How streaming metrics are collected, calculated, and interpreted

Digital radio platforms provide access to a wide range of audience and listening data. Metrics such as Total Listener Hours (TLH), Average Time Spent Listening (ATSL), sessions, and Unique Listeners can all help broadcasters understand how audiences engage with their content across apps, websites, smart speakers, connected cars, and third-party platforms.


However, unlike traditional broadcast audience measurement systems such as RAJAR, streaming metrics are not collected through listener surveys. Instead, they are derived from technical activity generated when devices connect to streaming services.


This is an important distinction.


Streaming analytics are based on information recorded by streaming infrastructure during playback sessions. These systems can provide extremely useful insight into listening behaviour and digital platform performance, but they also rely on statistical interpretation and estimation. As a result, some metrics are naturally more reliable than others, particularly over longer reporting periods.


Understanding how these metrics are generated, what they represent, and where their limitations exist is essential when interpreting digital listening data.


This article explains:


How Streaming Metrics Are Generated

When a listener starts a radio stream, their device connects to a streaming server. As part of its normal operation, the streaming platform records technical information about that connection in server logs.


Each playback connection is commonly referred to as a session.


During a session, the streaming platform may record information such as:

  • When the session started and ended
  • How long playback continued
  • Which stream or station was requested
  • What type of player or device was used
  • Platform or app information
  • Network information such as IP addresses
  • Additional identifiers provided by apps or players


This information forms the foundation of digital audio analytics.


For example:

  • If a listener plays a station in a mobile app for 30 minutes, this generates a listening session
  • If they reconnect after losing signal, a second session may be created
  • If they later continue listening on a different device, this may appear as additional listening activity


Streaming platforms aggregate and process this technical activity to produce reporting metrics.


In addition to standard connection data, some modern apps and web players may also provide additional information through technologies sometimes referred to as URL decoration or listener identification. This can include anonymous listener IDs, consent signals, or platform identifiers that help reporting systems interpret listening activity more consistently.


The quality and consistency of this information can vary significantly between platforms, devices, and player implementations. This is one of the reasons streaming metrics may behave differently across apps, websites, aggregators, and smart speaker platforms.


Common Streaming Metrics Explained

Digital audio platforms provide a wide range of metrics, each designed to measure different aspects of listener behaviour. Some metrics are more stable and reliable than others, particularly over longer periods of time.


Understanding what each metric represents is important when interpreting station performance.


Total Listener Hours (TLH)

Total Listener Hours measures the total amount of listening time consumed across all listeners.

For example:

  • One listener consuming 1 hour of audio = 1 TLH
  • Ten listeners consuming 1 hour each = 10 TLH


TLH is generally considered one of the most reliable streaming metrics because it is based on total playback duration rather than audience identification.


As a result, TLH is typically:

  • Less affected by device changes
  • Less sensitive to listener identification methods
  • More stable across platform migrations or app upgrades
  • Well suited to long-term trend analysis


For many broadcasters, TLH is one of the strongest indicators of overall listening performance.


Sessions

A session represents an individual playback connection to a stream.


A new session is typically created whenever:

  • A listener starts playback
  • A player reconnects
  • A listener changes device
  • A network interruption causes playback to restart


Sessions are useful for understanding listening activity and connection behaviour, but they do not directly represent individual people.


For example, one listener may generate multiple sessions during a single day if they:

  • Move between networks
  • Experience connectivity interruptions
  • Switch devices
  • Restart playback manually


Average Time Spent Listening (ATSL)

Average Time Spent Listening measures the average duration of listening sessions.


This is typically derived by dividing:

  • Total listening time
    by
  • Total number of sessions


ATSL can provide useful insight into listener engagement and content retention.


For example:

  • Longer ATSL may suggest highly engaged audiences
  • Shorter ATSL may indicate more casual or transient listening


However, ATSL can also be influenced by session fragmentation. If listeners reconnect frequently due to connectivity issues, one continuous listening experience may appear as multiple shorter sessions.


Unique Listeners

Unique Listeners are an estimate of how many distinct listeners accessed a stream during a reporting period.


Unlike TLH, this metric relies heavily on listener identification methods such as:

  • Listener IDs
  • Device identifiers
  • IP addresses
  • Session analysis


This makes Unique Listeners one of the most complex and least precise streaming metrics.

For example:

  • One person listening across multiple devices may appear more than once
  • Multiple listeners sharing a device may appear as one
  • IP addresses may change over time
  • Different platforms provide different levels of listener identification


Because of this, Unique Listeners should generally be treated as an estimate rather than an exact audience count.


Shorter reporting periods, such as daily reporting, are usually more reliable than monthly reporting because listener identifiers are less likely to change within a smaller time window.


Why Streaming Metrics Are Not Perfect

Streaming metrics are extremely valuable for understanding digital listening behaviour, but no digital measurement system can perfectly identify real people and listening activity in every scenario.


This is not unique to any one platform or provider. It is a broader characteristic of streaming analytics across the digital audio industry.


Listener Identification Limitations

Streaming systems identify activity based on technical information generated by devices and applications. They do not directly know who is listening in the same way that a survey respondent can self-report listening behaviour.

For example:

  • A connected car may be used by multiple people
  • An office or shop stream may be heard by several listeners simultaneously
  • One person may listen across multiple devices during a single day


In all of these situations, the technical data may not perfectly reflect the number of individual listeners.


IP Addresses Are Not Permanent Identifiers

Historically, many reporting systems relied heavily on IP addresses to estimate unique audiences.


However, IP addresses are not stable identifiers:

  • Mobile networks may assign different addresses over time
  • Home broadband addresses may rotate periodically
  • Multiple users may share the same outward-facing address
  • Carrier-grade network technologies can cause large groups of users to appear under similar network identifiers


Over short periods this may have limited impact, but over longer reporting windows these behaviours can significantly affect audience estimation.


Connectivity Interruptions and Session Fragmentation

Modern streaming relies on internet connectivity, which is not always continuous.


Listeners may:

  • Lose mobile signal temporarily
  • Move between Wi-Fi and mobile networks
  • Experience short buffering interruptions
  • Restart playback unintentionally


In many cases, these events generate additional sessions even when the listener experiences what feels like uninterrupted listening.


This can affect metrics such as:

  • Sessions
  • ATSL
  • Unique Listeners


Many analytics systems attempt to reduce the impact of this behaviour through statistical processing and session analysis, but no method is perfect.


Automated and Non-Human Traffic

Streaming platforms may also receive requests from:

  • Search engine crawlers
  • Monitoring systems
  • Aggregators
  • Automated services


Modern reporting systems attempt to identify and filter non-human activity wherever possible. Improved platform identification and listener identification technologies have significantly improved this process over time.


However, filtering and classification methodologies can vary between platforms and providers.


Why Consistency Matters

Because streaming metrics rely on technical interpretation, consistency is often more important than absolute precision.


Changes in:

  • Apps
  • Web players
  • Listener identification methods
  • Platform integrations
  • Reporting methodologies

can all influence how metrics behave.


For this reason, metrics are generally most valuable when:

  • Comparing trends over time
  • Using consistent methodologies
  • Reviewing multiple metrics together
  • Understanding broader audience behaviour rather than focusing on a single number in isolation


Why Streaming Metrics Behave Differently to RAJAR

Streaming analytics and RAJAR figures are often discussed alongside one another, but they are produced using fundamentally different methodologies. Because of this, they should not be expected to align directly.


RAJAR Measures People and Behaviour

RAJAR is a survey-based audience measurement system. It collects listening information from selected participants who record or report their listening behaviour across radio platforms.


This methodology is designed to estimate:

  • Audience reach
  • Listening habits
  • Time spent listening
  • Broader population behaviour


Importantly, RAJAR measures people and reported listening behaviour, not technical playback sessions.


This means:

  • Multiple people listening together can still be represented
  • Broadcast listening can be captured even where no digital data exists
  • Audience estimates are derived from statistical population sampling


Streaming Metrics Measure Technical Activity

Streaming analytics work differently.


Digital audio metrics are derived from technical information generated when devices connect to streaming infrastructure. Reporting systems interpret this activity to estimate listening behaviour.


Streaming systems typically measure:

  • Playback sessions
  • Listening duration
  • Device activity
  • Platform behaviour
  • Listener identifiers where available


However, they do not directly know:

  • Who is physically present
  • How many people are listening together
  • Whether one person is listening across multiple devices


As a result, streaming metrics are behavioural and technical estimates rather than direct audience measurement.


Why the Numbers May Not Match

There are several reasons streaming metrics may differ from survey-based audience figures:


Different methodologies

RAJAR and streaming platforms are measuring different things in different ways.


Shared listening environments

One stream may represent multiple listeners in:

  • Cars
  • Offices
  • Shops
  • Shared smart speaker environments


Multi-device listening

One person may generate multiple digital sessions across:

  • Mobile apps
  • Websites
  • Smart speakers
  • Connected cars


Different platform visibility

Some third-party platforms provide limited listener identification information compared to first-party apps and websites.


Reporting windows

Streaming identifiers become less stable over longer reporting periods, particularly where systems rely heavily on IP addresses.


Why Both Types of Measurement Still Matter

Despite their differences, both methodologies provide valuable insight.


Broadly speaking:

  • RAJAR provides market-level audience and reach measurement
  • Streaming analytics provide operational and behavioural insight into digital listening activity


Streaming metrics are particularly useful for understanding:

  • Platform usage
  • Engagement trends
  • Listening duration
  • Device behaviour
  • Digital audience growth over time


The important thing is to understand the strengths and limitations of each methodology rather than expecting them to produce identical results.


One of the most important principles when interpreting streaming analytics is understanding the difference between:

  • exact audience counts
    and
  • trend analysis.


Streaming metrics are generally most valuable when used to identify patterns and movement over time rather than as precise headcounts of individual listeners.


Why Trend Analysis Matters

Digital audio reporting relies on technical signals and statistical interpretation. Because of this, some level of estimation is unavoidable, particularly for metrics such as Unique Listeners.


However, even where individual numbers are imperfect, consistent reporting methodologies can still provide extremely valuable insight into:

  • Audience growth
  • Listening behaviour changes
  • Platform adoption
  • Engagement trends
  • Campaign and programming performance


For many broadcasters, understanding whether listening is increasing, decreasing, or changing behaviourally is more useful than attempting to determine an exact number of individual listeners.


Consistency Is More Important Than Perfection

Metrics are most reliable when:

  • The same methodology is used consistently
  • Comparisons are made across equivalent periods
  • Platform changes are taken into account
  • Multiple metrics are reviewed together


Changes to apps, players, or listener identification methods can influence how reporting systems interpret audience activity.


For example:

  • Modern listener identification may reduce duplicate counting
  • Improved filtering may remove non-human traffic
  • Better session handling may affect ATSL calculations


This can make metrics appear to change suddenly even when underlying listening behaviour remains relatively stable.


No Single Metric Tells the Full Story

Every streaming metric has strengths and limitations.

For example:

  • TLH is generally stable and reliable for long-term trend analysis
  • ATSL can provide insight into engagement
  • Sessions help explain listening activity
  • Unique Listeners provide directional audience estimation


Reviewing these metrics together usually provides a much more accurate understanding of station performance than focusing on any single figure in isolation.


Key Takeaways

  • Streaming metrics are derived from technical listening activity rather than direct audience surveys

  • Different metrics measure different aspects of listener behaviour
  • Some metrics, particularly Unique Listeners, are estimates rather than exact audience counts
  • Shorter reporting periods are generally more reliable for unique audience estimation
  • App, player, and platform changes can influence how metrics behave
  • Total Listener Hours (TLH) is often one of the most stable long-term performance indicators
  • Streaming analytics are most valuable when used consistently for trend analysis over time
  • Streaming metrics and RAJAR figures use different methodologies and should not be expected to align directly