Data storage
Every organisation on AI Insights 365 gets 2 TB of data storage, included with every plan. It holds the files you upload and the data you import, so a report can work with millions of rows without slowing down.
What counts towards your storage
| Uploaded files | Excel workbooks and CSV files you upload |
| Imported data | Large files and OData feeds are imported as compressed Parquet files that your charts read |
| Included | 2 TB per organisation |
| More space | 100 GB for $100, one-time, no subscription |
Live connections such as SQL Server, Power BI, Google Sheets or a file URL don't use your storage: their data stays where it is.
See how much you are using
Organisation admins see a Data storage card in Settings → Users & Licensing: a usage bar, how much is used out of your total, and how much is included or purchased.
The number is counted again every time a file is uploaded, replaced or imported.
Large Excel and CSV files (up to 500 MB)
Files up to 25 MB are uploaded and read straight away, as before.
Files from 25 MB up to 500 MB upload directly to your organisation's storage, with a progress bar:
- Choose File Upload and pick your file. The upload starts at once and shows a percentage.
- Save the datasource when it reaches 100%.
- The file is imported in the background. Every sheet becomes its own table, and the datasource panel shows the progress, for example Importing… 1,250,000 rows so far.
- When it says Imported, build charts as usual. Totals and averages are calculated on the server over every row, not a sample.
Replacing the file starts a fresh import. The old copy is removed, so it stops counting towards your storage.
How imported data is stored: Parquet
Imported data is saved as Apache Parquet, the open columnar file format used across modern analytics:
- Compressed: each table is stored column by column and compressed, so it is usually much smaller than the original workbook. That leaves more of your 2 TB for data.
- Fast: a chart reads only the columns it needs. A Sales by Region chart over 10 million rows reads two columns, not the whole table.
- Typed: column types (numbers, dates, text) are worked out from the whole file, not only the first rows, so a value on row 4 million is never cut off or misread.
- Queried on the server: charts, filters and AI questions run as SQL over the Parquet files with a fast analytics engine (DuckDB), and the results come back already totalled.
Each table of a workbook, and each OData entity set, is its own Parquet file. A re-import writes new files and removes the old ones, so a report never mixes old and new data.
You upload Excel or CSV as usual: the conversion to Parquet happens for you.
OData feeds: an imported copy you refresh
An OData datasource keeps an imported copy of each entity set you select, so reports open fast and don't depend on the feed being up.
Open the datasource to manage it:
- Refresh now imports the latest data straight away.
- Refresh daily re-imports it once a day.
The panel shows when it was last imported and how many rows came in.
When storage is nearly full
When your organisation reaches 90% of its storage, every organisation admin gets an email and an in-app notification. It says how much is used, and links to the storage card.
The warning is sent once. If usage drops below 85%, because you added space or removed data, you'll be warned again the next time it reaches 90%.
When storage is full, new uploads and imports are refused with a clear message. Existing reports keep working.
Add more storage
- Go to Settings → Users & Licensing → Data storage.
- Pick how much to add: 100 GB, 200 GB, 500 GB or 1 TB.
- Choose Add storage (one-time) and pay by PayPal or card.
The space is added as soon as the payment goes through, and you get an invoice by email. It's a one-time purchase: there's no monthly charge, and the space stays with your organisation.
Free up space
- Delete datasources you no longer use: their files and imported data are removed.
- Replace an old file instead of uploading a second copy.
Security
- Your files and imported data are stored encrypted in Microsoft Azure, kept apart per organisation and readable only through AI Insights 365 with your permissions.
- Large uploads use a short-lived, single-use link that can only write that one file.
- Row-level security applies to imported data exactly as it does to live data: each reader sees only their rows.