How to Create a Monthly Report in Looker with AI
Creating a monthly report in Looker often feels like a necessary chore - a time-consuming cycle of navigating Explores, filtering dimensions, and tweaking visualizations. This article will show you how AI, specifically Gemini in Looker, can streamline this process, turning a tedious task into a quick conversation. We'll cover the traditional method and then show how you can get to the finish line faster with an AI-powered approach.
Why Monthly Reporting Is A Pain
Monthly reports are your business’s health check. They reveal what’s working, what's not, and where you should focus your energy next. For a marketing team, this could mean tracking campaign performance, lead generation, and cost per acquisition. For a sales team, it's all about pipeline velocity, close rates, and quota attainment. But while the insights are valuable, the process of getting them is often draining.
The traditional reporting workflow involves:
Manual Data Pulls: The familiar routine of exporting CSVs from multiple platforms - Google Analytics, your CRM, ad managers - and trying to stitch them together.
Repetitive Tasks: Building the same report visuals month after month, with slight tweaks and adjustments.
Technical Hurdles: You often need a solid understanding of Looker’s specific language, LookML, or at least be very comfortable navigating its Explores and dashboards. Asking a simple follow-up question can mean starting the whole process over.
By the time you compile the report, answer follow-up questions, and send it out, a significant chunk of your week is gone. This is where AI assistants, like Gemini in Looker, promise a different way of working.
The Old School Way: Building a Report Manually
Before jumping into the AI solution, let’s quickly walk through the manual steps you’d normally take to create a monthly report in Looker. Understanding this process highlights exactly what AI automates.
Step 1: Define Your KPIs
First, you have to know what you’re trying to measure. Don't drown in data. Pick a handful of key performance indicators (KPIs) that align with your monthly goals. For an e-commerce business, this might be:
Monthly Revenue
Average Order Value (AOV)
Customer Acquisition Cost (CAC) vs. Customer Lifetime Value (LTV)
Traffic-to-Purchase Conversion Rate
Step 2: Start with a Fresh Explore
In Looker, you would head over to 'Explore' to start building a single visualization. From the Explore page, you click 'Explore' on the left side of your Looker homepage to show the available list of models for your instance. You would next pick a model to start a session. Now, let's explore your dataset. In the next section, choose a desired visualization from the ‘visualization bar', for example, a bar chart or a treemap.
Step 3: Begin Pivoting a Field
The field menu enables a user to add, reorder, sort, and pivot fields on the main explorer page. Filters and additional visualization options from the visualization toolbar and data table may also be applied to explore pages.
Step 4: Running a query by tapping ‘Run’
You’ve set up some filters and visualizations, so it's time to use the 'Run' function (or command-Enter on a Macbook, on Windows, it is Control-Enter) to run a particular query. If you've made other prior edits, those queries do not take effect until you run a new query. If enabled on the instance, Quick Start can assist with building a pre-populated Explore page that can be edited or employed in its original shape.
Step 5: Name and Run Another Query
Now you're ready to create reports for others on the team (or for yourself). It’s time to move things over to a folder for organization where you can name individual reports for editing at future dates. From the main directory, you should move the file to a selected folder and name the selected folder. This step is crucial for creating easy editability for recurring weekly/monthly events and helps your team collaborate seamlessly. This feature provides a simple search and retrieve functionality that reduces overall build time.
Step 6: Assemble Individual Reports into a Single Looker Dashboard Reporting
Looks are not dashboards, but dashboards are made from multiple independent reports, so Looks are the building components of a final dashboard. Once several queries have been saved as reports, the dashboard process can finally commence:
Go to: Folders >>, My Folder and tap the “New” button at the top and select “Dashboard.” You can decide to categorize folders in groups: Shared Folders, My Folder, and Boards which act in the same manner as a mood board where you can gather individual dashboards.
Add your first "Tile". Tap “Add Tile.”
Select a pre-existing Look that you want added to the dashboard.
Choose the preferred data for the report, next, you can edit it for title, add notations, or adjust parameters as needed.
Set the most desired frequency and scheduling option via ‘Run in Query' to make sure the data stays regularly updated.
Save your edit for future additions. In Looker, as we’ve discussed, things need to be manually created. There are no autosaves that work on your behalf in case you get distracted with work. You can return later to add more dashboard tiles at any time.
Add whatever filters you want the dashboard viewers to use as default options for interaction. These could include date ranges for M/M, Q/Q, or Y/Y reporting. You could easily go wild with filters across regions, countries, campaigns by source, and so on, making for an impressive granular viewing experience in a single view for all.
Again, while powerful, you can see how this becomes a lengthy, click-heavy process, especially if your report has ten or more visualizations. Each chart requires its own mini "build" process.
Using Gemini in Looker to Speed Things Up
Looker’s integration with Gemini aims to replace the majority of those manual clicks with natural language. Instead of building queries by selecting dimensions and measures, you can simply ask Looker what you want to see via 'Explore with Gemini'.
Step 1: Enable Gemini access in Looker
Admins for instances hosted on Google Cloud must follow specific steps to enable Gemini integration with their Looker instances. First, they must enable certain APIs via the Google Cloud console API Library page:
"Compute Engine API"
"Looker API"
"Vertex AI API"
Enable these for a billable account because there are pay-as-you-go costs. For other instances or those self-hosted, refer to the "Pre-Sales FAQs for Gemini" to see all requirements and documentation for all use cases.
Step 2: Just type what you want Looker to build
After clicking to open the 'Explore with Gemini’ button, you just type a natural language question about what visualization you want generated after selecting your model from your dataset. Then, press enter like any conventional chat experience. Looker will try its best to estimate what charts meet your needs based on the inputted question.
Step 3: Select A Data Visualization
You receive an answer to any query as a preliminary generated visualization, for example, a column chart that shows “Orders Count” aggregated by some time factor, say Months, and “Orders Created" from two categories of data "Jackets” and "Pants” to see if any sort of correlation may exist between sales of those garments over a certain period. Looker even takes it upon itself to name relevant axes such as "Order Count versus Order Creation Month." Once you like the chart, you can add it to an existing Looks folder or start an entirely new dashboard.
The Good, the Bad & the Ugly: A Candid Perspective on In-Platform Generative AI in Looker
The Good
Looker now enables you to perform very specific (and often technical) reporting tasks, like helping you debug Looks immediately by asking conversational prompts about errors you’ve seen pop up. You get an AI suggestion for any corrections instantly. You can use AI to generate formulas from a prompt about what calculation you specifically like to perform. It can save you a mountain of time on more strategic analytical initiatives that your team cares about, for example, analyzing terabytes of ad performance data for Google Ads that's in your BigQuery database rather than downloading it as a CSV file to your personal computer. For expert data users, this is a welcome addition to the Looker stack. As they continue to improve this functionality as native to Looker, it will continue to grow.
The Bad
Most BI providers might claim their generative AI tools are ready-made for all business customers, but that might not be true. They might only work inside certain data warehouses. For example, you must be a Looker-hosted customer in a particular cloud availability zone. And your Looker billing plan will likely need to be a business tier version with advanced tools. On many Looker plans, you’ll be tasked with provisioning Gemini, which might be challenging. These new generational AI updates in your favorite BI tools often sound too good to be true during a free trial, only to uncover hidden requirements in Looker's documentation later. This can be frustrating for anyone, regardless of technical skill, especially if documentation is outdated or incomplete. Approach these updates with caution.
Tips for Getting More From Any Analytics AI
Whether you're using it inside Looker or another tool, getting useful outputs from a data AI is a skill. It’s less about knowing SQL and more about knowing how to ask good questions. The machine is the order-taker, but you still need to provide an excellent order for it to cook up the report you need.
Be Clear and Specific
An AI can't read your mind. Ambiguous questions lead to generic answers.
Vague: "Show traffic."
Specific: "Create a line chart showing daily website sessions from Google Analytics for the last 30 days."
Build and Then Refine
Your first prompt doesn’t have to be perfect. Start with a basic chart and then tell the AI what to change. This iterative process feels more like a conversation and helps you drill down to the exact insight you need.
Initial Prompt: "Show me a bar chart of our top 5 countries by revenue from Shopify this quarter."
Follow-up 1: "Now, break that down and stack the bars by product type."
Follow-up 2: "Okay, filter out all sales from the 'Accessories' category."
Follow-up 3: "Change this to a table and add a column for average order value for each country."
This "chat-your-way-to-the-answer" method is far more intuitive for most people than hunting through menus and filters.
Ask for Interpretations
A good data AI shouldn’t just give you a chart, it should help you understand it. After generating a visual, ask questions that push for insights:
"What's the most significant trend in this traffic data?"
"Why is there a spike in revenue on the 15th of last month?"
"Based on this, which marketing channel seems to be the most cost-effective?"
This moves you from simply reporting numbers to creating a narrative around them, which is exactly what a good monthly report should do.
Final Thoughts
Monthly reporting doesn't have to be a drag. The traditional method of building dashboards in Looker, while powerful, often requires a significant investment of time and technical knowledge. By using Gemini with your Looker instance in a natural, conversational way, you can go from empty dashboards to a finalized monthly report much faster, allowing you to transition from data analysis to actionable insights for the business.
However, it's much easier if you don't even need to be in Looker at all. At Graphed , you simply describe the dashboard you want in a few short sentences, and Graphed builds it for you. All we require is that you connect some accounts (Google Analytics, social or marketing platforms - whatever you need). Then just ask - all without navigating Looks. We connect directly to your most critical marketing and sales platforms, so you can create live, interactive dashboards using only plain language prompts. Graphed automatically updates your live dashboard in real-time, so your entire team never has to lift more than a single mouse click to get all the answers they need for the future of your company.