Introduction
Amazon Quick Sight Topics are collections of one or more datasets that represent a subject area such as Sales, Media, or Marketing. They enable business users to ask questions using natural language, eliminating the need for technical expertise in SQL or data modeling. By properly configuring Topics, organizations can democratize data access while maintaining accuracy and relevance in query responses.
Creating successful Topics requires addressing key challenges that span technical configuration and business alignment. Domain knowledge forms the foundation, requiring deep understanding of industry-specific context and data relationships that underpin your organization’s data ecosystem. Business terminology becomes critical as you map technical data fields to the language users naturally speak, ensuring accessibility across different roles and departments. Complex calculations must accurately reflect organizational performance measures, translating raw data into meaningful business metrics. Synonyms capture the multiple ways users refer to the same concepts, enabling flexible natural language interaction that feels intuitive rather than constrained. Custom instructions tailor the Topic’s behavior to specific business requirements, aligning responses with organizational standards and ensuring consistency across user interactions.
This guide outlines best practices for building and maintaining Quick Sight Topics in Quick that deliver exceptional user experiences and enable data-driven decision-making throughout your organization.
Create Quick Sight Dataset
Datasets are the foundation of your Quick Sight analytics, serving as the prepared and structured data sources that power your analyses and dashboards.
Once you’ve created datasets from your data sources, you need to manage them effectively throughout their lifecycle to ensure reliable, secure, and collaborative analytics.
To create a new dataset
1. On the Quick homepage, choose Datasets under Quick Sight section. You can then create a dataset based on an existing dataset or data source, or connect to a new data source and base the dataset on that.
Here is an example of creating dataset by uploading a file (Saas-Sales.csv).
Click “Edit settings and prepare data" to review and adjust field types, rename columns and configure data preparation settings for optimal results. Select save and publish for dataset creation.
Select save and publish for dataset creation.
Navigate to the dataset section, select the newly created dataset, and verify that all rows have been successfully imported.
Optimize Quick Sight Dataset size
When creating a Quick Sight dataset, select only the fields that are necessary for anticipated analysis and reporting needs, avoiding “just in case” field additions that bloat the dataset without clear use cases. This selective approach delivers multiple benefits: it reduces dataset size and storage costs, improves topic tuning efficiency by focusing on relevant fields, enhances query performance through streamlined data structures, and enables faster refresh times. Ultimately, strategic field selection translates directly into faster insights, lower operational costs, and a better user experience for your business users. Leverage SPICE (Super-fast, Parallel, In-memory Calculation Engine) to pre-load data into optimized columnar storage for near-instantaneous query responses.
Create Quick Sight Topic
Before creating a Topic, identify the key business questions your users need answer. Understanding these requirements will guide your decisions about which datasets to select and which fields to expose, and how to configure the Topic for optimal user experience. This upfront planning ensures that your Topic directly addresses real business needs rather than simply exposing available data.
To create a topic
1. On the Quick homepage, choose Topics under Quick Sight section.
2. On the Topics page that opens, choose Create Topic at upper right.
3. On the Create Topic page that opens, do the following:
a. For Topic name, enter a descriptive name for the topic.
b. For Description, enter a description for the topic.
4. Click Continue and on the Select a dataset page
To add one or more datasets that you own or have permission to, select the dataset or datasets that you want to add.
5. Click Create
Your topic is created and the page for that topic opens. The next step is to configure the topic metadata to make it natural-language-friendly for your users. For more information, see Making Quick Sight topics natural-language-friendly.
Here is example Topic to get insights from Product Sales.
Exclude the unused fields
When you add a dataset to a topic, all columns (fields) in the dataset are added by default. If your dataset contains fields that you or your users don’t use, or that you don’t want to include in answers, you can exclude them from the topic. Excluding these fields removes them from answers and the index and improves the accuracy of answers that your users receive. In the Fields section, under Include, toggle the icon off.
Here is an example to exclude City and Customer ID from the topic.
Configure Date fields
Specify default date when multiple date columns exist, such as order date versus shipped date, to ensure queries return the intended results. Set the time basis to define the lowest level of time granularity supported, with options including daily, weekly, monthly, quarterly, or yearly aggregation. This configuration helps aggregate metrics across different time dimensions and is particularly critical for denormalized datasets with many metrics, where clear date field instructions prevent ambiguity and improve query accuracy. Here is an example to set Order Date as the default date.
Assign Field Roles
Every field in your dataset is either a dimension or a measure. Dimensions are categorical data, and measures are quantitative data. Knowing whether a field is a dimension or a measure determines what operations can and can’t perform on a field. For Example, Customer falls into the dimension category because it’s a categorical attribute that describes a person.
Specify Field Semantic Types
Always assign the most specific semantic type and subtype available to your fields. This significantly improves the accuracy of natural language query interpretation and ensures consistent, professional data presentation across all user interactions. Semantic types help the system understand the nature of your data, whether it represents locations, dates, currencies, or other specialized data types. An example for the Region field, Semantic type is Location and sub type can be Country, state, city.
Set Field Aggregations
Setting field aggregations helps determine which function should or shouldn’t be used when those fields are aggregated across multiple rows. You can set a default aggregation for a field, and a not allowed aggregation.
A default aggregation is the aggregation that’s applied when there’s no explicit aggregation function mentioned or identified in a user’s question. For example, let’s say one of your users asks, “How many products were sold yesterday?” In this case, Topics uses the field Product ID, which has a default aggregation of count distinct, to answer the question. Doing this results in a visual showing the distinct count of Product ID.
Not allowed aggregations are aggregations that are excluded from being used on a field to answer a question. They’re excluded even if the question specifically asks for a not allowed aggregation. For example, let’s say you specify that the Product ID field should never be aggregated by sum. Even if one of your users asks, “How many total products were sold yesterday?” sum isn’t used to answer the question.
If aggregate functions are incorrectly applied on a field, we recommend that you set not allowed aggregations for the field.
Format Field Values
Define formatting for values such as U.S. dollars or percentages to ensure consistent presentation in answers. For example, format “Sales” as U.S. currency to maintain professional and standardized data display across all responses.
Add Synonyms to Fields and Field Values
Synonyms
Users use different terms for the same field, such as “Sales” versus “revenue” or “orders” Adding synonyms to fields and field values improves accuracy by mapping user questions to the correct data elements, ensuring users can ask questions using their preferred terminology while retrieving accurate results. Consider your industry-specific terminology, common abbreviations, and regional variations when defining synonyms.
Here is an example of adding synonyms revenue, orders for field sales
Field value synonyms
Field value synonyms enhance user experience by making data exploration intuitive and user-friendly, allowing users to query using familiar terms rather than exact data values. For example, if your dataset contains value “United States” in the Country column, adding synonyms like "america”,”usa”,”us” enables users to query using any common variation and receive accurate results immediately, eliminating query failures and creating a more natural analytical experience.
Here is an example of configuring field value synonyms for value United States.
Calculations
Create calculated fields to implement standardized business metrics, combine multiple fields, apply business rules, and simplify user queries. Use descriptive, business-friendly names with units specified (e.g., “Customer Segment Classification”). Common patterns include revenue calculations (quantity × price), customer metrics (lifetime value, segmentation), date calculations (days between dates, period extraction), and performance metrics (growth rates, conversion rates).
Here is an example for creating a calculated field.
Custom Instructions
Custom Instructions enhance the Topic responses by adding domain-specific knowledge that can’t be captured through standard topic metadata, such as synonyms or semantic types. This ensures more accurate, relevant, and tailored answers aligned with business needs. For example, you might add an instruction like " Define Q1 as August through October, Q2 as November through January, Q3 as February through April, and Q4 as May through July. When users ask about quarterly sales, always use these fiscal quarters instead of calendar quarters." to ensure quarterly sales queries return complete information.
Create a Space
A space in Quick is a collection of data and Quick resources scoped for a particular team or domain. You can use spaces to aggregate topics into a customizable knowledge center for your team. Spaces integrate seamlessly with Quick agents for contextual conversations and are designed to scale across personal, team, and cross-team use cases.
To create a space
1. Log in to the Quick console.
2. From the left navigation menu, select Spaces, and then select Create space.
3. On the space creation page that opens, do the following:
a. (Optional) Enter a name for your space.
b. (Optional) Enter a description for your space.
4. After your space is created, you can start interacting with your space using any available agent.
Add topics to space
In the space, add knowledge to begin adding Topics to your space. From the dropdown menu, you can select Topics from available resource types. After you finish adding topics, Space knowledge displays a list of all topics added to your space.
Create Custom Chat Agent
Chat agents in Quick help users explore data, analyze information, and take actions. Users can interact with chat agents using the Quick chat interface. You can use chat agents to Generate content and provide answers through natural language conversations from connected quick sight topics.
In this section, you create a custom chat agent in Quick . Complete the following steps:
1. On the Quick console, choose Chat agents in the navigation pane.
2. Choose Create chat agent.
3. On the Agent creator page, chose Skip.
4. Add a name for your custom chat agent. This is the name that your chat agent will be identified by.
5. Add an optional description for your custom chat agent that helps users understand the purpose of the chat agent.
6. On the Configure chat agent page, provide the following information:
a. For Agent identity, define the identity of your chat agent. For example: You are a specialized SaaS Sales Intelligence Assistant with deep expertise in analyzing B2B SaaS transaction data and business performance metrics. Your primary role is to help users understand sales trends, customer behavior, product performance, and revenue optimization opportunities within our SaaS business. When responding to queries, you should analyze data across multiple dimensions including temporal trends, customer segments, geographic markets, industry verticals, and product performance.
b. For Persona instructions, enter instructions on how your chat agent interacts with users during chat. For example: your primary role is to help users understand sales trends, customer behavior, product performance, and revenue optimization opportunities within our SaaS business. When responding to queries, you should analyze data across multiple dimensions including temporal trends, customer segments, geographic markets, industry verticals, and product performance. When presenting analysis, prioritize metrics that matter most to sales and revenue operations: revenue trends, profit margins, customer acquisition patterns, product mix optimization, and segment performance. Always ground your responses in the actual data available in the SaaS Sales topic. When users ask about specific customers, products, time periods, or segments, provide precise metrics and contextual analysis.
c. Under KNOWLEDGE SOURCES choose Link SPACES, choose the space Saas Sales you created, then choose Link.
d. Choose Launch chat agent to create your custom chat agent. You will see the progress Launching chat agent… and in a few minutes, you will see Successfully launched chat agent.
Chat with the agent
In the Quick Console, select the Chat agent from the chat agent’s navigation pane and begin querying the agent. Here is an example of how to get the total sales and margins for the last quarter. The Chat agent searches through the attached topic and retrieves the relevant data, as shown in the following screenshots.
Share the Agent, Space, and Topic with business users
When you create a chat agent, you are assigned as its owner by default. You can choose to share a chat agent you own with other Quick users. However, sharing the agent doesn’t grant automatic access to linked resources (e.g., spaces, topics). To enable full functionality, share the associated spaces and topics with the user as well.
Monitoring & Maintenance
Effective Topic management requires continuous refinement based on user feedback and evolving business needs. Regularly enhance Topic configuration by adding synonyms for terminology gaps, adjusting custom instructions for recurring issues, and expanding calculated fields for emerging analytical requirements. Maintain documentation of configuration changes to track Topic evolution and inform future optimization decisions. Optimize data architecture through pre-aggregation of metrics at source, partitioning data by common query patterns, and scheduling refreshes based on actual data update frequency. Monitor key performance indicators including answerable rate, query performance, refresh success, and SPICE capacity utilization to identify optimization opportunities and manage costs effectively.
A critical component of Topic optimization is understanding and managing SPICE storage capacity. Understanding SPICE storage capacity limits is essential for effective capacity planning and budget management. Quick Suite Enterprise edition provides 1 TB (or 2 billion rows) of SPICE capacity per dataset, with recent enhancements supporting up to 2 TB per dataset when using the new data preparation experience, while Standard edition offers 25 GB (or 25 million rows) per dataset. Beyond dataset limits, every Quick Sight Author license includes a 10 GB SPICE allocation, and every Quick Sight Author Pro license includes a 10 GB SPICE credit as part of the base subscription, providing immediate analytical capabilities without additional procurement across both Standard and Enterprise editions. Additional SPICE capacity can be purchased as needed and can be added or removed dynamically at any time during the month, with associated costs that scale based on usage. When monitoring SPICE utilization, consider these capacity thresholds to proactively manage storage allocation, optimize dataset sizes through aggregation strategies, and forecast infrastructure costs as your analytical requirements grow.
Dataset /Topic Refresh
Configure dataset refresh schedules based on data update frequency and business requirements. Align refresh timing with source data processing schedules while balancing data freshness needs against computational costs and SPICE capacity consumption. Operational datasets may require frequent refreshes during business hours, strategic datasets typically benefit from daily off-peak refreshes, and reference datasets often need only weekly updates. Monitor refresh performance and adjust schedules as requirements evolve to maintain efficiency and cost effectiveness.
Topic refreshes ensure your analytical environment reflects current data definitions and values. Refresh topic by refreshing datasets within the topic—manually refresh all datasets at once, refresh individual datasets as needed, or establish recurring schedules. Dataset refresh history enables performance monitoring and troubleshooting. For SPICE datasets, synchronize topic index refresh schedules with SPICE refresh schedules to maintain consistency. Regular topic index updates ensure the latest field definitions, calculated metrics, and data values are accurately recorded and available for analysis. Here is an example to configure dataset and Topic refresh.
Conclusion
Quick Sight Topics transform how business users interact with data by enabling natural language queries without requiring SQL expertise or technical knowledge. Success depends on thoughtful configuration—optimizing dataset size, enriching metadata with synonyms and semantic types, setting appropriate field roles and aggregations, and adding custom instructions that reflect your business context. When combined with Spaces for resource organization and custom chat agents with tailored personas, Topics create comprehensive knowledge environments that empower self-service analytics across your organization.
Ongoing monitoring and maintenance ensure long-term success. Regular review of answerable rates, query performance, and user feedback enables continuous refinement while optimizing costs through strategic field selection and refresh frequency management. By following these best practices, you establish a foundation for data-driven decision making that reduces the burden on technical teams and accelerates insights discovery across your business.
References
Semantic Types https://docs.aws.amazon.com/quicksuite/latest/userguide/topics-natural-language.html
Authors
Satyanarayana Adimula is a Senior Builder in AWS Generative AI Innovation & Delivery.Leveraging over 20 years of data and analytics expertise, he specializes in building agentic AI systems that enable large enterprises to automate complex workflows, accelerate decision-making, and achieve measurable business outcomes.
Munesh Siddappa is a Senior Builder at AWS Generative AI Innovation delivery, focusing on Amazon Quick implementations for enterprise clients across industries. He brings over a decade of experience in enterprise database solutions and currently specializes in generative AI products, working with customers and partners to deploy Amazon Quick solutions on AWS Cloud.