Moving DataFlow and your projects to a new computer

What moves and what doesn't

Your DataFlow account follows you. Your projects stay on the computer they were made on, because each is encrypted with keys kept only there, so a copy of its files is no use to whoever has the copy.

Why a project's folder can't just be copied

Each project is a folder in DataFlow's projects folder (Settings → Folders shows where). It holds project.json (the project's name and ID), data.duckdb (interviews, corrections, checks and the activity log) and pii.duckdb (the personal data you protected). Each of the two stores is encrypted with AES-256-GCM under its own random key, and the keys are kept in the system's keychain (Windows Credential Manager, the macOS Keychain, the Secret Service on Linux), never in the project files.

DataFlow has no way to export those keys today: no key file, no recovery phrase, no password to type in elsewhere. Copy a project's folder to another computer and DataFlow lists the project but says it can't be opened, because its key isn't in that computer's keychain. Nothing in it is changed, and Take it off the list moves the folder to set-aside without deleting it. So a project moves the long way: download its data on the old computer, then set it up again on the new one.

What setting a project up again brings back

Connected to the same form on the same server, the new project's first sync downloads every interview the server holds. The checks DataFlow writes from your form are added again on their own.

What doesn't come across: the corrections you made, which problems were fixed or confirmed, interviews you set aside, checks you added or changed yourself, the activity log and the assistant's conversations. They stay in the old project, and its download keeps the record: corrections.csv lists every correction with who made it, when and why, including ones later undone, and dataflow.do replays them in Stata from raw.csv. The new project starts its own log, so problems you dealt with on the old computer come up again, and the correction log shows what you decided about each. That makes the best time to move the gap between survey rounds; if fieldwork is running, consider finishing the round on the old computer.

The vault

If a project writes raw data to a vault, the vault is a single .hc file in VeraCrypt's format, wherever you chose to keep it. It opens with its password, its keyfile or both, in VeraCrypt as well as DataFlow, so it moves like any other file: copy it, and its keyfile if it has one, keeping the keyfile somewhere other than beside the vault. Without the keyfile no one can open it. If DataFlow kept the vault's password in the old computer's keychain, make sure you know the password before you leave; DataFlow doesn't show it.

DataFlow can't take over an existing vault, so on the new computer create a new one (on the Download page, under Outputs after each step) with a new file name, so the old one isn't replaced. The old vault stays a record of what came before, and VeraCrypt still opens it.

Your survey server

Server passwords and sign-ins are kept in the keychain too, so the new computer asks for them again as you connect each project:

• SurveyLoopr: sign in once with your SurveyLoopr account, in Settings → Account. No server password is needed.
• SurveyCTO: the server, username and password, and for an encrypted form its private key file. Bring the key file with you: DataFlow's copy stays in the old computer's keychain.
• ODK Central: sign in with your Central account. DataFlow keeps only a session token, never the password.

If the old computer is lost or broken

Then its projects can't be opened anywhere. Their keys were on that computer, and DataFlow never sends them to us or anyone else. Set the projects up again from your server, and use your last download, your output folders and the vault for what was decided on the old computer.

To be ready for that, switch on the outputs on each project's Download page. After each sync, DataFlow then writes the vault and the cleaned data, with its codebook and correction log, to folders you choose. Put them somewhere you back up.

Step by step

1. Download each project on the old computer. On the Download page, choose All of them, tick Codebook and Do-file and raw data, and save; the correction log is always included. Add personal data only if you need it (that's logged), and move the folder as your data protection plan allows.
2. Note what you set up by hand. Write down the checks you added or changed, the project's settings, its outputs and schedule, and any back-check surveys linked to it.
3. Copy the vault and its keyfile. Copy the vault's .hc file and its keyfile, if it has one, and make sure you know its password.
4. Install DataFlow on the new computer. Download it from letdataflow.com/download. On a Mac, open it the first time with Open Anyway in System Settings → Privacy & Security.
5. Sign in. In Settings → Account, sign in with DataFlow, or with SurveyLoopr if your seat comes with a SurveyLoopr server. Your plan covers up to three computers.
6. Set each project up again. Start a new study with the same name and form, and connect it to the same form on your server, or import the same files; the first sync downloads every interview. Then review personal data on Protect, add your own checks back, create a new vault and turn the schedule on.
7. Check the move worked. Compare each new project with the old one: its interviews, checks and settings. Run a sync, and open the copied vault in VeraCrypt to be sure it arrived whole.
8. Stop the old computer. On the old computer, turn off Sync on a schedule on each project's Import page, then sign out in Settings → Account; it can still open and download each project. Once you're sure nothing is missing, remove DataFlow and its data as your data protection plan says.

Questions