
Can Tableau Connect to S3?
Carlos GarciaPublished Amazon S3 is where an enormous amount of business data now lives. It is cheap, it scales without anyone thinking about it, and it has quietly become the default landing zone for exports, logs, warehouse extracts and Parquet files that nobody wants to pay warehouse prices to store.
So the question comes up constantly on analytics teams: can Tableau just read it?
The short answer is yes. The useful answer is more interesting, because the connector most guides point you at carries a deprecation notice on one Tableau property and no notice at all on another, and a rebuilt version is in beta. If you are about to build reporting on top of S3, that distinction matters more than the setup steps.
Can Tableau Connect to S3? The Short Answer
Yes. Tableau ships an Amazon S3 connector, and Tableau's own documentation says it "Applies to: Tableau Cloud, Tableau Desktop, Tableau Prep, Tableau Server."
Two things about it surprise most people.
First, it is not a built-in native connector in the way the Redshift or Athena connectors are. You install it. Tableau's instructions begin: "Start Tableau and under Connect, select Amazon S3 from the list of Additional Connectors." It arrives through the Tableau Exchange rather than shipping in the box, and "Amazon S3" does not appear in Tableau's published tech-specs connector lists at all.
Second, and far more important: Tableau's Exchange listing for that connector is now titled "Amazon S3 - Deprecated."
The listing states the reason plainly: "Tableau is deprecating this connector for security concerns." It then points at a successor: "Please use the new Amazon S3 connector in beta today, and will GA in 26.3." That version string is Tableau's own; the page never expands it to a full year.
Is your reporting stack built on plumbing that quietly got deprecated? Get a free audit and find out what your analytics setup is actually standing on.
Meanwhile Tableau's live help page for the same connector carries no deprecation, retirement, legacy or sunset language anywhere, and does not mention the replacement. Two Tableau properties disagree about the health of the same product. That is worth knowing before you commit a quarter of reporting work to it.
Free audit
Where do you stand in AI search?
A manual SEO + AI search audit of your site: how ChatGPT, Gemini, Perplexity and Claude read and cite it today, and what to fix first. Report within 48 hours, no credit card.
Get my free auditOr have it all done: AI SEO services, $999 once
What "Connecting to S3" Actually Means in Tableau
S3 is object storage, not a database. There is no query engine inside it, so "connecting" means something different from pointing Tableau at a SQL server.
Reading Files Directly
The S3 connector reads files out of a bucket and loads them into Tableau. Tableau is doing the parsing, so the file format and its encoding matter enormously, and the volume you can handle is bounded by Tableau rather than by S3.
Querying Through an Engine
The alternative is to put a query engine in front of the bucket and connect Tableau to that engine instead. Amazon Athena is the canonical route, and it is the only Tableau connector page that explicitly ties itself to S3 — its setup asks for the "Name of the S3 staging directory," with Tableau's own example being s3://aws-athena-query-results-123456785678-us-eastexample-2/.
Why the Difference Decides Your Architecture
Direct file reads are simpler and need no extra AWS service. Engine-backed access scales past file-size ceilings, pushes work down to the engine, and handles partitioned datasets that a file reader cannot. Most teams that outgrow the connector end up at Athena, and Tableau's documentation nudges you the same way.
How to Connect Tableau to Amazon S3, Step by Step
Tableau's documented flow is short, but it has prerequisites that trip people up because they are AWS-side rather than Tableau-side.
Before you start, gather the three things Tableau asks for: "The AWS region of your S3 bucket," "The S3 bucket name," and "Your AWS IAM access key for your S3 bucket (key ID and secret access key)."
- Install the connector. In Tableau, under Connect, choose Amazon S3 from the list of Additional Connectors. Tableau restarts as part of the install.
- Reopen the connector. Tableau's instructions are explicit that the second step happens after a restart: "After Tableau restarts go to Connect, and then select Amazon S3 from the list of installed connectors."
- Authenticate. Supply the region, bucket name, access key ID and secret access key. The IAM user needs read permission on the bucket.
- Pick your files. Drag the file you want. If you need several files treated as one table, you can union them — Tableau says "You can union files from your S3 bucket."
- Publish with embedded credentials. This one catches teams late. Tableau requires that "Workbooks and data sources must be published using the 'Embedded password' authentication option," and notes that "'Prompt user' isn't currently supported."
The authentication model is narrower than most AWS shops expect. Tableau states that "Only Amazon IAM user secret key/access key authentication without session token is supported." There is no documented support for IAM role assumption, cross-account roles, instance profiles or temporary credentials. If your organization has banned long-lived access keys — and many have — this connector may simply be off the table for you on policy grounds.
Tableau does not say that the static-key model is why the connector is being deprecated for "security concerns." The two facts sit next to each other on vendor pages without Tableau drawing the line, so treat the connection as your own inference rather than as something Tableau published.
What the S3 Connector Supports, and What It Refuses
This is where the direct-file route either fits your data or does not, and the answer is usually visible in five minutes.
Supported formats, in Tableau's words: "Parquet, .csv, compressed .gz, and Excel files are supported at this time."
The qualifications matter as much as the list:
- CSV is narrower than it sounds. "Only comma-delimited .csv files are currently supported." Semicolon, tab and pipe-delimited exports do not qualify.
- Encoding is fixed. "Only UTF-8 encoding is supported."
- Everything arrives as text. "All data is, by design, imported in string format," so every numeric and date field needs retyping in Tableau.
- You cannot mix formats in one connection. Parquet and CSV files cannot be unioned or joined together inside a single S3 connection.
The size ceilings are hard numbers, which makes them easy to plan against. "The file size limit is 15gb," and "The cumulative result set of a join or union can't exceed 15gb." Excel is far more restricted: "Excel files currently can't exceed ~100 mb due to performance issues with the Excel file parser in the connector."
Pulling cloud data into dashboards nobody reads is an expensive habit. Get a free audit and see which of your reports actually drive decisions.
Parquet brings its own constraints. Files need to match what Tableau's Hyper engine accepts, which rules out nested columns and several logical types, and supports a specific set of compression codecs. If your Parquet comes out of Spark with nested structs, expect to flatten it before Tableau will read it.
When to Use the S3 Connector, and When to Route Around It
Use the direct connector when your S3 data looks like flat files that a person could plausibly open: a nightly CSV export, a flat Parquet extract, a set of monthly files you want unioned into one table. Under 15 GB, comma-delimited, UTF-8, non-nested, and in a bucket you can mint an access key for — that is the connector's home ground.
Route around it in four situations.
Your organization forbids long-lived IAM access keys. The connector documents no alternative, so this is a blocker rather than an inconvenience.
Your data is partitioned in Hive style, or is genuinely large. A file reader with a 15 GB ceiling and no documented partition handling is the wrong tool; Athena was built for exactly this.
You need Iceberg tables. Salesforce's own support article is unambiguous: "Tableau currently does not support Iceberg tables in any products/connectors." The documented workaround is to export Iceberg data to Parquet through an Athena CREATE TABLE AS SELECT query first, then read those Parquet files.
You are standardizing for the long term. A connector whose vendor page says it is deprecated for security reasons is not where you want three years of workbooks pointing.
Choosing between a file reader and a query engine is the easy half. Get a free audit and see whether the dashboards at the other end are earning their keep.
Limitations and Common Problems
Beyond formats and size, the S3 connector behaves differently depending on which Tableau product you are in — which is unusual, and is the source of most "it worked on my laptop" tickets.
Unions are the clearest example. "Only Tableau Desktop supports a wildcard union," while "Web Authoring only supports user-defined manual union (dragging files)." So a wildcard union built in Desktop is not something a colleague can reproduce in the browser.
Wildcard unions also have a prerequisite that reads like a riddle until you hit it. To union files sitting in subfolders, "the root folder or bucket must have at least one file," because "This file is the first file that you connect to when creating the union." An empty top-level prefix breaks the pattern.
Two more product-level gaps are worth noting:
- "Note: This connector isn't currently supported in Tableau Prep Web Authoring or virtual connections."
- "Wildcard union doesn't support Excel files."
Region placement affects both speed and your AWS bill. Tableau's own guidance is that "Same region access has better performance than cross region bucket support," and warns that "Cross region buckets may have additional costs due to data egress fees." Running Tableau Server on EC2 in the bucket's region, or using Tableau Cloud, is the documented way to improve this.
AWS China regions are a special case with a split answer. Tableau says connecting there "is possible, with these limitations": Excel files work with no restrictions across Desktop, Server and Cloud, but "CSV and parquet files aren't accessible in Tableau Desktop and Tableau Cloud," and only Tableau Server can be configured to reach them via a TSM command.
What Tableau does not tell you is as notable as what it does. The S3 help page gives no minimum Tableau version. No vendor page gives a retirement date for the deprecated connector. There is no documentation at all for the new beta connector — no setup page, no auth model, no limitations. The help page never states whether connections are live or extract-only. And there is no documented prefix or path syntax for the connector itself. Every blog that supplies those details is supplying them from somewhere other than Tableau.
The introducing version is contradicted across three Tableau and Salesforce pages, which say 23.1, 2023.2 and 2024.2 respectively. If you need a minimum version, test it rather than trusting any one page.
Tableau's S3 Connector vs Athena, Redshift and Presto
Tableau documents several ways to reach data that physically lives in S3. They are not interchangeable.
- Amazon Athena is the direct substitute and the one Tableau ties to S3 explicitly. It queries files in place, handles partitions and scale, and needs an S3 staging directory for query results. Cost shifts from "free to store" to "pay per query scanned."
- Amazon Redshift is for data you have loaded into a warehouse. Tableau's page covers connecting to "an Amazon Redshift database" and never mentions Redshift Spectrum, so do not assume Tableau documents the Spectrum path.
- Amazon EMR Hadoop Hive is documented, but notably Tableau's EMR page does not mention S3 at all. It also carries the footnote "Not available in Tableau for Apple Silicon."
- Presto and Trino are documented on one page, with Tableau noting that Trino users should "download the appropriate driver from the Trino page." Starburst appears as a partner listing from Starburst rather than a Tableau help page.
- The Web Data Connector is gone. Tableau's page says "The Web Data Connector 2.0 has been retired and is no longer available." Older S3 tutorials that route through a WDC are dead ends, and Tableau's 2026.1 notes say the new S3 connector "eliminates dependency on the deprecated Web Data Connector."
For most teams the honest decision tree is short. Flat files under 15 GB, and keys you are allowed to use: the connector. Anything larger, partitioned, or governed by a no-static-keys policy: Athena. Already warehoused: Redshift.
About the replacement, Tableau's 2026.1 feature notes say "New Amazon S3 Connector is available in Beta," describing it as letting customers "access Amazon S3 data files (CSV and Parquet)" without writing scripts, and claiming it "improves performance and security, enables self-service analytics directly from S3 storage." Note that only two formats are named — Excel and .gz are not mentioned for the beta — and that Tableau does not say which products the beta covers.
Before you rebuild a connector, make sure the reporting on top of it is worth rebuilding. Get a free audit and find the gaps in your analytics and search performance.
Final Thoughts
Can Tableau connect to S3? Yes, through an Exchange-installed connector that reads Parquet, CSV, .gz and Excel files directly out of a bucket, with a 15 GB ceiling and access-key authentication.
The part worth carrying away is the state of that connector. Tableau's Exchange listing calls it deprecated for security concerns and points to a beta replacement it says will reach general availability in 26.3, while Tableau's own help page for the connector mentions none of this. If you are starting a new S3 project today, that gap is the single most useful thing to know, and it is the reason Athena deserves a serious look even when your files would technically fit.
Nothing here requires you to abandon S3 as a source. It does mean deciding deliberately between reading files and querying them, rather than letting the first tutorial you find decide for you.
If you are weighing up how cloud storage fits a BI tool more generally, our guide to Tableau Cloud covers how the hosted version changes what connectors and authoring options are available to you.
Want this done for you?
SEO Stuff gets businesses cited by ChatGPT, Gemini, Perplexity and Claude, and ranking in Google. Start with the free audit, a call, or the package.
See where you stand
A free, manual SEO + AI search audit of your site, with a report within 48 hours. No credit card, no sales call.
Get my free auditTalk it through
A free 15-minute call with the founder about whether AI search is a meaningful opportunity for your business.
Free strategy callHave it done
The Done-For-You Package: audit, 10 pages of content, 3 DR50+ placements, dashboard. $999 once, delivered in 21 business days.
See the package

