Choosing a provider
Compare Google Drive, Dataverse, and Zenodo, and see how each one handles your files once data starts arriving.
Comparing the providers
DataPipe writes each participant's data into your own account with one of the three storage providers below. The data is yours the whole way through. DataPipe only ever asks for permission to add files.
Google Drive: your own Google Drive. Data lands in a folder in My Drive/DataPipe, or in a parent folder you pick.
Limits: free Google accounts share 15 GB across Drive, Gmail, and Photos. Uploads stop once that is full.
Dataverse: institutional repositories run by universities and consortia (Harvard Dataverse, Borealis, DataverseNL, and others). Data lands in a draft dataset, in a collection you name.
Limits: your installation sets its own file size and storage limits. API tokens expire, often yearly, and DataPipe cannot renew them. Data stops arriving until you reconnect.
Zenodo: an open repository run by CERN. Data lands in a deposition that stays private until you publish it, and publishing issues a DOI.
Limits: 100 files and 50 GB per record. DataPipe merges completed sessions into archives so a long study stays under the file limit.
Accounts you need
You'll need an account with whichever provider you choose: Google Drive, your Dataverse installation, or Zenodo.
Once you have an account, the next step is giving DataPipe permission to write to it. Connecting and reconnecting
One provider per experiment
This choice isn't permanent. You can use a different provider for each experiment, and you can connect more than one provider to your account and pick between them whenever you create a new experiment.
What you can't do is change the provider of an experiment that already exists. Each experiment writes to the one Drive folder, Dataverse dataset, or Zenodo deposition that DataPipe created for it. To collect the same study somewhere else, create a new experiment on the other provider and point your experiment code at the new experiment ID. Data you already collected stays where it is.
Provider-specific behavior
The three providers behave differently once data is arriving. These are the differences you're likely to notice mid-study.
Google Drive
DataPipe creates one folder per experiment, named after the experiment, either under a parent folder you pick or under a folder called DataPipe in your Drive. It also creates data and data/raw inside it right away, so they're ready before the first participant submits.
DataPipe sets no file count or file size limit on Drive. The real constraint is your account's own quota. Free Google accounts share 15 GB across Drive, Gmail, and Photos, and uploads stop when that is full.
Drive allows two files with the same name in the same folder and never reports a conflict. That means DataPipe's own filename record, its list of the names an experiment has already used, is the only thing preventing a duplicate. Drive stores a file under the last part of its name, so two submissions that differ only in their folder prefix count as the same name.
Dataverse
DataPipe creates a draft dataset in the collection you name and never publishes it. Publishing is your decision.
Dataverse accepts only one write to a dataset at a time. If two participants submit at the same moment, Dataverse refuses the second write, so DataPipe queues that submission and retries it about a minute later instead of failing it.
Dataverse never rejects a duplicate filename. It quietly renames the new file instead, so a second README.md arrives as README-1.md.
Dataverse converts uploaded CSVs into its own archival .tab format unless it's told not to. DataPipe tells it not to on every write, but installations older than Dataverse 5.11 ignore that instruction without any error. DataPipe checks your installation's version when you create an experiment and warns you on the form if it's too old. JSON data is unaffected either way.
Zenodo
DataPipe creates an unpublished deposition and never publishes it. The link from your dashboard opens the deposit editor, not a public record, and no DOI is minted until you publish.
A Zenodo record holds at most 100 files and 50 GB. No other provider caps the file count. That cap is why DataPipe merges completed sessions into archives on Zenodo as collection goes on, and why finalizing a study into one archive is a Zenodo-only feature.
Zenodo has no folders, so a record with metadata turned on shows flattened names rather than a folder tree.
Writing a file that already exists on Zenodo replaces it, with no error. DataPipe's filename record is the only thing standing between a repeated filename and a lost session.
How DataPipe keeps filenames unique, what your files end up named, and what the archives in your storage are. Filenames, archives and your storage