Back to the blog

Research

Back-channeling: handling silence on the phone

Silence is the most uncomfortable thing on a phone call. Back-channeling fills the gap with the short acknowledgment sound the assistant gives while the caller pauses.

The Freya team

While people talk, the listener keeps giving small signals without taking the turn: “mm-hm”, “yes”, “I see”, “sure”. On the phone we cannot see the other person, yet these small sounds tell us we are being heard. Linguistics calls this back-channeling. We all do it every day without noticing. It is one of the invisible reasons a conversation feels fluent.

A voice assistant on the phone has to do the same job. Below we walk through the problem, the common wrong fix, and the path we prefer.

The most uncomfortable thing on the phone: silence

Face to face, body language tells us the other person is listening. On the phone, sound is the only clue we have. When the other side goes quiet for a few seconds, the human brain gets nervous right away: “Did the line drop? Did they hear me? Should I say it again?”

This gets critical when the caller reads out information piece by piece. Picture someone pausing in the middle of an ID number or a phone number. Older or more careful callers read digits in groups, stopping in between. If the assistant stays completely silent through that pause, the caller either assumes the line dropped or loses confidence and starts over. Both outcomes ruin the experience.

Why the classic fix falls short

The first idea that comes up is usually this: “let the assistant wait longer, say 5 seconds”. That is not a good fix.

  • It punishes the fast speaker. You impose a pointless delay at every step on someone who speaks fluently. The conversation feels artificial and slow.
  • It does not solve the real problem. You cannot solve a silence problem by extending the silence. You only create a longer gap.

Tuning fixed wait times one by one is fragile as well. Every step of the flow then needs its own maintenance.

A better path: back-channel sounds

The natural fix is not to impose a fixed wait time but to let the assistant behave like a person. When the caller makes a short pause (around 2 to 2.5 seconds), the assistant gives a natural acknowledgment sound that does not disturb them:

mm-hm...

This is the spoken version of “I am here, I am listening, take your time”. It does not cut the caller off and it does not push them with a warning like “please repeat your number”. It only makes its presence felt.

The approach serves both kinds of caller:

  • The fast speaker: They never pause, so the filler never fires. Nothing is delayed and the conversation moves quickly.
  • The slow or careful speaker: After the pause they hear a gentle acknowledgment. They feel someone is waiting patiently for them and they do not panic.

If needed, the assistant waits a short while again and adds a gentle follow-up: “yes, go ahead”. What matters is that no step ever leaves a long, dead silence.

Closing

Back-channeling looks like a small detail, and it changes the caller experience directly. It is one of the clearest things that separate a well designed voice assistant from a robotic one. Knowing when the turn is actually over is a separate question, and we cover it in the smart turn post.

In short

  • Silence on the phone creates doubt, above all while the caller is reading out numbers or details.
  • Raising the fixed wait time is the wrong fix: it slows the fast speaker down and leaves the real problem in place.
  • A back-channel filler (“mm-hm”) fills the silence with a natural acknowledgment. It never fires for a fast speaker and it calms a slow one.
  • It is one of the clearest details that separate a well designed voice assistant from a robotic one.

Hear this approach in your own flow

Let us walk you through Freya on a scenario of your choice.

Request a demo