About 26; min
Fivetran, Airbyte, and Stitch are the three most popular data pipeline tools — they move data from your SaaS apps, databases, and APIs into your data warehouse. Fivetran pioneered the category with managed pipelines. Airbyte entered as the open-source challenger. Stitch (acquired by Talend, now part of Qlik) offers a middle ground.
The pricing difference is dramatic: Fivetran charges by rows synced (and bills can surprise you). Airbyte is free to self-host. Stitch charges a flat monthly fee. Here’s what each one actually costs.
Quick Pricing Comparison
| Tool | Free Tier | Paid Starting At | Pricing Model |
|---|---|---|---|
| Fivetran | Free plan (500K MAR) | $1/credit (~$460/mo starter) | Monthly Active Rows (MAR) |
| Airbyte | Free (self-hosted, unlimited) | Airbyte Cloud: $10/credit | Self-hosted free / Cloud per credit |
| Stitch | Free plan (5M rows, 10 sources) | $100/mo (Standard) | Rows loaded per month |
Fivetran: The Managed Leader
Fivetran’s pricing revolves around Monthly Active Rows (MAR) — the number of unique rows that were created or updated in your source systems during a billing month. A table with 1 million rows where 50,000 change monthly counts as 50K MAR, not 1M.
Fivetran Cost Reality
Fivetran uses a credit system where different connector types consume different credit amounts per MAR:
| Connector Type | Credits per MAR | Example |
|---|---|---|
| Free connectors | 0 | Google Sheets, webhooks |
| Standard connectors | 1 | Salesforce, HubSpot, Stripe |
| Premium connectors | 2 | SAP, Oracle, Workday |
| Database connectors | 1 | PostgreSQL, MySQL, MongoDB |
At $1/credit on Starter: syncing 500K MAR from Salesforce costs $500/month. Add HubSpot (200K MAR) and you’re at $700/month. Add a PostgreSQL database that changes 1M rows monthly and the bill jumps to $1,700/month.
Real-world Fivetran bills typically range from $500/month (small startup, 3–5 sources) to $5,000–$20,000/month (mid-market, 20+ sources with databases). Enterprise companies with SAP and Oracle connectors pay $50,000+/month.
Fivetran’s Strengths
- Zero maintenance: Fivetran handles connector updates, schema changes, API versioning, and error recovery automatically. When Salesforce changes their API, Fivetran updates the connector — you do nothing
- 300+ connectors: The broadest connector library. If a SaaS tool has an API, Fivetran probably has a connector for it
- Schema migrations: When your source adds a new column, Fivetran propagates it to your warehouse automatically. No manual ALTER TABLE statements
- Reliability: 99.9% SLA on Standard. Fivetran is battle-tested at enterprise scale. When your pipeline breaks at 3 AM, Fivetran’s team fixes it — not yours
Fivetran’s Downsides
- Expensive at scale: MAR-based pricing means costs grow with your data volume. A viral marketing campaign that creates 2M new CRM records in a month can spike your Fivetran bill significantly
- Cost unpredictability: Hard to forecast monthly costs when data volumes fluctuate. Set billing alerts and MAR caps to avoid surprises
- Premium connector pricing: SAP, Oracle, and Workday connectors cost 2x per MAR. Enterprise data sources double your per-row cost
Airbyte: Open-Source Data Pipelines
Airbyte’s self-hosted version is completely free with no row limits, no connector limits, and no feature gates. Deploy it on your own infrastructure and sync unlimited data from 350+ connectors at the cost of server hosting only.
Airbyte Self-Hosted Costs
| Deployment Size | Server Requirement | Monthly Hosting Cost |
|---|---|---|
| Small (5 connectors, 1M rows/mo) | 4 GB RAM, 2 vCPU | $20–$40 |
| Medium (20 connectors, 10M rows/mo) | 8 GB RAM, 4 vCPU | $50–$100 |
| Large (50+ connectors, 100M rows/mo) | 16+ GB RAM, 8 vCPU, Kubernetes | $200–$500 |
Self-hosted Airbyte syncing 10M rows/month from 20 connectors costs ~$50–$100/month in infrastructure. Fivetran for the same workload: $2,000–$5,000/month. That’s a 20–50x cost difference.
Airbyte’s Strengths
- Free self-hosting: Unlimited everything at the cost of infrastructure. The most cost-effective option for teams with DevOps capacity
- 350+ connectors: Comparable to Fivetran’s library. The open-source community contributes new connectors regularly
- Custom Connector Development Kit (CDK): Build connectors for internal APIs and niche tools using Python or the low-code YAML builder. Faster to build custom connectors than on Fivetran
- No vendor lock-in: Open-source means you own your pipeline infrastructure. Migrate between self-hosted and cloud without losing configurations
- Airbyte Cloud: If you don’t want to self-host, Cloud provides the same connectors managed by Airbyte. Credit-based pricing is more predictable than Fivetran’s MAR model
Airbyte’s Downsides
- Self-hosting burden: You manage upgrades, monitoring, error handling, and scaling. When a connector breaks after an API change, you wait for a community fix or fix it yourself
- Connector quality varies: Core connectors (Salesforce, PostgreSQL, Stripe) are rock-solid. Community connectors for niche tools can be buggy or outdated
- No automatic schema migration: When source schemas change, Airbyte can handle it but the experience isn’t as polished as Fivetran’s automatic propagation
- Cloud pricing: Airbyte Cloud at $10/credit is comparable to Fivetran for managed pipelines. The cost advantage disappears unless you self-host
Stitch: Simple Flat Pricing
Stitch’s flat pricing is refreshing: $100/month for 100 million rows. No credits, no MAR calculations, no connector tiers. You know exactly what you’ll pay every month. At 100M rows, Stitch costs $100 while Fivetran could cost $5,000+ for the same volume.
Stitch’s Strengths
- Predictable pricing: Flat monthly fee based on row volume. No surprises, no per-connector multipliers, no credit calculations
- Free tier: 5M rows/month from 10 sources at $0. Enough for a startup syncing CRM, marketing, and product data to a warehouse
- Simple setup: Point-and-click connector configuration. Less powerful than Fivetran’s transformation layer but faster to set up for basic syncs
Stitch’s Downsides
- Fewer connectors: ~130 connectors vs Fivetran’s 300+ and Airbyte’s 350+. Niche SaaS tools may not have Stitch connectors
- Qlik ownership uncertainty: Stitch was acquired by Talend (now Qlik). Product investment has slowed compared to Fivetran and Airbyte’s rapid development
- Limited transformations: No built-in dbt integration or transformation layer. You handle transformations in your warehouse separately
- Row-based counting: Unlike Fivetran’s MAR (only changed rows), Stitch counts total rows loaded — including unchanged rows in full-table replications. This inflates row counts for large tables
Cost Comparison
Startup (5 SaaS sources, 1M changed rows/month)
| Tool | Monthly Cost |
|---|---|
| Stitch Free | $0 (within 5M limit) |
| Airbyte self-hosted | $20–$40 (VPS) |
| Fivetran Free | $0 (within 500K MAR — may need Starter at $460) |
| Airbyte Cloud | ~$100–$200 |
Growth company (20 sources, 10M changed rows/month)
| Tool | Monthly Cost |
|---|---|
| Airbyte self-hosted | $50–$100 |
| Stitch Standard | $100 |
| Airbyte Cloud | ~$500–$1,000 |
| Fivetran Starter | $2,000–$5,000 |
Enterprise (50+ sources, 100M changed rows/month)
| Tool | Monthly Cost |
|---|---|
| Airbyte self-hosted | $200–$500 |
| Stitch Advanced | $250–$500 |
| Airbyte Cloud | $2,000–$5,000 |
| Fivetran Standard | $10,000–$30,000 |
The pattern is clear: self-hosted Airbyte is cheapest at every scale. Stitch is the cheapest managed option. Fivetran is 5–20x more expensive but the most reliable and feature-complete.
Feature Comparison
| Feature | Fivetran | Airbyte | Stitch |
|---|---|---|---|
| Connectors | 300+ | 350+ | ~130 |
| Auto schema migration | Best | Good | Basic |
| dbt integration | Native | Yes | No |
| Custom connectors | Functions (limited) | CDK (Python/YAML) | Import API |
| Self-hosting | No | Yes (free) | No |
| Open source | No | Yes | No |
| SLA | 99.9% (Standard) | Cloud only | Advanced plan |
| Incremental sync | Best | Good | Good |
| Error handling | Automatic retry + alerting | Basic retry | Basic retry |
| Monitoring/alerting | Best (built-in) | Basic (self-host) / Good (Cloud) | Basic |
Who Should Pick What
Pick Fivetran If:
- Reliability and zero maintenance are worth paying a premium for
- Your data team doesn’t have DevOps capacity to manage pipeline infrastructure
- Enterprise connectors (SAP, Oracle, Workday) are needed
- Automatic schema migration and error handling save your team hours per week
- SLA guarantees matter for business-critical data pipelines
Pick Airbyte (Self-Hosted) If:
- Cost is the primary concern — unlimited data for $20–$500/month in hosting
- Your team can manage a Docker/Kubernetes deployment
- Custom connectors for internal APIs are needed (CDK is the best builder)
- Data sovereignty requires pipelines on your own infrastructure
- You want open-source with no vendor lock-in
Pick Stitch If:
- Predictable flat pricing matters more than connector count
- $100/month for 100M rows is the right price-performance for your volume
- Your source list is covered by Stitch’s 130 connectors
- You want managed pipelines without Fivetran’s complex pricing
Best reliability (money no object): Fivetran — automatic schema migration, 300+ connectors, and 99.9% SLA. The premium choice for data teams that need pipelines to “just work.”
Best value: Airbyte self-hosted — unlimited data movement for infrastructure costs only. 20–50x cheaper than Fivetran at every scale.
Best predictable pricing: Stitch at $100/month — flat fee, no surprises. The simplest billing of any data pipeline tool.
Start here: Airbyte self-hosted if you have DevOps capacity. Stitch Free (5M rows) or Fivetran Free (500K MAR) if you want managed.
FAQ
What’s the difference between ETL and ELT?
ETL (Extract, Transform, Load) transforms data before loading it into the warehouse. ELT (Extract, Load, Transform) loads raw data first, then transforms it in the warehouse using dbt or SQL. All three tools on this list are ELT — they move raw data to your warehouse. Transformation happens separately.
Do I need a data pipeline tool or can I write scripts?
Custom scripts work for 1–3 simple integrations. Beyond that, maintaining API connections, handling rate limits, managing schema changes, and monitoring failures becomes a full-time job. A pipeline tool saves 10–40 hours/month of engineering time at 10+ integrations.
Can I migrate between these tools?
Pipeline configurations don’t transfer between tools. You recreate connector settings and destination mappings. The data in your warehouse stays — you just point new pipelines at the same tables. Migration takes 1–3 days for a typical setup.
Which data warehouse should I use with these?
All three work with BigQuery, Snowflake, Redshift, and Databricks. For cost: BigQuery (usage-based, cheapest for moderate queries). For flexibility: Snowflake (credit-based, scales well). For AWS shops: Redshift. See our Snowflake vs BigQuery vs Redshift comparison for details.
