There are over 300 SaaS applications that help you automate your data operations. In this crowd of potential solutions, how do you narrow it down to the winning horse?

Here we compare three top contenders for your data operations: from data ingestion, via ETL to full data management, we compare side-by-side Fivetran, Stitch, and Keboola to shed light on their respective strengths and weaknesses.

Introduction

Fivetran, Stitch, and Keboola are all cloud-based data platforms. They are built around the idea that software can help you automate your data processes at scale. Thus saving your engineers, scientists, and analysts time for more revenue-generating work. 

But the three contenders have very different ranges of capabilities:

• Fivetran and Stitch focus solely on data ingestion and ETL. That is, they help you automatically extract data from Third-Party Apps, load them into databases, data warehouses, or data lakes, and they might help you transform the data.
• Keboola, on the other hand, is an end-to-end data operations platform. By focusing on the entire data operations process - instead of just the first mile - Keboola builds tools to help you extract insights faster. Alongside its extensive ETL capabilities, Keboola can be used for data governance, machine learning prototyping, data pipeline monitoring, data management, and a myriad of other operations.
  

Integrations: Sources and Destination

Data integration is the process in which ETL tools extract raw data from a source, transform the data (e.g., clean it, remove duplicates, etc.), and load it into a target destination, such as a data warehouse (Amazon Redshift or Snowflake for instance). 

The three SaaS platforms offer extensive automation in the extract and load departments. By building automatic connectors between sources and destinations, the SaaS platforms allow you to automate data collection and syncs along your data pipelines. 

Sources can include a plethora of data generators: 

• CRM and sales software: Hubspot, Salesforce, Pipedrive, etc.
• Advertising platforms: Facebook Ads, Google Ads, Linkedin Ads, Twitter Ads, CJ Affiliate, BingAds, and others.
• Marketing communication: CustomerIO, Drip, Mailchimp, ActiveCampaign, Klaviyo, ...
• Project management: Asana, Jira, Emails, Freelo, Google Drive, OneDrive, etc.
• Inhouse data: Google Sheets, CSVs, Excel, Dropbox, MySQL, PostgreSQL, MongoDB, and others.
• Support and customer success software: Delighted, Zendesk, Intercom, Freshdesk, etc.
• Accounting solutions: QuickBooks, Fakturoid, FlexiBee, ...

‍

Destinations are where your data will reside once extracted. For example:

• Databases: MySQL, MongoDB, PostgreSQL, DynamoDB, Firebird, Microsoft SQL Server, Neo4j, Oracle, …
• Data warehouses: Apache Hive, Amazon Redshift, IBM dashDB, Snowflake, Synapse, Google BigQuery …
• Data lakes (for huge deployments)
  

All three platforms allow you to perform extract and load operations in a user-friendly manner. No need for technical knowledge, all you need is a couple of clicks in your browser to automatically connect your sources and destinations and start syncing data.

So how do the three providers differ?

1. Quantity of sources and destinations. Stitch offers 135 sources and 9 destinations, Fivetran integrates with 125 sources and destinations, while Keboola connects with almost 250 sources and destinations. These numbers vary, with new connectors being added every day, so make sure to check the links for new integrations.
2. Origin of integrations. Fivetran does not support community-built integrations. What the Fivetran team builds is what you get. In contrast, both Keboola and Stitch offer rich community-built applications alongside their core integrations.
3. Transparency and documentation coverage. All three providers offer varying levels of documentation for their integrations. While Fivetran’s integrations have how-to guides, Keboola and Stitch build open-source integrations. This level of transparency allows you to dive deeper into understanding what integration does and even alter it if needed. As an example, look at the differences in documentation and access to code for a Google Analytics integration: Fivetran’s implementation, Stitch’s platform, and Keboola’s solution.
4. Custom integrations. If you want to add a new source, you need to request it with Fivetran’s support (and wait for them to build it), while Keboola and Stitch developed an open-source toolkit for building new data sources. With Stitch, you can use Singer, an open-source project, to develop a new integration. While Keboola offers its Generic component to import data from almost any REST API and countless other APIs. The component acts as a customizable HTTP REST client. So you can build your own data source. The universal extractor does not require programming prowess, you can write your extractor as a JSON (check the tutorial for more information).

Transformations

Transformations refer to all the procedures needed to make your data valuable. For example, cleaning your data to exclude corrupted information, adjusting the data types to match source and destination data, removing duplicates, etc.

Both Stitch and Fivetran have been historically lacking in the transformation department, offering no or only post-load transformations.

The Fivetran team used to offer no transformation capabilities at all but has recently partnered with dbt to extend its SQL transformation abilities. It still only focuses on post-load transformations, and it is unclear how using dbt affects pricing.

Stitch data advertises itself as an ETL product but does not really do full data transformations. Instead, it offers the ELT tool paradigm, where you use Stitch to extract and load your data, and then rely on other software (such as Talend Data Fabric) for transformations. 

Keboola on the other hand offers pre- and post-load data transformations. From premade components that do the transformation for you, to fully custom-written transformations in the programming language your engineers and scientists love best (SQL, Python, R, Julia, …). 

When a transformation job runs, it takes the required data from your project Storage and copies them to a temporary staging storage where data manipulation is executed, then the newly cleaned data is written to the destination of your choice. 

Because transformations are run like a job, you can schedule them and automate them, making data cleaning an integral and reliable part of your data operations workflow.

Advanced analytics, data visualizations, and machine learning

Good and clean data on its own is not enough to gain better insights. You also have to analyze it, visualize it, and sometimes productionalize it with machine learning, to extract the business value from your data assets. 

For example, until you analyze historical user data you cannot discover which customers are likely to churn (and need to be contacted ASAP to avoid losing their business), and which ones you can upsell in the next quarter.

All three data platforms integrate with a variety of BI and AI applications, such as Looker, Google Analytics, Jupyter notebooks, etc.

Where Keboola truly shines, and what sets it apart from Fivetran and Stitch, is the ecosystem for analytics, data engineering, and data science built for data operatives. Keboola, unlike the competitors, does not just integrate with other data tools, but it also:

• Offers Sandboxes, which abstract from production code and production data and allow for experimentation. Your data professionals can test and try different data products without worrying about breaking the production environment to iterate faster and gain insights at unprecedented speeds.
• Use the Data Catalog for both documenting data and sharing data. Understanding the data you work with is integral to developing quality analytical products. Being able to share data allows collaboration across people and departments, as well as reproducibility of data experiments when building data-centered products.
• Data scientists and analysts can tap into the application marketplace to build their custom ETL pipelines, and automate them, without having to rely on engineers to do it for them. A data scientist can easily integrate Google’s NLP API into their workflow, or deploy their ML model to Azure as an API with a couple of clicks. No need to rely on the backend engineer to have a free afternoon. This cuts production times and speeds up time to insights. Automating the ETL pipelines via Orchestrators also gives you the ability to work with always fresh data.

Security

All three platforms offer state of the art security provisioning: 

• From HIPAA, SOC 2, and GDPR certificates of regulatory compliance.
• Data encryption at rest and in transit.
• MFA and other forms of access provisioning.
• External security audits and certification renewals.
• Etc.

The only difference is the type of security offered for different plans:

• Fivetran offers safe VPN tunnels for data transfers only on their most expensive plan (if you pay less, you will have to SSH or find another solution).
• Stitch guarantees HIPAA compliance only on their Enterprise plan. 
• Keboola does not think its paying customers are the only ones who need data protection. Your data is secure even on the freemium model.

Support

All three data platforms offer varying methods of support: phone support, in-app chat support, training services (like the Keboola Data Academy), enterprise SLA, etc.

However, there seem to be differences in the quality of support. As can be seen from reviews and ratings of the three data products, users are not equally satisfied with all three platforms:

• 4.8/5 - Keboola
• 4.7/5 - Stitch
• 4.2/5 - Fivetran

A lot of dissatisfaction stems from support complaints.

For example, looking at Fivetran’s customer feedback, it seems their delivery of timely and competent support is inconsistent across the customer base. While some clients praise Fivetran for their responsiveness, others complain about the technical support inefficiencies, delays, and lack of clear and prompt communication. 

Another issue seems to be the necessary reliance on support for solving common problems. Multiple reviewers comment about their inability to debug issues themselves and having to wait on Fivetran’s technical support to solve connector issues on Fivetran’s end. This is probably because Fivetran does not have a transparent codebase, preventing users from looking at integration implementation.

On the other hand, both Keboola and Stitch are praised for their support (e.g. users have consistently rated Keboola’s customer service with 5/5, Stitch reports 94% of customers rate their support as “good” or “great”).

Pricing

All three data platforms are transparent with their pricing (praise the world of SaaS which does not hide this information), but vary widely in how they price their products:

1. Fivetran offers a 14-day free trial then implements a custom calculation of “monthly active rows” (MARs) and associated credits. MAR refers to the unique number of rows that were inserted, deleted, or updated within a calendar month. Their credit-to-row calculations are not straightforward because the price of a credit varies based on the number of rows that are active in a given month and the tier of the plan your organization has paid. Fivetran offers no free plan.
2. Stitch offers a 14-day free trial, then starts at $100/month for 5 million processed rows (aka, all the imported, exported, transformed rows are counted at every touchpoint). Their standard plan does not offer custom sources and limits users to a maximum of 10 standard sources. Stitch data offers no free plan.
3. Keboola offers an always free plan ($0/month), where you can always expand use by purchasing credits directly on the platform, and a subscription plan. Instead of rows, you are charged for the Time of processing. Every account gets 300 free minutes a month (free plan), then pays 14 cents per minute. In case your needs are greater, custom subscription plans are available.

Conclusion: Which one do you pick?

The ultimate choice of which data platform best serves your company’s needs depends on, well, your company’s needs. 

In  this article we compared side-by-side three of the main contenders for your ETL and data operations needs:

‍

We know choosing the best data platform can be hard. We’ve been there ourselves. This is why we hate vendor lock-in and the limited resources you get before committing to a new tool.

This is why we created the always-free Keboola plan. So you can explore everything Keboola has to offer without any commitments. Feel free to give it a go or reach out to us if you have any questions.