Home › AWS › AWS data services › Quick Sight
AWS data services · BI and dashboards · about 11 min read
AWS Quick Sight (formerly Amazon QuickSight), part of AWS Quick
AWS's managed dashboard and BI tool, now the Quick Sight part of AWS Quick, with an in-memory engine called SPICE.
In plain words
What it is
Quick Sight is the business intelligence tool that AWS runs for you. Business intelligence, or BI, means turning tables of numbers into charts, tables and filters that people can read. You connect a data source, build a dataset, draw visuals on an analysis, and publish it as a dashboard. Readers open the dashboard in a browser or inside another app. You never run a server. AWS bills per user, and for some features by capacity.
The product was called Amazon QuickSight from 2016. In October 2025 AWS folded it into a wider product called Amazon Quick Suite. Quick Suite adds chat agents, research reports, automated flows and small apps on top of the BI part. The documentation now calls the whole product AWS Quick and the BI part AWS Quick Sight. AWS says every QuickSight API, SDK and integration keeps working, so old scripts and embedded dashboards do not break. Many people still say QuickSight, and this page uses both names.
A concrete example: a tyre maker wants a daily dashboard of scrap rate per plant, per line and per shift. Quick Sight connects to a gold table called fct_scrap_daily. That table could sit in Athena, Redshift or a Databricks SQL warehouse. The plant manager filters to her own plant. A row-level security rule makes sure she sees only her plant. The quality director sees all plants and gets an email snapshot every morning at seven.
Quick Sight has its own in-memory engine called SPICE. SPICE stands for Super-fast, Parallel, In-memory Calculation Engine. You can import a dataset into SPICE so charts load quickly and the source database is not hit on every click. You can also skip SPICE and query the source live, which is called direct query. Choosing between the two is the most important design choice you make in Quick Sight.
A shop window that AWS builds and cleans for you: you choose what goes on display, but the stock still lives in the warehouse behind it.
Big picture
Where it fits
The lit lane is where Quick Sight works. Boxes on the left feed it; boxes on the right use what it produces.
Quick Sight sits in the Use lane at the far right of the picture. It reads finished numbers from the Serve lane. On our reference platform those numbers are gold tables and metric views in Unity Catalog, served by a Databricks SQL warehouse. Quick Sight has a native Databricks connector, so a report can query the warehouse directly or import into SPICE on a schedule.
The site's reference design names Power BI or Databricks AI/BI as the main BI tool. Quick Sight is an alternative. The migration plan says: keep existing QuickSight reports, repoint each one to the SQL warehouse so it reads gold or a metric view, and build new reports on AI/BI. That way there is one definition of each metric. The BI tool only draws it.
Quick Sight is a good first choice on an AWS-only platform, where Athena or Redshift serve the data and there is no Databricks. It is also good when you must embed dashboards in a customer-facing portal and want per-session pricing. On the AWS plus Databricks platform it is a tool you keep if people already use it, not the place to grow.
How it works
The ideas you need
| Idea | What it means |
|---|---|
| Data source | A saved connection to a database or file store, such as Athena, Redshift, S3 or a Databricks SQL warehouse. |
| Dataset | A prepared query over a data source, with joins, calculated fields and column names that visuals use. |
| SPICE | Quick Sight's in-memory engine. Data is imported and refreshed on a schedule, so dashboards are fast and the source is not hit on each click. |
| Direct query | Visuals send SQL to the source every time someone opens or filters a dashboard. The data is fresh, and the source pays for the load. |
| Analysis and dashboard | An analysis is the editing canvas an author works on; a dashboard is the read-only, published version that readers see. |
| Row-level security | A rule table that limits which rows each user or group sees, for example one plant manager sees only her plant. |
| Author and reader | The two main user roles. Authors build datasets and dashboards; readers only view and filter. They are priced differently. |
| Topics and chat agents | Natural-language question answering over curated datasets. It was branded Amazon Q in QuickSight and now sits within Quick's chat agents. |
| Embedding | Showing a dashboard inside another web app through an API call that returns a signed URL for the session. |
Choosing
When to use it, and when not
Use it when
- Your platform is AWS-only, data is served by Athena or Redshift, and you want a managed BI tool without another vendor contract.
- You already have QuickSight dashboards that people use every day, and moving them now would cost more than it gives.
- You must embed dashboards in a portal for dealers, suppliers or customers, and you want capacity or session pricing instead of named users.
- Your sign-in is IAM Identity Center and you want the BI tool inside the same AWS account boundary and network controls.
- You need many light readers who only look at a few dashboards a month, where per-reader pricing is low.
Avoid it when
- The company standard is Power BI, with Microsoft 365 licences and trained report authors, because a second BI tool splits skills and metric definitions.
- You want the semantic model and the dashboards to live next to the data in Unity Catalog, which is what Databricks AI/BI and metric views give you.
- You need pixel-perfect financial statements with complex layouts, where a paginated report tool fits better.
- Readers need very fast refreshes of live data, every second, for a control room; a live board over a fast store fits better there.
Production grade
What goes wrong in production
| Problem | What breaks | What to do |
|---|---|---|
| SPICE goes stale | A refresh fails at night after the gold table changes shape, and the dashboard shows yesterday's numbers with no warning to readers. | Alert on failed SPICE ingestions with CloudWatch or EventBridge, show a last-refreshed time on every dashboard, and schedule refreshes after the gold job, not at a fixed clock time. |
| Metric drift | A calculated field in Quick Sight defines scrap rate differently from the gold table and from Power BI, so two screens disagree in a meeting. | Define each metric once in gold or a Unity Catalog metric view, and allow only display formatting in the BI tool. Review calculated fields on every publish. |
| Direct query load and timeouts | Every filter click sends SQL to the warehouse. Visuals time out after two minutes, and some drivers keep the query running on the source anyway. | Use SPICE for wide audiences, pre-aggregate in gold, set warehouse auto-stop and query limits, and cancel runaway queries at the source. |
| Row-level security gaps | A new plant is added to the data but not to the rule table, so its manager sees nothing, or a broad rule shows everyone everything. | Generate the rule table from the same entitlement data that drives Unity Catalog grants, and test it with a reader account after every change. |
| Two security models | Unity Catalog masks a column, but a SPICE import made by an author with wide rights copies the unmasked data into Quick Sight. | Connect through a service principal that has only the grants readers should have, and point datasets at views that already apply masks. |
| Licence sprawl | Every person who once clicked a link becomes a paid user, and author seats pile up for people who never author. | Review users each quarter, move inactive authors to reader, and remove leavers through the identity provider, not by hand. |
| Name confusion | Teams search for QuickSight docs and land on Quick Suite agent features, or think the BI product was retired. | Tell teams that the BI part is now called Quick Sight inside AWS Quick, and that dashboards and APIs are unchanged. |
Alternatives
What to compare it with
| Alternative | Pick the alternative when | Pick Quick Sight when |
|---|---|---|
| Power BI | The company already uses Microsoft 365 and has Power BI skills, and you want Direct Lake or DirectQuery over the Databricks SQL warehouse. | You want everything inside the AWS account boundary and billed through AWS. |
| Databricks AI/BI dashboards and Genie | The data, metric views and grants live in Unity Catalog and you want dashboards that use the same permissions and lineage, with no extra licence for Databricks users. | Most readers are outside the Databricks workspace and you need embedding at scale for external users. |
| Looker or Looker Studio (Google Cloud) | You are on Google Cloud, or you want a strong modelling layer (LookML) shared by many tools. | You are on AWS and want one bill and IAM-based sign-in. |
| Amazon Managed Grafana | Engineers need time-series operations boards over CloudWatch, Prometheus or a live store. | Business users need sales, scrap or margin dashboards with filters and email snapshots. |
| Apache Superset or Metabase | You want open-source BI you can run yourself and change freely. | You do not want to run, patch and secure a BI server. |
Cost
How you pay
Quick Sight is billed per user per month, with author and reader roles priced differently. Quick also has newer plans, such as Professional and Enterprise, which include the agent features and charge an account-level infrastructure fee. Reader pricing and capacity pricing for embedded or anonymous viewers are also offered. SPICE capacity beyond what each user includes is charged per GB per month. Exact prices change and differ by plan and region, so check the Quick pricing page before you size anything.
What drives the bill
- Number of authors, who cost several times more than readers.
- Number of active readers each month, or sessions if you use capacity pricing for embedding.
- SPICE capacity above the included amount, which grows with wide, unaggregated datasets.
- Agent and research features in the Quick plans, if you turn them on.
- Load on the source warehouse from direct query, which shows up on the Databricks or Redshift bill, not the Quick bill.
Cost traps
- Direct query dashboards opened by hundreds of people keep the SQL warehouse awake all day.
- Importing raw silver tables into SPICE instead of small gold aggregates uses capacity fast.
- The account-level infrastructure fee in the newer Quick plans is charged even if few people use the agent features.
- Inactive author seats stay billed until someone removes them.
Prices change; check the AWS pricing page for your region before deciding.
Security and governance
Keeping it safe
- Sign in through IAM Identity Center or SAML federation from the company identity provider, and manage groups there, not inside Quick Sight.
- Use row-level and column-level security in datasets, and generate the rules from the same entitlement source as Unity Catalog grants.
- Reach the Databricks SQL warehouse and other sources over a VPC connection so traffic does not cross the public internet.
- Encrypt SPICE with a customer managed KMS key where the plan supports it, and treat SPICE as a copy of the data that must follow retention and erasure rules.
- Record Quick Sight API calls in CloudTrail and include dashboard access in the quarterly access review.
- Personal data should already be tokenised or masked in gold; never import personal columns into SPICE for convenience.
Operating it
What to watch, and the limits
| Watch | Why |
|---|---|
| SPICE ingestion status | Failed or slow dataset refreshes. Alarm on any failure for a tier-1 dashboard. |
| IngestionLatency and IngestionRowCount | CloudWatch metrics for how long refreshes take and how many rows land; a sudden drop in rows signals a broken source. |
| DashboardViewCount and VisualLoadTime | CloudWatch metrics for use and speed of each dashboard; slow visuals point to direct query problems. |
| SPICE capacity used | Capacity consumed against what you pay for, per namespace. |
| Active users per month | From the user list and usage reports; drives licence clean-up. |
| Warehouse query load from Quick Sight | Query history in the Databricks SQL warehouse filtered to the Quick Sight service principal. |
| Limit | Value and notes |
|---|---|
| SPICE dataset size (Enterprise) | Up to 2 billion rows or 2 TB per dataset |
| SPICE dataset size (Standard) | Up to 25 million rows or 25 GB per dataset |
| Columns | 2,000 columns per file or direct query result |
| Direct query visual timeout | 2 minutes to generate a visual |
| S3 manifest | Up to 1,000 files per manifest |
Limits and quotas checked 2026-11-11; many can be raised through Service Quotas.
Decisions
Situations and verdicts
| Situation | Choice | Verdict |
|---|---|---|
| Forty QuickSight dashboards used daily by plant managers | Repoint each to the Databricks SQL warehouse over gold or a metric view, then retire the old source | Keep if you have it: moving them all to a new tool now adds risk and retraining with little gain. |
| New sales and margin dashboards for the commercial team | Power BI or Databricks AI/BI over metric views | Pursue: one metric definition in Unity Catalog, and the company BI standard. |
| Dealer portal showing each dealer its own orders and returns | Quick Sight embedding with row-level security, or AI/BI embedding, compared on per-session price | Pursue with care: embedding is a real strength, but the row rules must come from the same entitlements as the catalog. |
| An AWS-only platform with Athena on S3 and no Databricks | Quick Sight with SPICE over Athena | Pursue: native, managed and billed in the same account. |
FAQ
Questions people ask
Has Amazon QuickSight been retired?
No. In October 2025 it became part of Amazon Quick Suite, and the documentation now calls the product AWS Quick and the BI part AWS Quick Sight. AWS says existing QuickSight APIs, SDKs and dashboards keep working without changes.
Can QuickSight read Databricks tables?
Yes. Quick Sight has a Databricks connector that queries a SQL warehouse. Point datasets at gold tables or metric views so that Unity Catalog grants and masks apply, and use a service principal with only the rights readers should have.
SPICE or direct query: which should I use?
Use SPICE for dashboards with many readers and data that changes a few times a day, because it is fast and spares the warehouse. Use direct query when data must be fresh to the minute and the audience is small.
What happened to Amazon Q in QuickSight?
The natural-language question feature now sits within Quick's chat agents and topics. Separately, the standalone Amazon Q Business service is closed to new customers since mid-2026, which is a different product.
Should we move from QuickSight to Power BI?
Only when there is a reason, such as a company BI standard or a new semantic model. First repoint the existing dashboards to gold and metric views so the numbers match; then move them when they need rework anyway.
Sources
Checked against
- AWS docs: What is AWS Quick
- AWS docs: Quick Sight data source quotas
- AWS blog: Amazon QuickSight evolves to Amazon Quick Suite
- AWS: Quick pricing
- AWS docs: Services in maintenance (Amazon Q Business)
Checked 2026-11-11.
Related
Where it connects
Power BI vs Databricks AI/BI · AWS and Databricks reference platform · Amazon Athena · Amazon Redshift · Amazon Managed Grafana · All AWS data services
← Managed Grafana · Bedrock →