Looker is what BI looks like when a company decides metric definitions are too important to live inside individual dashboards. Everything routes through LookML, a modelling language kept in version control: define revenue once, and every chart, explore and embedded view inherits it. Since Google absorbed it, it’s sold as Looker (Google Cloud core), and it remains the most opinionated tool in our BI category. The opinion costs money and staffing, so let’s be precise about who should pay.
My verdict early
Looker is the right tool for organisations with a data team large enough to own a modelling layer and consumers numerous enough to justify governing them. For a small team it’s the wrong shape entirely: you’ll pay enterprise prices and spend engineer-weeks in LookML before the first stakeholder sees a chart that Metabase would have drawn on day two.
The pricing, decoded from the quote wall
There are no public dollar figures, but the structure is published and worth understanding before sales calls. Three editions, all annual, one to three-year terms. Standard is scoped for organisations under 50 users and includes one production instance, 10 Standard users, 2 Developer users and 1,000 query API calls a month. Enterprise raises the security features and the API ceiling a hundredfold. Embed is for customer-facing analytics at half a million query API calls. Additional people are licensed per seat in three flavours: Viewer, Standard and Developer, each priced differently, so the user mix drives the quote as much as the edition does.
Two line items people miss. The API quotas are real operational boundaries if you build on Looker programmatically, and crossing editions to raise them is a big step. And the conversational analytics features meter in tokens: each tier includes a monthly allowance (60 million input tokens on Standard, scaling up by edition), with overage billed at $3 per million input tokens and $20 per million output. AI questions against your data are a metered utility here, budget them like one.
Also, clearing up the perennial confusion: Looker Studio, the free dashboard tool formerly Data Studio, is a different product sharing a brand. This review is about the governed platform.
What the governance actually buys
Lived with, LookML’s discipline pays in three currencies. Trust: when finance and marketing disagree about revenue, the argument ends at the model definition instead of festering across dashboards. Reuse: a defined measure works in every explore, so analysts stop rebuilding the same logic. And embedding: serving governed analytics inside your own product is where Looker is genuinely best-in-class, which is why the Embed edition exists as a first-class citizen rather than an afterthought.
The costs are equally concrete. Every metric change routes through developers, so agility depends on your data team’s queue. Ad-hoc exploration is deliberately fenced. And the talent pool that writes good LookML is small and knows its price.
The scoreboard
Reviews are strong at serious volume: 4.4 from 1,653 reviews on G2 (64% five-star, 30% four, 4% three, nothing meaningful below), 4.5 from over a thousand on Gartner Peer Insights. The praise clusters on trusted metrics and embedding; the complaints on cost, the modelling bottleneck, and performance under heavy concurrency. Nothing there contradicts the shape I’ve described: this is a governance product, praised and resented for governing.
Looker or the lighter path
The modern alternative pattern splits Looker’s job in two: definitions live in dbt‘s semantic layer, and a lighter tool (Metabase, or Lightdash if you want the LookML-like feel without the platform) serves the charts. You give up the polish of one integrated system and keep perhaps ninety percent of the governance for a fraction of the spend. For companies already deep in dbt, that maths is very hard for Looker to beat, and I’d want the embedding requirement or a large consumer base on the table before recommending against it.
Shrinking the quote
Since the price is a negotiation, negotiate the structure rather than begging on the rate. The user mix is the biggest lever: Viewer, Standard and Developer seats price differently, and most organisations over-buy Standard seats for people who only ever open dashboards. Audit who genuinely needs to explore before the call. Term length is the second lever, with one to three-year options and the usual discount gradient; take the longer term only after a pilot has survived contact with your users. And if embedding is your real goal, price the Embed edition against building on the Enterprise tier’s API quota, because they solve overlapping problems at very different numbers.
If you ever leave
Plan the exit while signing, because with Looker the asset is the model, and LookML goes only where LookML is spoken. Lightdash reads the same conceptual language (it was built by dbt-and-LookML people for exactly this migration), and the dbt semantic layer can absorb the definitions with translation work. The dashboards themselves you rebuild anywhere; treat them as disposable. What you protect is the year of metric decisions encoded in the model, so keep it clean, reviewed and documented as if the migration is certain, and the option stays real.
Verdict
Best-in-class governed and embedded BI, honestly priced only for the organisations shaped like it: a real data team, hundreds of consumers, or analytics you sell. Everyone else is buying a cathedral to hold a book club. Know which you are before the sales cycle decides for you.
