TBD.

Krisp has been a noise cancellation app, with the ambition of improving the overall effectiveness of communicators - so going beyond clarity, and also act as a speech coach. We had access to the audio stream and were thinking on real time improvements. To test this, the idea was to give a small summary at the end of the call - this way, we can get feedback if the idea is sound, what people are interested in and we realised that real-time feedback is harder to design, so this was faster to test and contributed to our overall hypothessis.

So.

We designed the project in two rounds, first just to show talk time, second round also adding talk pace and filler words used.

First round: establish tech infrastructure, use existing mechanisms and tech as much as possible. Had some arguments with the PM if this should be just feedback or also go viral - after release saw many people sharing on social media, so it did get some viral traction.

Second round more tricky, also needed to balance attention and figure out how to not only tell data, but also judgments, so start giving feedback. With talk pace it’s easy, with filler words we did some research what would be acceptable cases. Also there were concerns about accuracy, so had to give some asterixes over the interfaces.

Both times: rapid iterations, check with engineering and lightweight testing.

Process - check competitors, check with stakeholders, rapidly explore open questions and variations, while keeping feasibility as much in check as possible. Keep principles in mind - don’t overwhelm the user.

Finally after a promising beta, the company decided to take a different strategic route and roll-out was halted.

What was interesting how we evolved our process as we went. Initial was: PM did PRD, and design came afterwards. We iterated on the process (design happens parallel to PRD, as we not only explore the product, but also our understanding)

Updated: