Impact
- Clear visibility into top customer pain points
- Cut manual review from several hours to minutes
- Prioritized on trends instead of individual reports
- Reduced reliance on intuition
We receive a steady stream of customer support requests every day. Each request is labeled and tagged, but the system was built for organization, not understanding.
As volume grew, it became harder to answer a simple question: What actually hurts our users the most?
Problem
- Labels existed, but did not reflect real user pain
- High volume made patterns hard to see
- Similar issues were described in different ways
- Everything felt equally important
We had data, but no reliable way to extract insight from it. Prioritization often relied on intuition or whoever spoke loudest.
Constraints
- Could not replace the existing tagging system
- Labels could be inconsistent, missing or misused across requests
- Requests mixed technical issues, UX confusion, and user error
- Needed a lightweight solution, not a full ML pipeline
Decisions
I focused on extracting signals, not improving labels.
-
Grouped requests by underlying pain, not labels
Different tickets describing the same issue were treated as one problem
-
Prioritized trends over individual tickets
One loud request did not outweigh recurring smaller issues
-
Introduced simple weighting between frequency and severity
High-impact issues surfaced even if less frequent
-
Filtered out noise:
One-off issues, misconfigurations, and user mistakes were deprioritized
-
Chose simplicity over completeness
The goal was clarity, not perfect accuracy
This shifted the system from organizing tickets to explaining problems.
Execution
Built a system to group requests into consistent problem areas and surface recurring patterns. Mapped recurring issues into a smaller set of meaningful categories. Continuously refined groupings based on new data and edge cases.
How does it look like
This is an example of the monthly report generated from support tickets:
# Account Access Support Tickets - Scorecard Report
Analysis Date: December 1, 2025
Total Tickets Analyzed: 161
Tag Filter: Account Access
## How severe is the pain?
Score: 4/5 - Mostly blocking
Rationale:
- 55% blocking (high severity) - cannot access account, preventing all account activities
- 35% high friction (medium severity) - can access but with significant workarounds
- 10% inconvenience (low severity) - minor questions
- Distribution shows majority of issues are blocking critical workflows
Evidence:
- "I'm unable to login" (blocking)
- "I cannot access account to renew" (blocking)
- "We are currently unable to access the account" (blocking)
- "I need to change email but can't reset password" (high friction)
- Many tickets about being completely locked out
## Can we see exactly what customers were doing right before asking for help?
Score: 5/5 - Extremely obvious and consistent patterns
...
This replaced manual scanning of hundreds of tickets.
Outcome
This changed how we prioritized product work
- Clear visibility into top customer pain points
- Faster, more confident prioritization decisions
- Reduced reliance on intuition
- Better alignment between support and product
What used to feel like noise became a consistent signal for decision-making.
Next steps
- Add a tighter feedback loop with support and product
- Provide tooling so anyone in the organization can explore and generate reports
- Continue identifying recurring issues and shaping work to address them
This system made support data usable, but there is still room to improve how signals are captured and interpreted over time.