Download audio audit files separately

Proposal

Audio audit files are very useful, but can be problematic to download. When HQ prepares data for export, there can be a fairy large load for preparing data files and associated binary files. For under-powered servers, exporting audio files can crash the server. When downloading the export archive, there can be problems downloading large archives with slow or unreliable connections. In this context, downloading data can either take longer (because of the larger size of the zip file transmitted) or require several retries (due to dropped or timed-out connections). For home-spun file servers, downloading files from Survey Solutions to that location (e.g., NextCloud) might place a high resource burden on the file servers until everything is downloaded and indexed on the destination file server.

Plus, you may not want audio audit files for every use. For example, when building a report or feeding data to high-frequency checks, getting audio audit files simply slows things down. For other use cases, however, audio audit files are what is needed. For example, if a team is listening to audio audit files to check whether questions are asked well and survey protocol is followed, audio audit files are what is needed. In fact, it might be the case that only those files are needed (since survey data can be more easily viewed in the interview review UI than via some statistical software package).

To address this, consider one of two features, if not both:

  • Create an API endpoint to download audio files for individual interviews.
  • Create a download option to get audio files only (both in the GUI and in the API)

Affected subsystems

This would affect Headquarters only.

Users

Anyone using audio audit files. Currently, from what I can tell, this would mainly address CATI surveys, where interviews are shorter (audio audit files smaller) and listening to interviews is more prevalent and feasible. However, the addressable market here is any survey that is willing to review audio files from interviews, following general advice to audit 10-20% of cases. See reference here. That market is likely to expand as survey implementation costs get squeezed and traditional face-to-face data quality supervisions missions get harder to fund.

Alternatives

Because of the size of audio audit files, audio audits are sadly under-utilized. File size matters both for transmission from the field and for download at HQ. See more here.

When they are used, to my knowledge, they’re typically downloaded only in high-connectivity contexts.

Compatibility

Depending on how the feature is implemented, users who expect exported data to contain audio audits might not find them there (or need to select an option to include them).

1 Like