This glossary defines 57 terms used in marketing intelligence, the research, measurement and AI vocabulary behind modern marketing strategy, from share of voice and consumer tension to MCP, GEO and MER. Each definition takes one to three sentences, is written for marketers rather than engineers, and most link to the SOMIN guide that covers the idea in depth.
How to use this page
Jump to a letter from the contents, or search the page for a term. Where SOMIN uses a word in a specific way, such as angle, moment, proposition, story or clickability, the definition says so. Every term has its own anchor, so you can link straight to it, for example share of voice. For the numbers behind terms such as engagement rate, see the social media benchmarks; for the brief that ties persona, tension, angle and story together, see the creative brief template.
A
A/B test
A controlled experiment that shows two versions of an ad, page or email to comparable, randomly split audiences and compares one outcome, such as click-through or conversion rate. It tells you which version won for that audience at that time, provided the sample is large enough to rule out chance. Read more: how pre-testing and A/B tests work together.
Ad library
A public, searchable archive of the ads a platform is running or has run, such as the Meta Ad Library, Google's Ads Transparency Center or the LinkedIn Ad Library. Marketers use ad libraries to see which creative, offers and messages competitors put money behind. Read more: how to research competitor ads in the Meta Ad Library.
AEO (answer engine optimisation)
Optimising content so that answer engines, such as AI assistants, voice assistants and search features that answer directly, use it in their responses. In practice it overlaps heavily with GEO: answer the question early, structure the page clearly and make the source easy to trust. Read more: how brands get cited by AI answer engines.
AI agent
Software that uses a language model to plan and carry out multi-step tasks, such as pulling data, analysing it and drafting a deliverable, by calling tools rather than only replying with text. In marketing, agents run parts of a workflow while a person approves the result. Read more: SOMIN's AI marketing agents; what to automate and what to keep human.
Angle
The reason a consumer tension exists, named by a theory from psychology, sociology or behavioural economics, such as overconfidence bias or self-determination theory. In SOMIN's persona, tension, angle and story framework, the angle is what turns a description of a problem into an argument. Read more: the persona, tension, angle and story framework.
Archetype
A recurring character pattern, such as the Sage, the Hero or the Explorer, used to describe a brand's personality and voice. The twelve-archetype model popularised by Margaret Mark and Carol Pearson draws on Carl Jung's work; SOMIN tags posts with archetypes so you can see how a category presents itself. Read more: SOMIN Strategy.
Attribution
The method of assigning credit for a conversion to the touchpoints that came before it, using rules such as last click or first click, or a data-driven model. Attribution shows the paths people took, but on its own it cannot prove which marketing caused the sale. Read more: the marketing mix modelling barrier.
Audience research
The study of who a target audience is, what they want and how they behave, through surveys, panels, interviews, behavioural data or analysis of the content they engage with. It answers who and why, where social listening answers what is being said. Read more: social listening vs audience research vs ad optimisation; how audience research firms market their data.
B
Benchmark
A reference value from comparable brands, channels or periods, such as the median engagements per post in a category, used to judge whether your own result is good. A benchmark is only useful when the comparison set matches your size, platform and audience. Read more: social media engagement benchmarks for martech brands.
Brand lift
The measured change in awareness, ad recall, consideration or purchase intent between people who saw a campaign and a comparable group who did not, usually collected through platform surveys. It is a way of measuring incrementality for upper-funnel goals.
Brief (creative brief)
A short document that tells a creative team who the work is for, what tension it addresses, the single thing it should say, the proof behind it and how success will be judged. A good brief traces each of those answers back to evidence. Read more: the SOMIN creative brief template; how to turn a social listening insight into a brief.
C
Clickability (creative clickability)
The predicted likelihood that a creative will earn more or fewer clicks than comparable ads, estimated from the creative itself before any media is bought. SOMIN expresses it as High, Normal or Low; it is a probability, not a forecast of click-through rate. Read more: what creative clickability is and how it is predicted.
Consumer tension
The conflict between what a person wants and what stops them getting it, described the way they would describe it, such as wanting to act fast but fearing a wrong choice. A tension has two forces pulling against each other; a pain point often has only one. Read more: what a consumer tension is and how to find one.
Content homogenisation
The drift of a category's content towards the same formats, phrases and visuals, often sped up when every brand uses the same AI tools and templates. It erodes distinctiveness and makes ads easier to ignore. Read more: why every brand sounds the same; why ads in one category look the same.
Content pillar
A recurring theme a brand commits to publishing on, such as product education, customer stories or company culture, used to balance a content calendar. Pillars work best when each one is tied to an audience need and a planned share of posts. Read more: SOMIN's social media manager.
Context engineering
Deciding what information a language model sees when it does a task: the instructions, examples, data, retrieved documents and tool results in its context window. In marketing, better context, such as real audience data and brand rules, changes the output more than a cleverer prompt does. Read more: context engineering for marketing.
Conversion rate
The share of visitors, clicks or recipients who complete a target action, such as a purchase or a sign-up. Always state the denominator, because conversions per click and conversions per session give different numbers.
CPA (cost per acquisition)
Spend divided by the number of conversions, such as purchases, sign-ups or leads. Agree what counts as a conversion before comparing CPA across campaigns.
CPC (cost per click)
Spend divided by the number of clicks on an ad. A low CPC is only good news if the clicks go on to convert.
CPM (cost per mille)
The cost of 1,000 ad impressions: spend divided by impressions, multiplied by 1,000. It measures the price of attention, not its quality. Read more: key social advertising metrics.
Creative fatigue
The decline in an ad's performance as the same people see it again and again, usually showing up as rising frequency, falling click-through rate and a higher cost per result. The usual fix is new creative, not only a new audience. Read more: SOMIN's paid media agent.
Creative pre-testing
Evaluating ads before launch, through surveys, panels or predictive models, to decide which to run, fix or cut. Predictive pre-testing scores new creative against patterns learned from ads with known results. Read more: how to predict ad performance before you spend.
CTR (click-through rate)
Clicks divided by impressions, usually shown as a percentage. It measures how often people who see an ad click on it, not whether they go on to buy.
E
Engagement
Interactions with a post: likes or reactions, comments and shares, and on some platforms saves or reposts. Platforms and tools count engagement differently, so check the definition before you compare numbers. Read more: how SOMIN counts engagement in its benchmarks.
Engagement rate
Engagements divided by an audience base: either followers (engagement rate by followers) or people reached or impressions (engagement rate by reach). The two can differ widely, so always say which one you mean. Read more: how to measure influence beyond follower counts.
Entity
A distinct, identifiable thing, such as a brand, person, product or place, that search engines and language models recognise and connect to facts. Consistent naming, an about page and Organization structured data help engines understand which entity you are. Read more: generative engine optimisation.
Expression (creative expression)
A finished creative execution of an insight or proposition, such as an ad, a post, a video script or a visual, made for a specific channel. In SOMIN, expressions are generated downstream of the persona, tension, angle and story. Read more: SOMIN Strategy.
F
Frequency
The average number of times each person reached saw an ad over a period: impressions divided by reach. Rising frequency with a falling response is a common sign of creative fatigue.
G
GEO (generative engine optimisation)
The practice of making content easy for AI answer engines to find, understand and cite. The term comes from a research paper by Pranjal Aggarwal and colleagues, presented at KDD 2024, which found that adding citations, quotations and statistics raised a source's visibility in generated answers. Read more: how brands get cited by AI answer engines.
H
Hallucination
Output from a language model that is fluent and confident but false or unsupported, such as an invented statistic, quote or source. Because models generate plausible text, every factual claim in AI-written marketing needs checking against a source. Read more: how to check AI copy before it ships; why marketers distrust AI insights.
Hook
The opening line, frame or first seconds of a piece of content, whose job is to earn the next few seconds of attention. In short video, the hook decides whether anyone sees the rest.
Hreflang
An annotation, placed in a page's HTML, in HTTP headers or in a sitemap, that tells search engines which language or regional version of a page to show to which users, for example English for Singapore versus English for Australia. The x-default value names the fallback page. Read more: content localisation in Asia.
I
Incrementality
The extra outcomes, such as sales or sign-ups, that happened because of a marketing activity and would not have happened without it. It is measured with controlled experiments, such as audience or geographic holdout tests, rather than read off an attribution report. Read more: the marketing mix modelling barrier.
Insight
A non-obvious truth about an audience that explains its behaviour and suggests what a brand could do about it. In SOMIN, an insight, also called a story, combines a persona, a tension and an angle into one line, with the posts behind it attached. Read more: how to make AI insights verifiable.
L
Large language model (LLM)
A neural network trained on very large amounts of text to predict the next word or word fragment, which lets it write, summarise, classify and reason over language. Claude, GPT and Gemini are examples. Read more: why the same prompt writes the same ad.
llms.txt
A proposed convention, published by Jeremy Howard in September 2024, for a Markdown file at a site's /llms.txt address that gives AI agents a short, curated guide to the site and links to clean versions of its key pages. It is aimed at agents answering questions rather than at model training. Read more: SOMIN's own llms.txt file.
Localisation
Adapting content to a market's language, culture, references, formats and timing, not only translating the words. Good localisation changes the message where local needs differ. Read more: content localisation in Asia is not translation; the APAC and MENA marketing calendar.
Lookalike audience
An ad audience that a platform builds from people who resemble a source list, such as existing customers or site visitors. It helps scale reach to new people who share the source audience's characteristics.
M
Marketing mix modelling (MMM)
A statistical method that estimates how much each channel and factor, such as price, seasonality and promotions, contributed to sales, using aggregated time-series data rather than tracking individual users. It also estimates diminishing returns, which makes it useful for setting budgets. Read more: the marketing mix modelling barrier did not disappear.
MCP (Model Context Protocol)
An open standard, introduced by Anthropic in November 2024, that lets AI applications connect to outside data sources and tools through one common interface. With an MCP connector, an assistant such as Claude can query live marketing data instead of relying on pasted exports. Read more: how to connect Claude to marketing data with MCP; MCP for marketing agencies.
MER (marketing efficiency ratio)
Total revenue divided by total marketing spend over the same period, across every channel. Sometimes called blended ROAS, it sidesteps attribution arguments but cannot tell you which channel drove the result.
Moment
The point at which an audience feels a tension and can act on it, such as a festival, a purchase decision, a search or a frustrating day. In SOMIN, moments pair a persona with a tension or relevance theme and come in types such as seasonal, life, purchase and micro-moments. Read more: how moments connect insight to brief; the APAC and MENA marketing calendar.
O
Organic reach
The number of people who see content without paid promotion. It depends on each platform's ranking systems, so it moves when those systems change. Read more: what still drives organic reach in 2026.
Owned, earned and paid media
Owned media is the channels a brand controls, such as its website and social accounts; earned media is coverage and sharing by others; paid media is advertising. Each does a different job, so most plans use all three.
P
Persona
A research-based profile of a target segment, with goals, needs, context and behaviour, used to keep work focused on a real reader. A persona is only as good as the data behind it. Read more: how to build personas from audience data and competitor ads.
Proposition
The specific promise a brand makes to a persona at a moment, joining an insight to one product truth, such as a feature, a reason to believe, a unique selling point or an offer. In SOMIN, each proposition comes with summaries, sample taglines and the takeaway the audience should leave with. Read more: where the proposition sits in a brief.
R
RAG (retrieval-augmented generation)
A technique in which a system first retrieves relevant documents or data and then passes them to a language model with the question, so the answer is grounded in that material rather than only in what the model learned in training. Read more: why category data beats a better prompt.
Reach
The number of unique people who saw a piece of content or an ad at least once in a period. Reach counts people; impressions count views.
ROAS (return on ad spend)
Revenue attributed to advertising divided by ad spend, often written as a ratio such as 4:1. It depends on the attribution model behind it, so compare ROAS only within one measurement method. Read more: SOMIN's paid media agent.
S
Schema markup (structured data)
Code, usually JSON-LD using the schema.org vocabulary, that labels what a page contains, such as an organisation, an article, a product or an FAQ, so that search engines and AI systems can read it reliably. Read more: generative engine optimisation.
Sentiment
Whether a mention or post expresses a positive, negative or neutral feeling, usually classified by a model. Sentiment works well as a trend line and poorly as a verdict on a single post, because sarcasm and context trip up classifiers. Read more: the essential guide to social listening tools.
Share of voice (SOV)
A brand's share of all the conversation, content or advertising in its category over a period, measured for example as a share of posts, mentions or impressions. Comparing share of voice with share of market shows whether a brand is out-investing its size. Read more: share of voice by market.
Social listening
Tracking and analysing public conversations and content about brands, competitors and topics across social platforms, news sites and forums, to spot themes, sentiment and shifts as they happen. Read more: social listening vs audience research vs ad optimisation; what social listening tools post about.
Story
In SOMIN, the one-line expression of an insight, built from a persona, a tension and an angle and backed by a short narrative that dramatises it. It should be a line the persona would recognise and only the brand could say. Read more: the persona, tension, angle and story framework.
T
Tone of voice
How a brand sounds, such as informative, playful, authoritative or hopeful, separate from what it says. SOMIN labels the tone of every post it analyses, which shows which tones a category over-uses and which ones earn a response. Read more: engagement benchmarks by tone of voice.
W
Whitespace
A message, audience, customer need or format that competitors in a category leave unclaimed and that buyers would respond to. A gap with no demand behind it is not whitespace; it is empty space. Read more: what marketing whitespace is and how to find it; finding whitespace by market.
Z
Zero-click search
A search that ends without a click to any website, because the answer appears on the results page or in an AI-generated response. It makes being cited inside the answer more valuable than ranking below it. Read more: the new rules of brand discovery.
FAQ
What is marketing intelligence?
Marketing intelligence is the collection and analysis of information about markets, competitors, audiences and campaign performance, used to make marketing decisions. It combines social listening, audience research, competitor ad analysis and performance data, and increasingly uses AI to turn that information into personas, insights and briefs.
What is the difference between social listening and audience research?
Social listening tracks what is being said in public conversations and content, as it happens. Audience research studies who the audience is and why they behave as they do, through surveys, panels and behavioural data. Most strategies need both: listening finds the signal, research explains it.
How are persona, tension, angle and story connected?
They are four questions answered in order. The persona is who you are talking to, the tension is what is bothering them, the angle is why it happens, and the story is the one line that says it. Each step narrows what the copy can say, which is why skipping one tends to produce generic work.
What is the difference between GEO, AEO and SEO?
SEO aims to rank pages in search results. AEO and GEO aim to have content used and cited inside the answers that AI assistants and answer engines produce. The basics overlap: clear answers, trustworthy sources, structured pages and original data.
What is the difference between ROAS and MER?
ROAS divides the revenue attributed to ads by ad spend, so it depends on an attribution model and is usually reported per channel or campaign. MER divides total revenue by total marketing spend, so it ignores attribution and shows overall efficiency, but cannot say which channel drove the result.
Why does MCP matter for marketers?
The Model Context Protocol lets an AI assistant connect to live data and tools through a standard connector, so it can analyse your actual campaign, social and audience data rather than a pasted spreadsheet. That makes AI analysis faster to run and easier to check against its source.