Skip to content

Latest US poll: 63% say gas prices are affecting their household budget a great deal or a moderate amount

See the poll

Audio ad testing

Hear how your audio ad lands

Play your radio, podcast or streaming spot to the people you want to reach, then ask what they remember. With no picture to lean on, the brand name, the voice and the music have to do all the work.

3AI follow-up questions on each voice answer, at most
5audience groups compared side by side, at most
20question types, each with its own chart

Why test the spot

No logo on screen to help

In an audio ad the brand only exists if someone says it, sings it or plays a sound people already know. Listeners are often driving, cooking or at the gym. A test shows whether the brand name gets through and whether the message survives a single listen.

What can go on the bench

Radio adsPodcast host readsProduced spotsStreaming audio spotsScratch recordingsJingles and sound logos

Spot by spot

Five measures for each spot

Each row is one measure, named the way the report names it.

Base: Two groups of US adults who listen to podcasts; each heard one ad, once.

Shares of listeners. Feeling is the share who picked a happy or excited face.

Audio ad testing · scorecardBrand A · audio ad test report
Measure
HostHost read
Prod.Produced spot
Brand recallCould they name the brand after hearing the ad?
58%B
46%
MessageWhat they say the ad was telling them.
47%Bremembered the offer
39%remembered the offer
AppealHow much they liked what they heard.
62%
64%
FeelingHow it made them feel, picked on a face scale.
51%
55%
Voice and musicWhether the voice and the music suit the brand.
70%
66%

Heard once

Watch a spot play once, as on air

  1. 01The spot plays once
  2. 02What came through?
  3. 03Recall builds up, spot by spot
  4. 04The version to air

Step 1 / 4

The spot plays once

Listeners hear the ad inside the survey, once, as they would on air, before any questions appear.

Study set-up

Audio ad testing

  • StimulusTwo podcast ads: one read by a show host, one produced with music and a voice actor
  • WhoUS adults who listen to podcasts
  • DesignTwo groups; each hears one ad
  • ListeningOnce, as on air, before any questions
  • ThenBrand, offer, feeling, and answers out loud

Set up with our team, or yourself in CatalyseAI

Step 2 / 4

What came through?

Which brand was it for, what was it saying, and did the voice and the music suit the brand? People can answer out loud.

Question 1 of 9

Which company was the ad for?

The mattress one. Brand A, I think.

How well did the voice suit the brand?

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5

Not at allVery well

Next

"The host sounded like he actually uses it."

"Nice music, but I missed the name."

"Thirty nights free, I remember that."

Voice answer

AI follow-up: What made you remember it?

"The host said he sleeps on it himself, and he said the name twice."

  • Length
  • Words
  • Topic relevance 0.82

Transcribed, checked and searchable. Up to three follow-ups per answer.

Step 3 / 4

Recall builds up, spot by spot

Who named the company and who caught the offer, for each version. Set against the script, it shows where the name was said and whether it landed.

Live tally, as answers arrive

Host readThe name, said twice by the host0:30

Name · 0:04Name · 0:24
  • Named the company58%
  • Remembered the offer47%

Produced spotThe name, said once, under the music0:30

Name · 0:26
  • Named the company46%
  • Remembered the offer39%

Step 4 / 4

The version to air

The voice answers explain the numbers: why one version is linked to the brand more often.

Verdict

The host read goes on air

The produced spot needs the brand name earlier.

  • Host readAir it
  • Produced spotSay the name sooner

Four steps

How an audio ad test works

From the spot to the version that airs.

  1. 01

    Send the spot

    A finished radio ad, a podcast read or a streaming spot. A scratch recording is enough to choose between two scripts.

  2. 02

    Play it once

    Listeners hear the ad inside the survey, once, as they would on air, before any questions appear.

  3. 03

    Ask what came through

    Which brand was it for, what was it saying, how did it make you feel, and did the voice and the music suit the brand?

  4. 04

    Let them answer out loud

    People can reply by voice. Answers are transcribed, and an AI interviewer can ask up to three follow-up questions.

  5. Then: decide

    The produced spot needs the brand name earlier.

    Plan your study

On air

Example: host read or produced spot?

A mattress company has two podcast ads: one read by a show host, one produced with music and a voice actor. Two groups of US adults who listen to podcasts each hear one ad. The test shows which version people link to the company more often and which offer they remember, and the voice answers explain why.

  • Host read

    Air it

    Linked to the company more often, 58% against 46%, and the offer is remembered.

    Keep the name in twice.

  • Produced spot

    Say the name sooner

    Liked as much, but the name gets lost under the music.

    Move the name to the first five seconds.

Ask the studyAnswers from your own data

Why was the host read linked to the brand more often?

In the voice answers, people say the host named the company twice and sounded like a real user. In the produced spot the name came once, under the music.

It checks its numbers against your data and draws the chart.

What you get

What you get

  • Output 01

    Brand and message results

    Who named the brand and which message they took away, for each spot you test.

  • Output 02

    Spoken reactions, transcribed

    Voice answers with transcripts, the words people used most, and themes across all answers.

  • Output 03

    Listener groups compared

    Results for up to five groups side by side, such as podcast listeners and radio listeners.

  • Output 04

    Deck and data

    A PowerPoint deck, an Excel workbook and CSV files.

FAQ

Questions, answered

Can listeners replay the ad?

You choose. In one-play mode the spot plays once, as it would on air, before the questions appear. Or leave the normal controls on.

Can I test a script before it is recorded?

A rough recording is enough to compare scripts. If there is no recording at all, test the written lines with message testing first.

Can people answer by voice?

Yes. Voice answers are recorded in the survey and transcribed, and the report shows the words and themes that come up most.

Have a spot ready for air?

Tell us the format and who should hear it. We will suggest the set-up and the listeners to recruit.

Nearest office: Orlando, FL (corporate office)