George Fraser co-founded Fivetran with Taylor Brown. After early attempts at a broader analytics product, customer requests for data connectors led them toward automated replication into data warehouses; they spent nearly two years finding their first customer. — A Word From Our Founders.

On Product Philosophy and Automation
Supported Principles:
- Make the complexity invisible: Fivetran hides source-specific engineering behind a simple source-to-warehouse connection. — Matt Turck Interview.
- Make data access reliable: Fraser described Fivetran’s mission as making data access as simple and reliable as electricity. — Software Engineering Daily Transcript.
- Narrow the product boundary: Fivetran concentrated on reliably replicating data into warehouses rather than trying to own each customer’s downstream analysis and business definitions. — Matt Turck Interview.
- Automate schema changes: Fraser argued that propagating source-schema changes should be handled by the service rather than by a person maintaining each pipeline. — Software Engineering Daily Transcript.
- Sell a managed outcome, not a toolkit: Fraser contrasted Fivetran’s responsibility for matching source data with integration toolkits that leave correctness to their users. — Matt Turck Interview.
- Maintain the connectors over time: Different APIs and schemas demand source-specific sync strategies; the continuing maintenance of those integrations is central to Fivetran’s service. — Matt Turck Interview.
- Make replication accessible: Fraser described a setup in which a customer connects a source and destination and receives replicated warehouse data without building the pipeline themselves. — Matt Turck Interview.
- Treat integration as ongoing engineering: Adding a connector is not the whole task: Fivetran engineers must handle changing APIs, source schemas and replication behavior across many systems. — Matt Turck Interview.
- Own data correctness: Fraser said mismatches between source and destination became Fivetran bug reports, and their fixes accumulated across customers. — Matt Turck Interview.
- Reduce configuration deliberately: Fraser wrote that not letting users configure warehouse schemas enabled automated maintenance and centralized bug fixing. — Evidence-Based Business.
Key Learnings:
- Solve the replication problem end to end: The company chose to own the source-to-destination copy and its correctness while leaving customer-specific data interpretation downstream. — Matt Turck Interview.
- Know when flexibility is warranted: Fraser acknowledged that a custom pipeline can be appropriate for a very large, unique, stable source; the managed service shines when common sources change frequently. — Matt Turck Interview.
- Take unglamorous infrastructure seriously: The founders describe data centralization as complex but important, and frame automation of that hidden complexity as their core work. — A Word From Our Founders.
- Let fixes benefit all customers: The founders wrote that a larger customer base exposed more source-system quirks, allowing the shared connector service to improve for everyone. — A Word From Our Founders.
On Entrepreneurship and Building a Company
Supported Principles:
- Persist through a long search: Fraser and Brown said it took almost two years and considerable determination to win their first customer. — A Word From Our Founders.
- Protect cofounder trust during the search: Fraser said their longstanding trust helped them persist through several unsuccessful product iterations before the first customer. — Matt Turck Interview.
- Follow the customer’s actual bottleneck: Customers using Redshift and their preferred BI tools asked Fivetran for the connector beneath its original pivot-table product; the team shifted to pipelines. — Fivetran Origins.
- Hire capable first-principles thinkers: Fivetran’s published values say it hires smart people and challenges industry assumptions through first-principles reasoning. — Fivetran Core Values.
- Fund the company’s actual stage: Fraser recounted modest early angel financing and low spending before a later seed round and faster fundraising after growth accelerated. — Matt Turck Interview.
Key Learnings:
- Let customer demand clarify product-market fit: Fivetran’s first connector customers were asking for replication into their own Redshift warehouses, clarifying which part of the original product they valued. — Fivetran Origins.
- Keep the early operation lean: The founders said early growth was supported by revenue and seed investors; Fraser also recalled paying themselves little while searching for their first customer. — A Word From Our Founders.
On Strategy and the Data Market
Supported Principles:
- Recognize the warehouse shift: Fraser identified Redshift’s lower-cost availability as important to the emergence of the cloud-warehouse market that Fivetran served. — Matt Turck Interview.
- Reconsider legacy integration assumptions: Fraser described earlier ETL tools as toolkits requiring customers to specify and maintain pipeline details; Fivetran instead took responsibility for managed replication. — AiThority Interview.
- Support multiple destinations: Fraser described Fivetran’s ambition to connect data sources to a customer-chosen database and ultimately support any database-like destination. — AiThority Interview.
- Tie billing to active data movement: Fivetran switched from fixed per-connection fees to usage measured in Monthly Active Rows, charging for changed rows rather than all stored rows. — Fivetran Usage-Based Pricing.
- Use specialized layers where they help: Fraser distinguished Fivetran’s extraction/loading role from warehouse analytics and customer-specific transformations done after the data lands. — Matt Turck Interview.
- Learn from a shared connector base: The founders said that more customers reveal more quirks in common data sources, making fixes reusable across the service. — A Word From Our Founders.
Key Learnings:
- Look for cost discontinuities: Fraser connected the modern data stack’s rise to dramatically cheaper warehouses, data storage and processing than the previous generation. — Matt Turck Interview.
- Preserve the original data in ELT: Fraser explained that loading an unmodified copy before transformation makes it easier to revisit or repair downstream transformations. — AiThority Interview.
On Leadership and Culture
Supported Principles:
- Culture is created by participants: Fraser told colleagues that participating in Fivetran’s culture also means creating it. — Fivetran Culture Visit.
- Seek teammates with shared values and varied perspectives: Fivetran’s published values say it seeks people who share its drive without hiring people who are all alike. — Fivetran Core Values.
- Define the team value concretely: Fivetran’s published “One Team, One Dream” value asks colleagues to share setbacks and successes and seek diverse perspectives. — Fivetran Core Values.
- Let evidence challenge a leader’s ideas: Fraser urges leaders to assess evidence quality and accept negative experimental results even when they overturn prior beliefs. — Evidence-Based Business.
Key Learnings:
- Make initiative an explicit value: Fivetran’s published engineering values define initiative as taking ownership, being accountable and learning through coaching. — Fivetran Core Values.
- Make integrity an explicit value: Fivetran’s published engineering values define integrity as doing the right thing even when that is harder in the short run. — Fivetran Core Values.