Looker Explore Assistant and how it helps reduce bottlenecks
Introduction
This blog is part of a series of blogs on the Looker Opensource GenAi Extensions. In this series, we will discuss the Why (value added) of these extensions to your organization’s data stack like in this blog. We also covered the How here where we introduce the Step-by-step setup guide after our trials and errors with the newly released extension.
What is the Explore Assistant?
Recently rolled out as an open-source extension package of Looker, the Explore Assistant bridges the challenging gap that non-technical users often face when using Looker. Now, even without a technical background, users can directly ask the Explore Assistant for the information they need.
Who is the Explore Assistant for?
This Explore Assistant is for you if:
- You are a manager or executive who needs to access data and make data-driven decisions more efficiently but does not have the technical skills and is not familiar with Looker. For example, you often find yourself waiting for reports from the data team or struggling to extract insights from complex datasets, this extension can empower you to query data using natural language
- You are a data manager who is facing a shortage of data analysts who can support other departments’ need for data timely and cannot provide Looker training to all business users. If you’re overwhelmed by requests for data from various departments, this tool can help bridge the gap by enabling non-technical users to perform their data exploration
How to use Explore Assistant?
Below is a quick demonstration of how Explore Assistant could help your organization foster a more data-driven culture. If you are interested in finding out how to set up this extension, check out our other blog.
How does Explore Assistant help benefit you?
Imagine yourself as the manager of digital marketing and you would like to evaluate the performance of your social media platform campaigns which were rolled out last month, this is how Looker Explore Assistant can significantly enhance your data-driven decision-making process.
1. Simplified Data Exploration
The Looker Explore Assistant revolutionizes data interaction by translating plain language input into complex queries. This eliminates the need for users to have a deep understanding of LookML.
Now let’s get back to your digital marketing campaign. Typically, you would rely on your analytics team to pull data and provide insights. However, due to the lack of analysts, there isn’t anyone available to guide you. This is where the Looker Explore Assistant comes into play. Your team members, even if they have never used Looker before, can now directly ask the Explore Assistant, “Show me the sales performance for the last week by region,” and the Looker Explore Assistant translates this plain language input into a Look.
This capability allows your team to quickly access and explore the data, empowering the team to make informed decisions without waiting for technical support.
2. Fine-Tuning with More Input Examples
One of the most significant benefits of the Looker Explore Assistant is its ability to fine-tune your data analysis by incorporating more input examples. For example, during a recent campaign, you noticed varying engagement rates across different social media platforms. By inputting multiple examples related to these metrics, the Looker Explore Assistant becomes better at providing deeper insights into platform-specific performance.
Data stack requirements
This sounds amazing. However, for Looker Explore Assistant to work these are the requirements:
- Google Cloud Platform: Your data stack should be on Google Cloud Platform. Specifically, your database needs to be stored in Big Query, and Looker is your primary BI tool
- Functional Explore: the Explore in use should be mature enough that it does not require frequent updates. The frequent updates would lead to frequent maintenance for the examples of input and output
- Consistent Naming Conventions: Your dataset should follow consistent naming conventions. This is essential for the language model to accurately understand and return insights.
- Example Inputs and Outputs: Provide a few examples of natural language inputs and corresponding outputs to serve as a foundation for the model.
What are the and 👍and 👎?
This tool is great at👍
- Replacing point-and-click data exploration with natural language: the Looker Explore Assistant will give your first-time Looker users a head start while getting familiar with Looker. They can simply ask questions and get the result.
- Basic prompts, sorting, filtering, period comparison, and pivoting: these require some basic Looker knowledge but the extension can handle the tasks accurately
- Adapting to your data update: by giving it a new example input, output, and refinement, the extension would be able to learn the new context and return your desired output. The great thing about extension is how accurate it could be just by being given a few examples, which surprised me.
What this tool still needs improvement on 👎
- Understanding visual cues: the Looker Explore Assistant still cannot be able to understand visual cues very well and the result Look would still need some manual adjustments. Hope this will be improved the the later releases
- Work on multiple explore: currently, the Looker Explore Assistant only works on one explore at a time and a completely new extension would need to be installed foreach explore
- Forecasting: is not in the scope of this extension
- Inconsistent result: when given the same prompt, the extension can return different fields for the result. This could be improved by giving more explicit examples input and output
Conclusion
The Looker Explore Assistant is an open-source feature built to bridge the gap between data analytics and everyday business needs. Although there is still room for improvement, it is a great head start for your users to get used to Looker and save time for other complex tasks. For additional tools to enhance your data consumption, explore our great blogs on VertexAI Actions and Dashboard Summarization.
