How Does Looker Work?

Looker reinvents how businesses interact with data, but understanding what happens behind the scenes can feel like peeling back the layers of a complex machine. It’s more than just a tool for creating pretty charts, its power lies in a unique architecture centered around a modeling layer that defines business logic once for everyone to use. This article breaks down Looker's core architecture, demystifies its "secret sauce" - LookML - and walks through how it all comes together to deliver interactive, real-time analytics.

What is Looker, Really? A Quick Overview

At its core, Looker is a cloud-based business intelligence and data analytics platform, now part of the Google Cloud family. Unlike some traditional BI tools that require you to import or extract data into their own proprietary engine, Looker operates with a different philosophy. It connects directly to your existing SQL-compatible database or data warehouse (like Google BigQuery, Amazon Redshift, or Snowflake) and leverages its power to run queries.

Think of Looker not as a container for your data, but as a smart and user-friendly interface for your database. It provides a web-based environment where both data analysts and business users can explore, visualize, and share insights from their company's data. Its primary goal is to create a reliable "single source of truth" where everyone in an organization can access the same governed metrics and definitions, ensuring consistency across all reports and dashboards.

The Core Concept: Looker’s "In-Database" Architecture

The single most important concept to grasp about Looker is that it almost never stores your data. Its architecture is "in-database," which means an overwhelming majority of the analytical processing happens directly within your data warehouse.

This is a major departure from tools like Power BI (which often uses its VertiPaq engine) or Tableau (with its Hyper engine), which frequently rely on pulling extracts of data onto a local machine or server for faster processing. Looker skips this step entirely.

Here’s how the in-database process works:

User Interaction: A business user opens a dashboard or an "Explore" (Looker's data exploration environment) and drags a few fields into their report, like "Total Sales" and "Region." They might also apply a filter, such as "Last 90 Days."
SQL Generation: Looker's brain, the LookML modeling layer, interprets these clicks. It translates the user's request into a perfectly formatted, efficient SQL query.
Query Execution: Looker sends this SQL query directly to your connected data warehouse (e.g., BigQuery).
Database Processing: Your data warehouse does the heavy lifting. It executes the query against your live data and returns a tidy result set back to Looker.
Data Visualization: Looker receives the results and immediately renders them as a chart, table, or number visualization for the user.

This approach comes with several powerful benefits:

Real-Time Data: Because Looker queries the database directly, the data is always as current as the data in your warehouse. There are no stale extracts or sync schedules to worry about.
Leveraging Database Power: Your analytics performance is directly tied to the performance of your data warehouse. If you've invested in a fast and scalable warehouse like Snowflake or BigQuery, Looker will be lightning-fast.
Security and Governance: Your sensitive data never has to leave the security of your database. Looker simply sends queries and receives results, minimizing data liability.
Centralized Logic: You don't have business logic scattered across dozens of different spreadsheets or dashboard files. It's all centralized in one place - the modeling layer.

The Secret Sauce: Understanding LookML

If the in-database architecture is Looker’s skeleton, then LookML (Looker Modeling Language) is its brain. LookML is a specialized language used to describe dimensions, measures, calculations, and data relationships in your database. It acts as a semantic layer, an abstraction that sits between your complex database schema and your business users.

Instead of forcing a marketing manager to learn SQL and remember that revenue is calculated as SUM(sale_price) from the order_items table, a data analyst can define this logic once in LookML. From then on, the marketing manager can simply select the user-friendly field labeled "Total Revenue" from a menu.

LookML allows data teams to curate the end-user experience, making data exploration intuitive and error-proof. It turns raw data tables into a governed, reusable set of business metrics.

Key Components of a LookML Model

A LookML project is just a collection of text files written in a YAML-like syntax, and it typically consists of these core components:

Models: A model file specifies the database connection and defines a collection of "Explores" that are available for that connection.
Views: A view file corresponds to either a table in your database or a derived table (a table that doesn't physically exist but is created by a SQL query). This is where you define the accessible fields. You declare fields as either dimensions or measures.
- A dimension is a grouping field, like customer_name, order_date, or product_category. It corresponds to a column in your data.
- A measure is an aggregate field that performs a calculation on one or more dimensions, such as count_of_orders, average_order_value, or total_revenue.
Explores: In the model file, you define Explores. An Explore is a logical object where you join multiple views together. For example, you could create an "Orders Explore" that joins the orders view with the users view and the products view. This creates the foundation for your business users' self-service analysis.

A Simple LookML Example

Imagine you have an e-commerce database with an orders table. A simplified LookML view for it might look something like this:

Here, we’ve told Looker:

This view is based on the ecommerce.orders table.
The fields order_id, created_at, and status are available as dimensions. The created_at field has been set up as a dimension group, so users can automatically group by date, week, or month without any extra work.
There are two measures: a simple count of all orders and a sum of the sale_price column, helpfully formatted as USD.

A data analyst builds hundreds of these definitions into a project. Once complete, a business user sees a clean, organized list of "Number of Orders" and "Created Date," free from the messiness of SQL syntax and database schemas.

The Day-to-Day: How Users Interact with Looker

With the powerful LookML model working in the background, the experience for a non-technical end-user becomes incredibly simple and empowering.

Step 1: The 'Explore' Interface

The journey begins in the "Explore" environment. A user selects a starting point curated by the data team, such as "Orders," "Website Traffic," or "User Demographics."

On the left side of the screen, they see a neatly grouped list of all the dimensions and measures available for that Explore. They simply click the fields they’re interested in - for example, "User Traffic Source," "Device Category," and "Number of Sessions."

As they click, Looker adds these fields to a report builder and automatically generates and runs the correct SQL query in the background. The results appear in a data table almost instantly.

Step 2: Customization and Visualization

From the data table, the user can easily:

Filter: Add filters using a graphical interface (e.g., "Status is Complete" or "Order Date is in the last 6 months").
Pivot: Pivot dimensions to explore relationships in the data, similar to a pivot table in Excel.
Visualize: With a single click, transform the data table into a bar chart, line graph, map, pie chart, or a variety of other visualization types.

Step 3: Building and Sharing Dashboards

Once a user creates a useful chart or table (called a "Look"), they can save it and add it to a personal or team dashboard. Dashboards in Looker are collections of these Looks, arranged on a grid. They are interactive, allowing users to apply dashboard-wide filters to update all visualizations simultaneously. This makes it easy to drill down and analyze performance for a specific product, region, or time period.

Putting It All Together: A Summary Workflow

To recap, here’s the end-to-end process that makes Looker work:

Connection: An administrator connects Looker to a company’s SQL data warehouse.
Modeling: A data professional writes LookML code, defining tables as views, setting up join logic in Explores, and creating dimensions and measures with business-friendly names. This code is version-controlled with Git for collaboration and rollback capabilities.
Exploration: A business user logs into Looker, selects an Explore, and starts combining dimensions and measures to ask questions of the data.
SQL Generation & Execution: Behind the scenes, Looker translates those user selections into a SQL query, sends it to the live warehouse for execution, and receives the results.
Insight: The results are displayed back to the user, who can then visualize the data, combine it with other charts on a dashboard, schedule reports via email, or set up alerts for when metrics cross a certain threshold.

This entire loop is designed to provide governable self-service analytics, balancing the freedom for users to explore with the structure and reliability data teams need.

Final Thoughts

Ultimately, Looker works by abstracting away the complexity of SQL through a powerful, centralized modeling layer called LookML. This enables it to query a database directly in real-time, giving business users a sandbox to explore governed, up-to-date data without having to learn code or risk misinterpreting raw database tables.

Building a robust LookML model is an investment that requires technical expertise, but it's one that can pay dividends by fostering a truly data-driven culture. If your organization lacks the dedicated data engineering resources for a tool like Looker, we built Graphed to solve this exact problem. It allows you to connect data from marketing and sales platforms like Shopify, Google Analytics, and Facebook Ads and use natural language - just plain English - to instantly create the real-time dashboards you need, getting you from data to insight in seconds, not months.