Glossary

Market

The competitive landscape a campaign measures. A market defines the category, geography, and set of AI models used. One market can contain multiple campaigns. For example, you might create one market for South African banking and run separate campaigns to track different questions or time periods within it.

Campaign

The configured measurement unit inside a market. A campaign specifies which questions to ask, which entities (brands, attributes) to track, which audience segments to apply, and which AI models to query. Run a wave to collect responses.

Wave

One full execution of a campaign: all questions, models, and segments run together in a single batch. Each wave produces a set of AI responses. Run a second wave to compare AI Perception over time. The Trends page shows how entity share of voice changes between waves.

Baseline

A campaign run with no audience segment applied. The baseline represents the AI's unframed view of the topic. It is used as the comparison reference point when measuring segment lift. Segment lift is calculated as segment SOV minus baseline SOV for a given entity.

Segment

An audience group defined by demographic or attitudinal characteristics (for example: "Urban millennials" or "Price-sensitive shoppers"). For each segment, the AI generates a synthetic persona that represents a member of that group. Responses are collected once per segment per wave, and metrics are aggregated across all personas in the segment.

Share of voice (SOV)

The percentage of all tracked entity mentions in the campaign that belong to a given entity. It reflects how much of the conversation (by mention volume) that entity holds compared with other entities in the same class.

All metrics on this page describe model outputs (LLM responses), not direct consumer survey data or human recall.

Top of mind (TOM)

The percentage of model responses where a given entity was mentioned first (among entities of that class). This reflects LLM recommendation order and salience, not consumer spontaneous recall. Higher TOM means the model positions the entity as a primary answer.

Penetration

The percentage of responses that mention a given entity at least once. High penetration with lower SOV can mean the entity appears in many answers but with relatively few mentions per answer (broad but shallow).

Prominence

A position-weighted share of voice: entities mentioned earlier in a model response receive more weight. This rewards visibility near the start of an answer, not only raw mention count.

Mean position

The average normalised position of an entity in the mention order among its class in each model response (0 = first, 1 = last). The average is taken only over responses where that entity appears. Lower values mean the entity tends to appear earlier when it is named.

Saliency (composite score)

A single headline score on entity tables, computed as a weighted blend: Top of mind (TOM) 40%, share of voice (SOV) 25%, penetration 20%, and prominence 15%. It summarises how strongly an entity shows across several model-output signals. This is an LLMEKNOW composite, not a standard market research term. Always read it alongside the component metrics.

Sentiment

An average score derived from the sentiment of mentions (typically 0 = negative, 1 = positive), summarised per entity. Use it alongside SOV: high share with poor sentiment can signal risk.

Segment lift

The difference in SOV between a segment condition and a baseline. Positive lift means the entity gains share when a targeted audience frame is applied; near-zero lift means the entity is known similarly across audiences.

Segments and personas

A segment defines an audience group by demographic or attitudinal characteristics (e.g. Urban millennials, Price-sensitive shoppers). A persona is the synthetic AI-generated individual that represents a segment during a campaign run. One segment may produce multiple personas; metrics aggregate over all personas in a segment.

SOV trend (longitudinal)

On campaigns with at least two completed research waves, the direction of change in share of voice from the earliest to the latest wave for each entity. It is a descriptive pointer to momentum, not a statistical forecast. See the Trends page for full time series.