Looker
Looker is Google Cloud's enterprise BI platform, acquired from Looker Data Sciences for $2.6B (closed Feb 2020) and rebuilt in 2026 around Gemini for "agentic BI" (Conversational Analytics, Dashboard Agents, a LookML Agent, Managed MCP — PayPal runs 5,000 employees on Claude Desktop via Looker MCP). Google publishes no list pricing; Vendr's analysis of 355 real contracts puts average annual spend at $150,000, with a recorded high of $1.77M. Editions are Standard, Enterprise, and Embed layered on per-seat licenses (~$1,665/yr Developer, ~$799/yr Standard, ~$400/yr Viewer per third-party procurement data). Its defining technical trait is LookML, a git-versioned modeling language compiled to SQL against BigQuery/Snowflake/Redshift. Google names Looker a Gartner Magic Quadrant Leader for BI for the third consecutive year as of 2026; current release is 26.12 (July 2026), shipped biweekly. Current version/status: 26.12 (July 2026, biweekly release channel). License: proprietary SaaS. Pricing: No public list price (Google requires a sales call); Vendr's dataset of 355 real contracts averages $150,000/year with a recorded max of $1.77M, layered on estimated per-seat fees of roughly $1,665/yr Developer, $799/yr Standard, and $400/yr Viewer. Maintained by Google Cloud (acquired Looker Data Sciences, deal closed Feb 2020, $2.6B).
Pick Looker if you're already BigQuery-native and need one governed semantic layer serving both dashboards and embedded/agentic use cases — Managed MCP is genuinely ahead of Tableau and Power BI here. Skip it under 50 users or without budget for a dedicated LookML engineer; Sigma or a managed Metabase gets self-serve analysts to a dashboard faster and cheaper. The weakness Looker's own consultants concede: LookML isn't free after the license. Firms like Mammoth.io peg ongoing LookML development and maintenance at 40-60% of total Looker spend — on a $120K instance, that's another $50-70K/year in BI engineering time just to keep the model current, on top of BigQuery compute. Budget the semantic layer like a second product, not a config file. Known weakness: Consultants and Looker practitioners widely document that LookML modeling and maintenance consumes 40-60% of total Looker spend annually, on top of the license — the semantic layer is an ongoing engineering cost, not a one-time setup.
See how it ranks →