// startup autopsy

NetPerceptions

Built the recommendation engine Amazon used — then Amazon rebuilt it in-house

competitionBankruptcy

Fatal mistake: Core clients internalised recommendation technology once it became strategically central — no replacement pipeline

Founded	1996
Closed	2003
Country	USA
Sector	Saas
Founder	Steve Snyder

// the model, blind

Evaluating only NetPerceptions’s profile at its peak — without knowing the outcome — the model ranked Acquisition gone wrong as the #1 likely cause. Documented cause: Competition.

Key Events Timeline

1996-01

FOUNDING

NetPerceptions founded

1999-06

PIVOT

Strategic pivot under pressure

2003-06

SHUTDOWN

Bankruptcy: NetPerceptions ceases operations

Full Analysis

Free · no account needed

Documented cause

NetPerceptions built collaborative filtering software — the technology that powers "customers who bought this also bought" recommendations. Amazon was an early customer, licensing NetPerceptions' RecommendIt engine. As Amazon scaled, it internalised the capability and rebuilt recommendation logic that was too core to its business to leave in third-party hands. NetPerceptions found that enterprise clients viewed personalisation as a competitive differentiator they wanted to own, not license. Unable to replace Amazon and other departing flagship clients, the company filed for bankruptcy in 2003.

Lesson

“Identify whether your enterprise clients view your technology as infrastructure (they will keep licensing) or as competitive capability (they will eventually internalise). Price and retention strategy differ fundamentally between the two.”

Failure anatomy

Collapse type

Bankruptcy

📉 MEDIUM

Hype cycle

peak of inflated expectations

Moat type

Technology

Fatal mistake

Core clients internalised recommendation technology once it became strategically central — no replacement pipeline

FAQ

Is collaborative filtering the same as the technology Netflix and Spotify use?

Yes — collaborative filtering is the foundational technique behind modern recommendation systems. NetPerceptions was doing this at commercial scale in 1997-2000, years before it became mainstream. The underlying algorithms have since evolved enormously, but the core "users like you also liked" pattern is the same.

Did any personalisation software company survive from this era?

RichRelevance (founded 2006) and Certona (founded 2004) built successful personalisation businesses for retail by learning from the earlier generation's mistakes — focusing on mid-market retailers who could not build in-house rather than targeting Amazon-tier clients who would inevitably internalise.