Why Nobody Is Listening to Your Live Stream (Our Data)

Most likely because they arrived, waited a few seconds, heard nothing happening, and left — before you ever knew they were there.

We can put a number on it. Across 30 days we logged 1,842 people who opened a live audio platform, waited, and gave up without connecting to anything. The median gave up after 13 seconds. Nine in ten were gone inside a minute.

This is our own data, from our own logs, and we're publishing it because we have not seen anyone else measure the specific moment where a live audience is actually lost.

How Long Do People Actually Wait?

Thirteen seconds. That is the median, not the floor.

The mean is badly misleading here — a handful of people who left a tab open for hours drag it to 40 seconds. The median is the number that describes real behaviour, and it is brutally short.

Note the shape of the distribution: a quarter of people are gone in six seconds. That is less time than it takes to read a station description.

How Does This Compare to Known Streaming Research?

It measures a different failure, and that distinction matters.

The established research is about startup delay — how long the player takes to produce sound after you press play. The figure everyone quotes comes from a peer-reviewed study rather than a vendor blog, and it is worth reading at source: Krishnan and Sitaraman's Video Stream Quality Impacts Viewer Behavior (ACM Internet Measurement Conference, 2012; full text, UMass), built on 6.7 million viewers watching 23 million videos over 216 million minutes.

Their finding:

"Viewers start to abandon a video if it takes more than 2 seconds to start up, with each incremental delay of 1 second resulting in a 5.8% increase in the abandonment rate."

Two further results from the same paper are rarely quoted and matter more to a broadcaster than the headline. Rebuffering equal to just 1% of a video's duration cut watch time by 5%. And a viewer who hit a failure was 2.32% less likely to return to the site within a week — the cost of a bad session outlives the session.

The industry rule of thumb is well established:

Startup time should be under two seconds — every fraction of a second beyond that directly impacts viewer retention.

Our 13 seconds is not that. Our player starts instantly; there is nothing wrong with the technology. What our listeners are waiting for is something to exist — a broadcaster to go live, or another person to arrive.

That makes the two numbers complementary rather than contradictory. Two seconds is the tolerance for technical delay. Thirteen seconds is the tolerance for emptiness. If a stream that loads slowly loses half its audience, a stream that loads instantly into silence gets roughly eleven extra seconds of patience — and no more.

What Does 13 Seconds Mean for a Broadcaster?

It means your first impression is not your first minute. It is your first ten seconds, and silence is the most expensive thing you can put there.

Three practical consequences:

1. Dead air at the top of a show costs you the audience that arrived for it. The people checking whether you are live do it in the first seconds. If they hear nothing, they cannot distinguish "not started yet" from "broken."
2. Announced start times matter more than the content that follows. Someone arriving at 8:00 for an 8:03 start has already made their decision by 8:00:13.
3. Something audible beats an accurate "starting soon" card. A visual placeholder does not survive a listener who opened the tab in the background.

This is also why we run always-on channels in the gaps. When nothing is scheduled, our platform still has audio playing — classical, an eclectic mix, and 1940s old-time radio — so that arriving at a quiet moment is not the same as arriving at silence.

Is It the Platform's Fault or the Broadcaster's?

Usually neither. It's a coordination problem.

The clearest example we have: on one evening, four people arrived from four different countries within 48 minutes — and not one of them ever overlapped with another. Our logs recorded others_in_queue=0 at every single check while each was waiting. Two of them, from Saudi Arabia and the United States, arrived 60 seconds apart and each stayed about three seconds.

Any one of them would have had a conversation if they had waited a minute. None knew the others were coming.

That is the structural problem with live audio: supply and demand have to be present simultaneously, and at small scale they almost never are. Scheduling is the standard fix. Always-on audio is the other one — it converts "nothing is happening" into "something is happening, and you can stay while you wait."

What Should You Actually Change?

Ordered by how much it moves the number, based on what we've measured:

1. Start on time, audibly. Not a countdown graphic — sound. Talk over your own intro if you have to.
2. Publish a fixed schedule and keep it. Habit beats quality for retention at small scale. A mediocre show every Tuesday outperforms a great show whenever.
3. Fill the gaps with something legal to play. Public domain and CC0 audio costs nothing and needs no licence — see where to find it.
4. Tell people how long to wait. "Starting at 8, doors open 7:55" sets an expectation that survives thirteen seconds.
5. Give arrivals a reason to stay one more minute. That minute is the entire ballgame.

Frequently Asked Questions About Stream Retention

If people leave in 13 seconds, does audio quality even matter?

For the people you are losing, no — and that reordering is the useful part of the finding. Someone who leaves at 13 seconds has not heard enough to judge your microphone, your mix or your content. Production value determines whether the listeners who stay come back; it has almost no influence on the ones leaving before anything happens. Broadcasters routinely spend their first budget on audio gear when the loss is occurring upstream of anything that gear could fix.

Does having listeners help you get more listeners?

Almost certainly, and it is the awkward part of the problem: the empty room is self-reinforcing. A listener who arrives to silence and a zero next to the station name has two signals telling them nothing is happening, and both are accurate. That is why the first listener is the expensive one and the tenth is nearly free — and why a stream that plays something between shows is attacking the cause rather than the symptom.

Is it better to broadcast on a schedule or always-on?

A schedule is what builds an audience; always-on is what stops you losing the one that arrives off-schedule. They solve different halves. A published time gives people a reason to come back deliberately, which is the only mechanism that compounds. But most arrivals are not deliberate — they are someone clicking a link at a random hour — and for them the question is simply whether anything is playing. If you can only do one, do the schedule; the value of always-on is that it stops the schedule leaking.

An Honest Caveat

We do not yet know whether always-on audio actually increases how long people wait. We shipped it recently and have no post-launch comparison. What we can say is that the 13-second median is measured over 1,842 real sessions, and that people are demonstrably not leaving because they were disappointed by content — they leave before there is any content to judge.

We would rather publish the number with that caveat attached than imply a result we haven't earned.

Where This Data Comes From

Client-side telemetry from HereSay, a voice chat platform, covering 30 days and 1,842 sessions that ended without a connection. "Abandoned" means the person opened the platform, entered a waiting state, and left without ever connecting. Timings are measured client-side from the moment waiting began.

For context on the platform generating it: 15,000+ conversations across 60+ countries to date, with a median call length far shorter than most people expect — the reason we started measuring the waiting period in the first place.

Sources: our own client telemetry (1,842 sessions, 30 days) · Akamai/Limelight startup-delay research summarised by Bitmovin · OTTVerse: video startup time explained · Mux: live streaming metrics that matter

Thinking about broadcasting? See how to start a talk radio station without licensing fees, or compare the platforms.