Big Data Analytics
Turn this teardown into a decision-ready prompt for ChatGPT, Claude, or your agent.
If you only have a few minutes to spare, here’s what investors, operators, and founders should know about GrepData (W13).
GrepData was a YC W13 big-data startup whose product never became a standalone business. Its engineers could build large-scale predictive systems, but customers pulled the team toward consulting and resisted outsourcing a general data platform. In 2013, the team joined Amanda Kahlow, who brought a proven B2B sales-prediction use case, to form 6sense.[1]
This was a terminal product pivot rather than a conventional acquisition. GrepData disappeared, while its technical founders and data infrastructure became the base of a company valued at $5.2 billion in 2022.[2]
Viral Bajaria learned large-scale analytics at Hulu. In a 2025 interview, he recalled that a Super Bowl advertisement forced the young streaming company to prepare for a traffic surge that could melt its systems. He took over reporting and analytics, moved beyond SQL Server, and learned Hadoop and machine learning on the job.[3]
In 2012, Bajaria and former classmate Premal Shah committed to starting a company. Dustin Chang and Shane Moriah joined them. Eric Feng advised the team to apply to Y Combinator, and GrepData entered the Winter 2013 batch.[3]
The product aimed to make big-data analytics accessible to companies without building the full Hadoop stack themselves. The technical premise was real, but the buying motion was weak. Bajaria later said the company was “becoming a consulting business.” Business buyers invited their engineers into meetings, and those engineers often replied, “I don’t want to have somebody else do my big data.”[3]
Amanda Kahlow held the missing half. A consulting project for Cisco had shown that behavioral data could predict which B2B buyers were approaching a purchase. She needed a product team. In a 2014 joint interview, Bajaria summarized the fit: “For the Grep Data team it was the opposite; we had the product, and we needed the business side of things. It really was the perfect match.”[1]
GrepData's original product simplified large-scale data processing and predictive analytics. Public descriptions are thin, and no archived product manual was located. The team appears to have offered an end-to-end layer over Hadoop: ingest company data, run analysis, and return predictions without requiring the customer to assemble its own pipeline.
That breadth caused the commercial problem. A generic data platform crossed organizational boundaries. Business leaders wanted answers; internal engineers wanted to own infrastructure and data movement. Each deployment pulled GrepData toward bespoke consulting.
6sense narrowed the system around a concrete decision: which business account is likely to buy, what it may buy, and when sales should act. The product combined customer CRM and marketing data with external behavioral signals, identified anonymous activity at the account level, estimated buying stage, and sent recommendations into sales and marketing workflows.[4]
The technical work survived, but the unit of value changed. Customers did not buy a general big-data abstraction. They bought prioritized accounts and coordinated action. That product boundary gave 6sense a recurring annual subscription and a buyer in marketing or revenue operations.[1]
GrepData targeted companies with large datasets but insufficient internal analytics infrastructure. 6sense targeted B2B sales and marketing teams, especially enterprises with long buying cycles, anonymous research, and fragmented CRM and marketing data.
Read the complete post-mortem, the rebuild playbook, and the exact reasons GrepData is still worth studying now.