Do we really need to write our own search engine?
We asked ourselves this question more than once while working on improvements to search in March 2006. At the time, there were bigger search engines out there, including some good open source ones, that we could have used to power Facebook search results.
Context & Ripple Effects
This post closes a question Facebook says it asked itself repeatedly during its March 2006 search work: whether to power results with one of the larger existing engines or a good open-source alternative, or build its own. It went in-house — a decision that only got more consequential after the Facebook Platform launch drew developer chaos and an analyst case that Facebook is 'the new AOL', a closed destination users may not need to leave.
First-order effects
- Every Facebook query now runs on infrastructure Facebook controls end to end, so relevance tuning, ranking changes, and new searchable fields ship on Facebook's schedule rather than an outside engine's.
Second-order effects
- With comScore confirming a flood of new traffic and thousands of third-party apps multiplying inside the platform, in-house search gives Facebook the only index that covers its own social and application data — content the bigger external engines were never built to reach.
Third-order effects
- All Facebook's framing of the move — 'Facebook Becomes the People Search Engine' — points at the structural stakes: if search over people and their activity proves more useful than keyword web search for many queries, a walled-garden platform starts competing with general-purpose search engines on their home turf.
The trend: Large consumer platforms are increasingly building proprietary search over their own non-public data rather than licensing external engines, turning search from a utility into a strategic moat.