Retailtech ranking methodology
Indicative comparison across different industries: these companies do not answer the same questions and are not all direct competitors. Not a representative market study.
Measured 21/10/2026 · GPT-6 Luna (OpenAI) · 200 answers.
Sample and procedure
The study covers 21 explicitly selected companies from the Retailtech ecosystem. The selection is predefined and not representative. The ranking is independent of the event organiser and the existing Switzerland GEO ranking.
20 categories, each with ten prewritten German-language questions. On and Salomon share one category. Each question receives 1 answer from GPT-6 Luna (OpenAI) via the API, without browsing or individual conversation history. The system prompt requests a short, factual answer with up to five suitable providers.
The English pages translate this existing German-language measurement. The questions are translated for readers only; no new answers were collected. Scores, rankings and the measurement date remain unchanged. Use the German pages to see the original prompts.
Measurement questions contain no target brand names. All brand mentions are then extracted and classified for category relevance. Answers, classifications and configuration are versioned for traceability. The ten questions and mention results are disclosed on every company page.
Proper names and documented variants are merged. Subsidiary and portfolio brands are not automatically attributed to the target companies. Product or subsidiary brand names without the target company’s name count neither as target brand mentions nor as additional providers. A product or regional name that explicitly contains the target brand name does count. Former company names are included only through the disclosed aliases.
How the metrics are calculated
- Mention rate: the proportion of the ten answers that mention the company at least once.
- Average position: the average order of first mention among detected providers, only in answers that mention the company. No mention is shown as a dash rather than rank 0.
- Share of Voice: answers mentioning the brand divided by the sum of answer-level mentions of all providers in that category. Providers outside the 21 companies are also included in the denominator.
- Visibility Score: the mean AIRank score across all ten answers, including 0 for answers without a mention. It combines frequency, position, sentiment and share of text. Details of the shared score calculation (German).
- Positive sentiment: the proportion of positive extracted mentions among all detected mentions of the brand. This is not a measure of customer satisfaction.
- Sorting: descending Visibility Score, then mention rate. Identical values share a rank; alphabetical order is used only for display.
Limitations
The cross-industry comparison is indicative only. Questions, market structure, awareness and competition differ. A CRM provider and an underwear brand cannot be compared as direct competitors. The chosen category also does not cover the entire portfolio.
One answer per question and one model do not support robust conclusions about all AI systems or the stability of these values. Model knowledge, randomness and wording affect results. Ten questions yield mention rates in steps of ten percentage points.
AI-assisted extraction, classification and sentiment detection may contain errors. Answers are not verified factual claims or purchasing recommendations. The JTI category covers factual company analysis only, not recommendations to consume tobacco.
This snapshot is not updated automatically and says nothing about market share, company quality or the success of GEO measures.