Transparency
How Trackhound handles press content, external data, and what it will, and will not, do.
How press content is handled
Trackhound reads press headlines as one signal among many, to gauge how much coverage is building around a show. It works from headlines only, never full article text, and only from sources whose licence allows it. Press stays in the background as context, and is never surfaced directly.
Data sources
Met Office
Weather forecasts and observations per venue, retrieved via the Met Office DataHub API. Forecasts are retrieved each morning; actual observations each evening.
Data supplied by the Met Office.
National Rail / TfL
Transport disruption signals retrieved from TfL StopPoint API (London) and the Rail Data Marketplace. Raw observations (e.g. a disruption category at a nearby station tonight) are stored per project, per day, and purgeable. Trackhound does not produce its own train-time predictions; no forecasted arrival or departure time is displayed unless it is a verbatim pass-through of the live National Rail value at the moment of rendering.
Station coordinates from NationalRail (Train Information Services Ltd), provided via the Rail Data Marketplace.
Cohere
Cohere Embed is used to recognise report formats and find similar shows. The results are stored privately, per organisation, as part of your own knowledge base. No user data crosses organisational boundaries.
Ticketing systems
Where Trackhound connects to your ticketing system, it does so with your permission and read-only: it never sells, refunds, or emails on your behalf. Any personal buyer details those systems return are hashed in memory and never stored in plain text.
AWS Bedrock (EU region)
The AI models used for parsing, analysis, attribution and answering your questions run exclusively on AWS Bedrock in EU regions, so your data stays in Europe. They include models from Anthropic (Claude), Meta (Llama), Mistral and NVIDIA. Our most powerful option, used for deep research, always asks for your approval and shows the cost before it runs.
The benchmarking pool
Trackhound's forecasting models are stronger when similar shows can be found. Cooper organisations can opt in to the anonymised benchmarking pool, which allows sales curve shapes from closed shows to contribute to cross-org similarity search.
Opt-in is free and entirely voluntary. Contribution means that the normalised trajectory of a closed show (percentage of run complete versus percentage of capacity sold, at 50 data points) can be used to identify shows with similar sales dynamics for other organisations. No show name, organisation name, venue name, ticket price, or gross revenue crosses organisational boundaries; only the anonymised curve shape is shared.
Cross-org comparison only happens above a minimum segment size (at least six shows per segment), and any feature that surfaces cross-org results goes through legal review before it is released.
What we don't do
- No creative-asset generation. Trackhound will never write ad copy, generate images, draft social posts, or produce any marketing text. Intelligence, attribution, and detection are its purpose; creative decisions belong to you.
- No individual buyer profiling. Trackhound works with aggregate sales data (snapshots of totals, not individual transactions). Customer email addresses and phone numbers are hashed in memory and never persisted in plain text.
Forecast disclaimer
Forecasts (weather, sales, attribution) are predictions, not statements of fact.
Weather forecasts are derived from Met Office DataHub data. Sales and attribution forecasts are probabilistic model outputs; confidence intervals are displayed where available. Past performance of similar shows does not guarantee future results for your production.