Choosing a Destination

When Stitch replicates your data, we’ll load it into the destination of your choosing.

As Stitch currently only allows you to connect one destination to your account, we recommend asking yourself the questions below before making your selection. By fully assessing each choice first, you’ll decrease the likelihood of needing to switch destinations or re-replicate all of your data at a later date.

• Does it support all (or most of) your data sources?
• Will the structure of the data replicated by Stitch work with how you plan to use it?
• Do you need a fully-managed solution, or can you perform maintenance tasks on your own?
• Is the destination’s pricing structure suitable?

---

Side-by-side comparison

In the tabs below is a quick look at each of Stitch’s destinations and how they compare to each other.

The remaining sections of this guide expand on the information in these tabs.

• General
• Pricing
• Stitch details
• Destination details
• Replication and transformations

---

Getting started, now

If you simply want to try Stitch and Redshift, or if you don’t have the ability to spin up a Redshift cluster of your own in AWS, we recommend trying Panoply. With just a few clicks, you create your own fully-managed Redshift data warehouse and start replicating data in minutes.

Note: If you decide to switch destinations later, you’ll need to queue a full re-replication of your data to ensure historical data is present in the new destination.

---

Destination and data source compatibility

Some integrations may be partially or fully incompatible with some of the destinations offered by Stitch. For example: Some destinations don’t support storing multiple data types in the same column. If a SaaS integration sends over a column with mixed data types, some destinations may “reject” the data.

For integrations that allow you to control how data is structured, you may be able to fix the problem at the source and successfully replicate the data. If this is not possible, however, Stitch may never be able to successfully replicate the incompatible data.

Refer to the Integration and destination compatibility reference for more info.

---

Replication, transformations, and data structure

While the majority of your data will look the same across our destinations, there are some key differences you should be aware of:

• How updates to existing records are handled

• How destinations handle nested data structures

Loading behavior and updates to existing records

Loading behavior determines how data is loaded into your destination. Specifically, how updates are made to existing rows in the destination.

Stitch supports two loading behavior types:

• Upsert: When data is loaded using the Upsert behavior, existing rows are updated in tables with defined Primary Keys. Stitch will de-dupe records based on Primary Keys before loading them, meaning that only one version of a record will exist in the table at any given time.
• Append-Only: When data is loaded using the Append-Only behavior, records are appended to the end of the table as new rows. Existing rows in the table aren’t updated even if the source has defined Primary Keys. Multiple versions of a row can exist in a table, creating a log of how a record has changed over time.

The table below lists the default loading behavior for each destination and whether it can be configured.

Note: If a destination supports and is configured to use Upsert loading, Stitch will attempt to use Upsert loading before Append-Only. All other conditions for Upsert loading must also be met.

Refer to the Understanding loading behavior guide for more info and examples of each loading bheavior type.

Destination	Version	Default loading behavior	Loading behavior is configurable?
Amazon Redshift	v2	Upsert
Amazon S3	v1	Append-Only
Google BigQuery	v1	Append-Only
Google BigQuery	v2	Selected by you
Microsoft Azure SQL Data Warehouse	v1	Upsert
Panoply	v1
PostgreSQL	v1	Upsert
Snowflake	v1	Upsert
data.world	v1	Upsert

Nested data structures

Some destinations don’t natively support nested structures, meaning that before Stitch can load replicated data, these structures must be “de-nested”. During this process, Stitch will flatten nested structures into relational tables and subtables. As a result of creating subtables, a higher number of rows will be used.

If a destination does natively support nested structures, no de-nesting will occur and Stitch will store the records intact.

Check out the Handling of Nested Data & Row Count Impact guide for an in-depth look at what we mean by nested records, how Stitch handles nested data, and what those records will look like in your data warehouse.

Destination	Version	Support	Notes
Amazon Redshift	v2		Nested data structures will be flattened into relational objects.
Amazon S3	v1		Depending on the data storage format you select, nested data structures may be kept intact or flattened into relational objects: CSV: Nested data structures will be flattened into relational objects. JSON: Nested data will be kept intact.
Google BigQuery	v1		Nested data structures will be maintained.
Google BigQuery	v2		Nested data structures will be maintained.
Microsoft Azure SQL Data Warehouse	v1		Nested data structures will be flattened into relational objects.
Panoply	v1
PostgreSQL	v1		Nested data structures will be flattened into relational objects.
Snowflake	v1		Nested data structures will be loaded intact into VARIANT columns.
data.world	v1

---

Maintenance and support

With the exception of a self-hosted PostgreSQL instance, all the destinations offered by Stitch are cloud-hosted databases, meaning you won’t have to factor in server maintenance when making a decision.

In the table below are:

• Fully-managed destinations that include routine maintenance and upgrades in their plans
• DIY destinations that will require you to perform and schedule maintenance tasks on your own

Fully-managed destinations	DIY destinations
Google BigQuery Google BigQuery Microsoft Azure SQL Data Warehouse Panoply Snowflake data.world PostgreSQL (Heroku, Amazon RDS, Amazon Aurora)	Amazon Redshift Amazon S3 PostgreSQL (self-hosted)

---

Destination pricing models

Every destination offered by Stitch has its own pricing structure. Some providers charge by overall usage, others by an hourly rate or the amount of stored data. Depending on your needs, some pricing structures may fit better into your budget.

In the next section, you’ll find each destination’s pricing structure, including a link to detailed price info and whether a free option (trial or plan) is available. Here are a few things to keep in mind:

Destination	Version	Pricing model	Notes
Amazon Redshift	v2	Hourly	Currently, Amazon Redshift bases their pricing on an hourly rate that varies depending on the type and number of nodes in a cluster. The type and number of nodes you choose when creating a cluster is dependent on your needs and data set, but you can scale up or down over time should your requirements change.
Amazon S3	v1	Storage	Amazon S3 pricing is based on two factors: The amount of data stored in and location (region) of your Amazon S3 bucket.
Google BigQuery	v1	Usage	Google BigQuery ‘s pricing isn’t based on a fixed rate, meaning your bill can vary over time. To learn more about how Stitch may impact your BigQuery costs, click here.
Google BigQuery	v2	Usage	Google BigQuery ‘s pricing isn’t based on a fixed rate, meaning your bill can vary over time. To learn more about how Stitch may impact your BigQuery costs, click here.
Microsoft Azure SQL Data Warehouse	v1	Usage	Microsoft Azure SQL Data Warehouse bases their pricing on your compute and storage usage. Compute usage is charged using an hourly rate, meaning you’ll only be billed for the hours your data warehouse is active. Compute usage is billed in one hour increments. Storage charges include the size of your primary database and seven days of incremental snapshots. Microsoft Azure rounds charges to the nearest terabyte (TB). For example: If the data warehouse is 1.5 TB and you have 100 GB of snapshots, your bill will be for 2 TB of data. Refer to Microsoft’s documentation for more info and examples.
Panoply	v1	Monthly	Panoply charges based on the amount of data stored and offers several plan options for your needs. Refer to their pricing page for more information.
PostgreSQL	v1	Varies	Pricing depends on the type of PostgreSQL instance you’re using. Heroku and Amazon RDS, for example, have a variety of plans to choose from.
Snowflake	v1	Usage	Snowflake pricing is based on two factors: The volume or data stored in your Snowflake destination and the amount of compute usage (the time the server runs) in seconds. Snowflake offers two types of plans, each with varying levels of access and features. There are On Demand plans which are commitment-free and usage-based. The alternative is a Capacity option, which guarantees secure price discounts. Learn more about Snowflake plans and pricing here.
data.world	v1	Monthly	data.world plans vary depending on the number of private projects/data sets, size limits per project/dataset, external integrations, and total number of team members that can belong to an account. All plans, however, include unlimited public projects/datasets, API access, joins, queries, activity alerts, and other standard features.

---

Additional resources and setup tutorials

Ready to pick a destination and get started? Use the links below to check out a more in-depth look at each destination or move on to the setup and connection process.

• Amazon Redshift
  Overview | Setup
  Loading reference
• Amazon S3
  Overview | Setup
  
• Google BigQuery (v1)
  Overview | Setup
  Loading reference
• Google BigQuery (v2)
  Overview | Setup
  Loading reference
• Microsoft Azure SQL Data Warehouse
  Overview | Setup
  
• Panoply
  Overview | Setup
  
• PostgreSQL
  Overview | Setup
  Loading reference
• Snowflake
  Overview | Setup
  Loading reference
• data.world
  Overview | Setup
  

Related	Troubleshooting
Choosing a Destination Destination & Integration Compatibility Loading Data into Your Destination Switching Destinations	Destination Connection Errors

---

Questions? Feedback?

Did this article help? If you have questions or feedback, feel free to submit a pull request with your suggestions, open an issue on GitHub, or reach out to us.