Summary
Designing Glassnode Query for in-depth analytics
Glassnode Query enables users to create custom queries for Bitcoin and Ethereum blockchains, extracting insights into transaction volumes, user activity, and protocol performance. This customization ensures users get precise data for informed decision-making.
Product team
Marta Gajowczyk, Sr. Frontend Developer
Andy Jessop, Engineering Lead
Rafal Bromirski, Sr. Frontend Developer
Timeline
June 2023- November 2023
The initial idea
The initial idea behind this project started with using AI to query the blockchain. instead of users having to know SQL, they can simply write a natural language prompt, and it will process the query via openai and present the results.
POC
Proof of concept
The POC centered around the hypothesis that users could input natural language prompts using AI to obtain desired results.
However, during internal testing, it became evident that the POC did not fully meet the expected standards.
MVP
MVP
In response to internal feedback on the POC, which highlighted accuracy issues with the AI-generated results, we pivoted to a SQL-first approach for the MVP.
This strategic shift positioned AI as a secondary assistant, allowing us to prioritize the reliability and precision of query results.
Usability testing
Testing and Validation
After completing the MVP build, we were left with several unknowns that required validation. These unknowns included:
Can a non-technical user understand how to use AI to generate SQL queries?
What will be required for them to integrate this tool into their workflow?
How easy and intuitive do they find the query creation process?
We recruited 30+ diverse participants matching our user persona, including those proficient in SQL queries. See logos of sourced companies below.
Synthesis
Synthesis
After conducting user interviews, we reviewed all the recordings during Synthesis and extracted valuable insights to share with stakeholders as actionable items.
Key takeaways
Our most newsworthy takeaways
We were wrong about
• Our table structure is easily understandable.
• Users will find example queries helpful.
We were right about
• One chart type won’t be enough for them to represent the data.
• AI will make writing SQL faster and retrieving data faster.
Surprises!
• 3 companies decided to purchase Glassnode query after using it.
• Many participants who are using Dune rather wait longer then pay for a plan to execute faster.
Solution
Solutions
The strategic pivot to a SQL-first approach proved effective in addressing user concerns and enhancing the reliability of query results. User testing indicated improved satisfaction and usability.
Impact
Impact
After releasing Glassnode Query to existing customers, we observed a positive impact on retention and sales.
200+ queries created
12% increase in retention
4% increase in sales