Speech-to-Text at POS

AI-based technology for understanding your customer.

Step 1: Voice Recording
Qualitative insights
through audio recording at the POS
Gather more context information through recording of uncut customer statements via audio.
Step 2: Speech-to-Text
Transcription through AI Numerous audio files are transcribed to complete text statements by means of speech-to-text-algorithms in shortest time.
Step 3: Coding
Coding by means of AI Text statements will be coded by a trained and AI-controlled model independent of any environmental influences and human signs of fatigue.
Step 4: Insights
Qualitative insights &
authentic storytelling
You receive comprehensive insights and can, therefore, validate, or confirm, the quantitative assessments once again via the statements. Furthermore, precise recommendations for action can be derived by means of descriptive, multivariate and inferential statistics and driver analysis.

Immerse into the world of your customers at the POS

AI-based technology finally allows analyzing
a large number of customer stories at the POS,
accurate, fast and in high quality.
We identify your success drivers for more insights.

Speech-to-Text – Video Switch on sound

Searching for storytelling at the POS

Over the years, the spoken word has become hard currency in market research. In the past, it was usual to finish a quantitative survey with open-ended questions, whether the respondents had further remarks. Today, more and more qualitative questions are seen even in standardised surveys.

The need to constantly generate more qualitative insights even with large caseloads means the respondent’s open-ended statements are becoming ever more important – especially in combination with a closed valuation.

Challenge „Human“ at the POS

The other world – people always have less time per se, especially at the POS. As personal surveys are becoming ever more challenging, those surveyed also have to be convinced about how meaning participation is. So how can more meaningful insights be obtained through more qualitative answers at the POS? IWD deals with this question on a daily basis in the more than 3 million personal surveys conducted every year at the POS in 25 European countries.

The solution can only be: the spoken word shall prevail.

More qualitative insights through audio recording at the POS

To be able to gather completely qualitative statements in a short time, IWD relies on audio recordings. For this, IWD uses the recording ability of smartphones; the indispensable device for computerized, personal surveys. The respondent’s statements are gathered via audio recording during the survey mode. The interviewer neither needs to type nor interrupt the survey, they only need to start the recording at the correct place. The statements will be recorded for the entire interview and directly uploaded to the server.

High acceptance with the respondents and interviewer

Our experience shows a high willingness to participate and provide information amongst the respondents. 85% of the respondents allowed for their statements to be recorded. The remaining 15% had to be manually noted down by the interviewer as text within the survey software.

AI-based evaluation of the audio files

To be able to evaluate the large amount of audio files, these are transcribed with the help of a speech-to-text algorithm. Thanks to the automatic transcription, one gets complete text statements in the shortest time. A manual transcription in scientifically acceptable times would not be possible in this instance.

As the speech-to-text algorithm has not yet mastered all facets of human language (something that we know all too well from our assistants on our smartphones), the statements are checked and corrected when necessary. Through machine learning, the transcription will become better every time and human correction will become less and less necessary.

When needed, the texts are translated and coded after the transcription. Even here, IWD employs an AI-algorithm. This is trained and monitored as a virtual coder. Therefore, we ensure an ever more constant coding, independent of any environmental influences and human signs of fatigue.

The coded statements can subsequently be used with a higher significance, for example, for descriptive, multivariate and inferential statistics as well as for driver analysis.