Sequel: Reverse Engineering Voiceflow Data Export

After the breakthrough with Voiceflow, we went on to redo our HCI design (for the second time) and start working on an MVP implementation using that. As we were almost there, we noticed the appearance of an interesting button in Voiceflow’s web UI:

voiceflow export

Voiceflow have implemented a data export feature![1] And it turns out that this .vf file is actually in a similar JSON format to the data we scraped earlier. Considering using this file as the input to our system instead, we see:


  • Likely a more stable file format
  • Lower maintenance overhead (won’t break when Voiceflow update their API)
  • More intuitive UX (won’t ask the user for their Voiceflow credentials)


  • We had spent 1 month (25% of this project’s timeline!) on something that may not have a key role in our project.

Having this feature does not render our previous breakthrough meaningless. In fact, it adds value to the project. Having discovered our reverse engineered method earlier, this reduces our dependency on Voiceflow’s file export feature and formatting. The client recognises the value of the procedure we discovered as adding redundancy and hence reducing the overall risk of the system.

In the end, we ended up implementing our system with this file input while keeping our reverse engineered approach in the form of proof of concept code for redundancy.

[1] “The Ultimate Guide to #VFV2: What’s New With Voiceflow? | Voiceflow”,, 2021. [Online]. Available: [Accessed: 21- Mar- 2021].