Transforming Selectel’s Database Connection Experience
New users were dropping off after failing to connect to a database. Through research, prototyping, and design changes, we reduced churn by 12%, decreased support requests by 75%, and increased engagement by 3×.
Overview
About the product
A Cloud Managed Database is a type of database hosted, maintained, and managed by a cloud service provider. Instead of setting up and managing their own database servers, companies rely on the provider to handle everything — including storage, security, backups, and updates.
Users simply configure their desired settings through a creation form, making setup quick and hassle-free. This allows teams to focus on leveraging their data, rather than dealing with the technical overhead of database management.
Users
Cloud Managed Databases are primarily used by DevOps engineers in small, medium, and large enterprises because they streamline database management, ensuring scalability, security, and automation.
Problem
Analytics revealed a high churn rate: new customers were creating Managed Database clusters, then deleting them and leaving the platform. Through conducting UX research, I discovered that many new users were overwhelmed by the database connection process, which ultimately led them to abandon the product.
Goals
🔹 Decrease support load: enable clear connection process, reducing the need of Customer Support intervention.
🔹 Introduce clear milestones: establish milestones to help new clients easily navigate through the connection process.
🔹 Decrease churn rate: improve UX of connecting to a database to reduce drop-offs.
Research
To identify the cause of the high drop-off rate and explore solutions, I conducted in-depth research using both qualitative and quantitative methods.
Expert interviews
Since I was new to the product team and haven't used cloud managed databases before, I interviewed colleagues who use them daily to understand the user flow better.
Afterwards, I created this chart to show the flow of a new user:
.png)
We were focusing on registered users with active accounts, so it was clear they faced issues in either one of the final stages of the flow:
1️⃣ Cluster Creation form → The first step in using a Cloud Managed Database is filling out the form, where you select the database type and settings.
2️⃣ Connecting to a Cloud Managed Database → Clients must either copy scripts from the cloud panel or write them manually, then paste them into a console or other tools outside of control panel to connect to the database.
Heuristic Evaluation
I began by reviewing the current database creation and connection flow to identify friction points and inefficiencies. Based on that, I developed two hypotheses:
- Visibility issues: clients might overlook the Connection section because they don’t expect it to appear at the bottom of the Overview page.
- Poor information architecture: clients may struggle to find the information they need in the Connection section.
Old Connection section at the bottom of Overview page
User sessions analysis
To find the drop-off points I analysed user sessions using Posthog. The Cluster Creation form showed no major issues, but the Connection section with scripts for database connection had little interaction.
Support tickets analysis
Since the connection process happens outside our platform, I reviewed support tickets for related issues. The most common problem users faced was not finding the correct port for connecting to a database.
User interviews
To better understand user pain points, I conducted semi-structured interviews with 10+ users, focusing on their struggles with the current flow.
Problem Statement
Issues identified through support ticket analysis and user interviews overlapped by 70%. It was clear that the initial database connection process was overwhelming for new users. Instead of a seamless, self-service experience, they encountered significant friction with:
- Connection Section Visibility: Most users had trouble finding connection scripts in the UI.
- Poor Information Architecture: Even users who found the Connection section struggled to locate the correct ports for connection.
- Lack of Programming Languages in UI: The initial UI only included Bash scripts, while many users were looking for other programming languages.
- Lack of DNS Addresses for Replicas: The initial UI only included the DNS address for the master node, while many users were also looking for DNS addresses for the replicas.
- Support Dependency: Many users required direct assistance from Support, which slowed their onboarding and increased frustration.
Business impact
- High Churn Rate: Many new users stopped using our platform altogether after failing to connect to a database.
- Increased support load: The support team was overwhelmed with queries related to connecting to a database.
Design Process
Our design process follows the Double Diamond framework and Lean UX methodology, focusing on continuous discovery, definition, ideation, and implementation.
Low-Fidelity Prototyping
I started by identifying a more visible placement for the Connection Panel. User interviews showed that most users expected it to be at the top of the screen. Based on that insight, I proposed three layout options and conducted quick hallway testing with support team members who had little to no experience using Cloud Managed Databases.
Connect button that opens a modal, placed at the top right of the Overview page.
A link opening documentation page at the top left.
A separate Connection tab.
While the first option is common in the market (used by our main competitor, Yandex.Cloud), users found it misleading. They expected the “Connect” button to connect them directly to the database from the browser, but instead, it opened a modal with connection scripts.
🏆 A separate tab proved more intuitive for our users, so we decided to go with that option.
Information architecture
Before jumping into prototyping, I wanted to clearly structure all the information users need to successfully connect to a database. To make sure I didn’t miss anything, I had a quick call with our engineers.
I highlighted all the essential parts needed to successfully connect to a database.
Then I sketched a draft of the tab structure and asked my colleagues to review it for technical accuracy. Analytics showed that almost no clients used IP addresses to connect to Cloud Managed Databases. Since DNS addresses are more reliable technically, we decided to remove IPs and include only DNS addresses in the interface.
At this stage, we needed to decide which programming languages to support. We quickly ran a survey to identify the most commonly used languages among our clients. Based on the results, our engineers began creating sample scripts while I moved forward with high-fidelity prototypes.
High-Fidelity Prototyping
I created a few prototypes and quickly tested them with colleagues from the support team. After choosing the version that best addressed user pain points, I shared it with our design team for review. Following a few final tweaks, the prototype was ready:
Final version of the interactive prototype
Changes made to old UI
Outcomes&Impact
📉 12% reduction in churn rate.
📝 75% reduction in support requests related to the connection process.
📈 3× increase in users copying information from the Connection section.
😊 13% increase in Customer Satisfaction Score (CSAT).
