Audio guide planning
Museum visitor analytics: what audio guide data can tell you
A practical guide to museum visitor analytics from audio guides: uptake, playback progress, language use, selected-stop sequences, privacy and reporting.
Museum visitor analytics combines ticketing or footfall, observation, surveys and events from audio guides or apps. This article covers the audio guide layer: language selection, stop starts, playback progress, selected-stop order and device reporting. These events establish system activity. Attention, learning, satisfaction, physical dwell time and reasons for skipping require observation or direct visitor research. Every report needs a metric definition and denominator. Language share describes guide sessions. Playback progress describes started tracks. Used with gallery observation and feedback, these measures can reveal under-visited exhibits, confusing route moments and content worth revising, then show whether a change helped visitors discover and use the guide more effectively.
What audio guide analytics record
An audio guide produces event data. Depending on the system, an event may record a session start, language selection, stop selection, playback position, trigger detection, device identifier, content version or sync time. Those records describe system activity. Visitor motivation, attention and interpretive impact require observation, interviews or surveys.
Treat analytics as one evidence source. UK Government Service Manual guidance recommends combining digital analytics with user research and defining the purpose of a service before choosing metrics. Museum evaluation follows the same logic. In 2026, the American Alliance of Museums' guidance on intercept surveys described direct visitor feedback as a way to understand who is visiting and why. Direct feedback supplies motivation and audience context alongside the device events.
| Recorded signal | Reasonable use | Interpretation limit |
|---|---|---|
| Language selected | Compare language demand among valid guide sessions | The language profile of every museum visitor |
| A stop started | Measure the share of guide sessions reaching that stop | The visitor understood or enjoyed the object |
| Playback end position | Estimate how much of a started track played | That the visitor listened continuously or retained the content |
| Order of selected stops | Find common guide sequences and unexpected jumps | An exact physical route through the building |
| Device sync or upload | Check whether reporting data reached the platform | The time at which every recorded action happened unless playback timestamps are also stored |
Define every metric and denominator
The same dashboard label can hide different calculations. One supplier may call a device boot a session; another may require a language selection or first track start. One may call a track complete at 90%; another may require playback to reach the final second. Put the formula, exclusions, aggregation rule and reporting timezone in the data dictionary before comparing results.
| Metric | Working definition | Decision it can support | Main caution |
|---|---|---|---|
| Valid guide sessions | Sessions meeting an agreed activity rule after staff tests, resets and duplicates are excluded | Usage volume by day and hour | Requires a separate total-visitor or issued-device measure |
| Guide uptake | Issued guides or qualified digital sessions divided by eligible visitors in the same place and time window | Distribution, signage and staffing | Needs an independent visitor denominator from ticketing, a door count or observation |
| Stop reach | Sessions playing a named stop divided by sessions that started the relevant tour | Route and content investigation | Later stops naturally have fewer possible listeners; closures and trigger faults can lower the result |
| Playback progress | Recorded end position divided by track duration for a started play; report active playback duration separately where the channel records it | Find tracks that may be too long, unclear or poorly placed | Represents recorded playback position; seek, replay and repeat-play rules must be stated |
| Completion rate | Started plays whose normalised end position reaches a stated threshold divided by all started plays | Compare versions of the same stop | Set the project threshold and normalise repeated plays before aggregation |
| Language share | Valid sessions in one language divided by all valid guide sessions | Plan guide-language capacity and review content demand | Scope: guide users within the reported period |
| Fleet utilisation | Devices issued or active divided by devices available in the same operating window | Fleet size, charging and spare planning | Unavailable, charging and unsynced units must be classified consistently |
Definitions matter even for apparently simple counts. The Smithsonian public-engagement dashboard states that one person visiting three Smithsonian museums on one day counts as three museum visits. Under that definition, the value describes museum visits. Counting unique people requires a separate measure. Audio guide reports need the same clarity.
Why each delivery channel produces different data
Dedicated devices, tablets, native apps and web tours observe different events and report on different schedules. A cross-channel dashboard requires shared stop IDs and metric definitions.
| Channel | Typical data path | Useful strengths | Common gap |
|---|---|---|---|
| Dedicated audio guide | Events stored on the device and uploaded when the guide reaches a connected dock or charger | Consistent hardware, issued-device counts and reliable offline use | Data may arrive after the visit; issuance still needs a front-desk or ticketing denominator |
| Museum tablet | Tour application records local events and uploads over Wi-Fi or an administration sync | Rich interaction events and shared stop identifiers with multimedia | Offline or unsynced tablets create a temporary reporting gap |
| Native app | App events sent when the visitor's phone has connectivity and tracking permissions allow | Pre-visit and post-visit use can be measured | Downloads, opens, tours started and physical visits are different denominators |
| QR-code web tour | Browser events sent during the session, subject to connectivity, privacy settings and tracker blocking | Browser access removes the app-install step from the scan-to-start funnel | Track scans, page views and audio starts as separate events |
| Live tour guide receiver | Channel assignment, handout or receiver connection may be recorded | Group size and simultaneous channel use | Usually little or no stop-level playback data because the guide controls the tour |
In Look2Innovate fleets, a network-connected Smart Charger can upload usage statistics to Live Visitor Statistics in Look2Guide on configurable sync cycles as often as every ten minutes. Treat upload time and playback time as separate fields. Preserve playback timestamps when the channel records them. Label server-side reporting timestamps as upload time. At Berlin Cathedral, Trend units collect visitor statistics, including selected-stop sequences, alongside multilingual tours. Selected-stop data describes guide use. Indoor positioning requires separate validated location technology.
Decisions the data can support
Look2Guide brings period review, exhibit editing and fleet updates into one operating cycle: review trends and signals for languages, stops and playback; edit exhibit audio, translations or playback settings; publish approved updates through connected Smart Chargers; then compare the next period. This supports small, testable improvements to the visitor experience.
Revise content with a testable hypothesis
A track with many starts and consistently low playback progress is worth investigating. Listen in the gallery, check the object sightline and ambient noise, then interview or observe visitors. If the script seems too long, shorten one version while keeping the stop, language and distribution conditions stable. Compare the same metric over a comparable period. Follow-up checks should separate script length from trigger failure, a closed gallery, sightline problems or a school group passing through.
Plan languages across comparable periods
Language share can show sustained demand among guide users and reveal whether a translated tour is being selected. Review it by season, day type and distribution point before changing the language budget. A low share has several possible causes: low demand, difficult language discovery, inconsistent front-desk offering or incomplete fleet loading.
Investigate under-visited exhibits and route problems
Stop reach can identify exhibits selected by a small share of guide sessions. Compare stops with similar route positions and availability, then observe the gallery. Check the previous cue, object visibility, signage, trigger performance and language labels. After a change, compare the same visitor window. This process can help more visitors discover overlooked objects and expose confusing moments in the route. Reserve the term heat map for systems with validated position data. Audio guide records provide selected-stop sequences.
Operate the fleet around real demand
Validated session or playback timestamps with the correct timezone can help identify demand periods. Channels that record upload time use issuance, ticketing or front-desk records for hour-of-day demand. Device status and sync completeness show whether part of the fleet is stale. Combine these sources with arrivals, average tour duration and returns when applying the museum audio guide fleet-sizing method.
Protect data quality before reading the trend
Most reporting errors begin before analysis. A content rename, staff test or missed upload can create a false trend that looks like visitor behaviour. The UK Government Data Quality Framework uses six dimensions: completeness, uniqueness, consistency, timeliness, validity and accuracy. They translate directly into questions such as whether every guide synced, retries were deduplicated and stop IDs remained consistent. Keep a short quality log alongside the dashboard.
- Use stable stop IDs across devices, apps, languages and content revisions.
- Record the active tour and content version with each event or reporting period.
- Exclude staff tests, installation checks, repeated resets and training sessions through a documented rule.
- Show stale or unsynced devices, last successful sync, expected-batch gaps and data-completeness status. Treat never-uploaded sessions as unobserved.
- Separate playback time, device clock time, upload time and report timezone.
- Keep the eligible-visitor denominator in the same venue, date and operating window as the guide numerator.
- Compare like periods, accounting for closures, school holidays, events, seasonal languages and changes at the handout desk.
- Aggregate or suppress small language and selected-stop segments where precision could identify individuals.
Check whether event definitions, the visitor population or data completeness changed before attributing a trend to visitor experience.
Privacy and data governance
Start with the least detailed data that answers the museum's question. Aggregate device statistics can avoid stable visitor identifiers. App profiles linked to tickets, email addresses or location history carry additional identification risk. Pseudonymous identifiers remain personal data when they allow a person to be singled out or linked across visits.
| Data level | Example | Governance question |
|---|---|---|
| Aggregate | Daily sessions and language totals with no visitor-level export | Are small groups suppressed and is the retention period necessary? |
| Pseudonymous session | A random session ID with stop sequence and playback events | Can the identifier be linked across visits, devices or other datasets? |
| Identifiable or linked | Guide use connected to a ticket, email, account or personalised report | What is the legal basis, notice, access rule, retention period and deletion process? |
For organisations subject to the GDPR, Regulation (EU) 2016/679 requires principles including purpose limitation, data minimisation, accuracy, storage limitation and security. It also sets requirements for processor contracts and, where processing is likely to create high risk, data-protection impact assessments. Document which party is controller, processor or subprocessor and the evidence supporting any GDPR compliance claim.
For web or app audience measurement in France, CNIL guidance updated in July 2025 says a consent exemption is limited to conditions including strictly necessary audience measurement, anonymous statistics, publisher-specific purpose, exclusion of cross-site tracking and exclusion of third-party reuse. It also recommends informing users and limiting tracker and data retention. The framework applies in France. Other deployments follow their applicable law and institutional policy.
What procurement teams should require
Analytics requirements should name the decisions, evidence, definitions and acceptance criteria. Data reporting already appears in museum audio guide procurement. Imperial War Museums included data download and provision in its audio and multimedia guide procurement, alongside content updates, maintenance and a five-year service period.
- Provide a metric dictionary with formulas, denominators, completion thresholds, exclusions and timezone.
- List every field actually collected; include stable stop, language, tour and content-version identifiers, and add a device identifier when operationally justified.
- Explain offline storage, upload timing, retry behaviour and how missing or unsynced data appears.
- Allow filters by site, tour, date, language, stop, content version and device type, with suppression for very small groups.
- Provide aggregate reports plus a documented CSV or API export suitable for the museum's own analysis.
- State data ownership, hosting location, subprocessors, international transfers, security controls and breach process.
- Define user roles, administrator permissions, access logs and account removal.
- Set retention and deletion rules for raw events, session-level data, exports and backups.
- Guarantee open-format export and documented deletion at contract end.
- Use an acceptance dataset with known sessions, missing uploads and test events to prove that calculations match the specification.
A practical first 90 days
In the first 30 days, verify the event dictionary, remove staff tests and establish a baseline with no content changes. By day 60, choose one question and make one controlled change, such as shortening a track with low playback progress or improving signage before a low-reach stop. By day 90, compare the same metric and visitor window, check the result through observation or feedback, and record whether the change is retained. This creates a repeatable decision log and supports regular review throughout the year.
FAQ
What is the most useful museum audio guide metric?
The decision determines the metric: uptake for distribution, stop reach for route investigation, playback progress for track review, language share for translation planning and fleet utilisation for operations. Define the denominator and exclusions before setting a target.
Can audio guide analytics measure total museum visitors?
Audio guide analytics measure valid guide sessions or issued devices according to the system's definition. Measuring guide uptake needs an independent count of eligible visitors from ticketing, a door counter or a matched observation window.
Does a high audio guide listen-through rate mean visitors liked the track?
Check the supplier definition first. In Trend data, playback progress is the recorded end position. Active-playback duration, where available, gives a stronger time measure. Satisfaction, understanding and attention require surveys, interviews or observation.
Can selected audio guide stops create a visitor-flow heat map?
The recorded order of selected stops can suggest route patterns. Exact physical paths, room time and positions between stops require validated location data. Label the audio guide measure selected-stop sequence.
Do museum analytics always require visitor consent?
Consent requirements depend on the technology, data, purpose and applicable law. Aggregate device statistics can avoid stable visitor identifiers. Cookies, app identifiers, location history and ticket-linked profiles introduce additional requirements. Document the legal basis, transparency, retention and processor roles with the museum's privacy lead.
