How to upload a CSV or Excel file to DataLens
The shortest path from a file on your laptop to a profiled dataset you can clean, model, chart and publish.
What you will do
You will load a delimited or spreadsheet file into DataLens, check the columns and types it detected before committing to them, and finish with a dataset that every other screen in the platform can use.
This is the usual first step. Almost every other guide assumes you have a dataset loaded, and this is the fastest way to get one.
When this is useful
- Someone sent you an extract and you want to understand it before writing any code.
- You are prototyping a model or a pipeline and want real data in front of you quickly.
- You need to check the quality of a file — nulls, duplicates, odd types — before it goes anywhere near a warehouse.
- You are learning data architecture and want a real dataset to practise on.
Before you start
- A file in a supported format — CSV, TSV, Excel (.xlsx or .xls), JSON, or OpenDocument (.ods). Those are the exact formats the upload control accepts.
- A workspace — Every dataset belongs to the workspace you are signed into. Nothing else needs to be set up first.
Steps
Open Data → Connections
In the top navigation, select Data. The second row shows two entries: Connections and Datasets. Choose Connections.
Connections opens on a row of source tabs — Upload File, Documents, Database, REST API, FTP / SFTP, CDC Sync, CRM, Data Lake and AI Model. Upload File is selected by default.
You should see the Upload File tab active, with a drop area accepting CSV · TSV · Excel · JSON · ODS.
Choose your file
Drop the file onto the upload area, or click it to open your file browser. You can select more than one file at a time.
The upload area accepts .csv, .tsv, .xlsx, .xls, .json and .ods. Anything else is not offered by the file picker.
The file appears in the list with a progress indicator while it is read.
Check the detected schema before committing
Select the file in the list. DataLens shows the columns it found, the type it inferred for each one, and how many nulls are present.
This preview is the point at which to catch a mis-read file — a header row treated as data, a numeric column detected as text because of a stray value, a date format that did not parse.
A table of Column, Type and Nulls for every column in the file.
Pick the sheet, for Excel files
An Excel workbook can hold several sheets, and a dataset is one sheet. Choose the sheet you want before loading.
The sheet name is kept with the dataset and is shown later in Datasets, so you can tell two sheets from the same workbook apart.
Load the file
Confirm the upload. DataLens reads the rows, records the schema and profiles the columns.
If a dataset with the same name already exists you are told so rather than silently overwriting it.
The dataset appears under Data → Datasets with a row count and a column count.
Confirm it landed correctly
Go to Data → Datasets and select your new dataset. The summary shows Rows, Columns, how many are Numeric, how many are Text, and whether any column Has Nulls.
A row count that does not match the file, or a column typed as text when you expected a number, is worth fixing now rather than three screens later.
A dataset card showing Rows, Columns, Numeric, Text, Has Nulls and the Source.
What happens next
With a dataset loaded, every other stage of the platform opens up: Prepare for cleaning and transforms, Model for source and target models, Govern for profiling and PII detection, Analyse for charts, and Deliver for publishing and export.
If the file needs cleaning first — and most do — the natural next step is Prepare → Clean & Transform.
Example
An analyst is handed customers.xlsx with three sheets. They upload it, select the Customers sheet, and see in the preview that signup_date came through as text rather than a date because a handful of rows use a different format. They load it anyway, then fix the type in Prepare with a Cast Type step — which keeps the original file untouched and the change recorded.
Tips
- Check the Nulls column in the preview before loading. It is the fastest quality signal you will get, and it costs nothing.
- Upload several related files in one go — they become separate datasets and can be joined later in Prepare or Model.
- A column detected as text when it should be numeric is normally one bad value, not a bad file. Load it and fix it in Prepare rather than editing the source.
- Give files meaningful names before uploading. The file name becomes the dataset name, and it is what you will be picking from lists later.
Limitations
- The upload control accepts CSV, TSV, XLSX, XLS, JSON and ODS only. Parquet, Avro and ORC are not accepted here — use the Data Lake connector for object storage formats.
- One sheet becomes one dataset. A workbook is not loaded as a single multi-table object.
- PDFs and Word files are not tabular uploads. They belong in a Document Collection, which is a separate source tab and works differently.
- Uploading the same file again does not merge with the existing dataset; you are warned that the dataset already exists.
Related how-to guides
Connect a database
Connect PostgreSQL, MySQL, SQL Server, Redshift, BigQuery or Snowflake, browse the tables and load what you need.
Clean a messy dataset
Drop nulls, fill gaps, remove duplicates and cap outliers — with a live preview before anything is applied.
Profile a dataset and find PII
See what is really in each column, and find the personal data before it ends up somewhere it should not.
Build a document collection
Load Word, Markdown, HTML or text documents as a collection — stored whole, not parsed into rows.
Related questions
More of these on the DataLens FAQ page.
What file formats can I upload to DataLens?
The file upload accepts CSV, TSV, Excel (.xlsx and .xls), JSON and OpenDocument (.ods). Object storage formats such as Parquet are handled by the Data Lake connector instead, and documents such as Word files are handled by Document Collections.
Does DataLens change my original file?
No. Your file is read once on upload. Cleaning and transform steps are applied to a working copy inside DataLens, and the original file on your machine is never modified.
Can I upload more than one file at a time?
Yes. The file picker accepts multiple files and each becomes its own dataset. You can then join or append them in Prepare or Model.
Try this in DataLens
DataLens is in private beta. Request access and work through this guide on your own data.
Request beta access