converts best. Traditional agencies test 4–8 variants per campaign; AI-native agencies test 100s — that's the structural advantage.
Google's machine-learning attribution model — credits each touchpoint based on its actual statistical contribution to conversion. Better than last-click, still inside Google's walled garden.
Marketing designed to drive an immediate measurable action (purchase, lead, app install). Distinct from brand marketing, which builds preference over time. Most "performance marketing" is direct response.
A numerical representation of meaning. AI systems convert text, images, and audio into embedding vectors that can be compared mathematically. Underpins how AI search, RAG, and similarity-based audience modelling work.
The system that grades AI output before it goes live. For marketing creative: brand voice, factual accuracy, regulatory compliance, performance prediction. Without one, you have no idea what your AI just shipped.
A large general-purpose AI model (GPT, Claude, Gemini) trained on broad data, capable of being adapted to many tasks. The substrate underneath specialised agents like VIMC01 and VIMS01.
The maximum number of times a single user sees the same ad in a defined window. Too low and you under-influence; too high and you burn audience favour.
The practice of getting your brand cited inside AI-generated answers from ChatGPT, Perplexity, Gemini, and Google's AI Overviews. The successor to SEO for the AI-search era — though SEO and GEO overlap heavily.
A constraint applied to AI output to prevent harmful, off-brand, or off-topic content. The seatbelts of AI marketing. We have many.
An AI output that sounds confident but is factually wrong. Mitigated by retrieval-augmented generation, evaluation harnesses, and human review. Cannot be eliminated entirely — only reduced to acceptable rates.
Workflow where AI produces output and a human reviews it before action is taken. Essential for high-stakes outputs (large-budget creative, regulated industries). VIMDRIVE runs HITL on every campaign above a defined spend.
The lift caused by your marketing — i.e. customers you would not have got without it. Distinct from gross conversions, many of which would have happened anyway. The honest test of whether marketing actually works.
What a user actually wants when they search a term. Categories: informational, navigational, commercial, transactional. VIMS01 maps every keyword to its intent before content or paid is built.
Lifetime value divided by customer acquisition cost. The North Star ratio of digital marketing economics. A healthy LTV/CAC sits above 3.0 for most categories; below 1.0 means your marketing is destroying value.
The class of model that powers most generative AI in marketing today — GPT-4/5, Claude 4, Gemini, etc. Designed to predict the next token of text given a prompt.
A platform-built audience modelled on similarity to a seed list (e.g. your existing customers). Useful for prospecting; less reliable since iOS privacy changes shrunk the underlying signal.
A statistical model that estimates the contribution of each marketing channel to overall revenue. Top-down counterpart to bottom-up attribution. Larger budgets benefit most from running both.
Marketing-Qualified Lead and Sales-Qualified Lead — the standard handoff stages between marketing and sales. The definitions are usually wrong, which is why most B2B funnels leak.
Creative produced specifically for the platform it runs on — TikTok creative looks like TikTok, Instagram looks like Instagram. Cross-posting kills performance. VIMC01 produces platform-native variants by default.
The action you tell an ad platform to optimise toward (purchase, lead, view content). The platform's algorithm is only as smart as the event you give it. Picking the wrong event is one of the most common causes of underperformance.
Showing different content to different users based on what's most likely to convert them. Modern personalisation is cohort-level (segments of users) not individual-level (creepy).
A small piece of code on your site that fires when users perform an action, sending data back to ad platforms. The classic, browser-side conversion mechanism — increasingly augmented by server-side conversion APIs.
The text instruction given to an AI model to produce an output. Inside VIMDRIVE, agents work from briefs (multi-paragraph structured documents), not single prompts.
An AI architecture that retrieves relevant facts from a knowledge base before generating output — reducing hallucination and grounding the response in real data. Used inside VIMS01 to make sure strategy is informed by your data, not invented.
Revenue from an ad campaign divided by spend on that campaign. Useful for paid-channel comparison; useless as a profit measure (it ignores cost of goods, fulfilment, and CAC).
An AI training approach where the system learns by trying actions and being rewarded for good outcomes. Underlies most of the bid-optimisation logic in modern paid platforms — and inside VIMP01.
The practice of getting pages to rank in search results. Still alive in 2026 — particularly for commercial-intent queries — but compounded by GEO for informational queries.
Machine-readable tags on a webpage that help search engines and AI systems understand what the page is about. Implementing schema is half of GEO.
The unit of text an LLM processes — roughly a syllable. Pricing of AI APIs is usually per-token. Long prompts are expensive; concise briefs are economical.
Marketing aimed at people who don't yet know your brand or category. Awareness, content, broad reach. Typically lower-converting per impression but compounds over time.
A database optimised for storing and querying embeddings (see above). The infrastructure underneath RAG, semantic search, and similarity-based audience modelling.
A conversion that happens after someone saw an ad but didn't click it. Real (people are influenced without clicking) but easy to overcredit. Best measured with incrementality testing.
