Turning a complex report into actionable audio summary (FeedbackTime case study)
Recently, FeedbackTime partnered with Podego to turn their info-dense reports into easy listening audio recaps. Let’s dive into how that came about, and what resulted.
FeedbackTime’s mission and methods
The aim of FeedbackTime is simple, to help customers pursue and achieve various transformation goals. They help organisations with change management by collecting and analysing data and consolidating it into reports that contain insights as well as concrete recommendations.
It’s all about helping customers focus on implementing recommended steps instead of spending valuable time collecting and analysing data.
The Challenge: Communicating concrete actions to different target audiences
FeedbackTime reports can be long, dense, and difficult to consume for different sections of customer demographics. They wanted to see if an engaging audio version could be created and provided Podego with a 7-page white paper and a 1-page background document.
The client in question needed the audio to appeal to different target audiences:
For NGO project managers.
Extended workforce that were not native speakers and required easier wording.
“We have a customer who has a transformation project where communication to an extended distributed workforce in multiple languages and easy words would be helpful to support the transformation.”
The project goal: To cut down report consumption time and communicate concrete next steps to different target audiences.
Our Approach
Podego supported FeedbackTime by creating an engaging audio companion to the report. We had 3 aims.
To provide an alternative media option for those who prefer listening over reading text.
To create an audio summary that is easy to ingest and actionable so that listeners would be able to take actions without reading the lengthy report.
That the provided summary could generate curiosity and stimulate further inquiry into the report.
How did we achieve this?
System prompts with LLMs! For this project, a unique system prompt was created that could summarise the FeedbackTime data and highlight key findings, facts, and figures followed by actionable recommendations.
It was very important to capture FeedbackTime’s brand personality, which is based on the concept of Radical Candor, a term coined by Kim Scott.
Next, we turned that into an easy to understand and lightweight audio recap. We looked for a voice that resonated with FeedbackTime’s core values and selected one with their input. Now we were ready to give the script a voice with Text-to-Speech tools.
Some specific project challenges:
The entire project was in German, and therefore we needed to take special care with the Text-to-Speech model and utilise the help of a native speaker to ensure quality and authenticity.
The word “Engagement,” is an interesting case in German. There are two ways of pronouncing it in German. The pronunciation we’re aiming for was the one that sounds French. We found that the Text-to-Speech model couldn’t be consistent with the chosen pronunciation and instead use both pronunciation in one audio. We tried using the International Phoneme Alphabet (IPA) pronunciation wasn’t possible because that function currently works mostly for English models. Eventually, we achieved the desired outcome by using the Alias method and had the Text-to-Speech model literally pronounce it as “anggazhemong” to get the desired effect.
The Result
A Five minute audio recap that delivered an engaging experience to the listener while delivering key insights and recommendations. This was achieved in a very short amount of time and showcased how LLM assistance can help to streamline peripheral processes like audio recaps for companies like FeedbackTime.
