CUSTOMER SERVICE

Sentiment Analysis

Prioritizing customer emotions to improve support

Team

Product Manager
Product Designer (me)
Engineering Team

Contribution

UX Research & Strategy
Interaction Design
Feedback-Driven Iteration

Timeline

3 months

The Problem

1 in 3 escalations happened before support knew the customer was upset

Most support teams focused on meeting SLAs (Service Level Agreement) and relied on reactive CSATs to gauge how customers felt. But by the time feedback arrived, the damage was often already done. These metrics failed to capture emotional shifts as they happened, leaving no early signs for the support to act on. As a result, frustration quietly built beneath the surface, leading to preventable escalations and churn.

The Outcome

85% of agents actively used sentiment analysis to prioritize tickets in real time

72%

Feature Adoption Rate

72%

Feature Adoption Rate

85.5%

Agent Engagement

100%

Active Usage by Adopters

65

New customers onboarded

The Strategy

Customer calls revealed 3 critical gaps: sentiment blind spots, workflow disconnects, and feedback delays

📞 Conducted 8 customer interviews

🧑🏻‍💼 Spoke to Admins, Supervisors, and CX Analysts

✉️ Focused on ticket-based workflows

Here is what we learnt from customers:

Even without sentiment analysis, teams recognize emotional cues but the recognition is often inconsistent and delayed.
Sentiment is usually detected too late in the ticket lifecycle, often after escalation. This makes real-time sentiment tracking essential not just post-resolution analysis.

Different roles have distinct emotional insights needs:
- Agents: Use sentiment to help prioritize tickets.
- Supervisors: Need live visibility to intervene early.
- Admins: Seek macro-level insights and trends.

Teams are already acting on emotional signals but in way that is manual, reactive, and inconsistent.

They want sentiment to drive automated workflows that help them act faster and smarter.

Takeaway:

Sentiment solution needed to drive proactive action, enabling supervisors to intervene early, admins to automate workflows, and agents to prioritize effectively, beyond just reporting trends.

Competitive analysis showed most tools treated sentiment as a passive metric, giving Freshdesk a chance to make it real-time and action-oriented.

What we found to draw inspiration from the competitors:

Gaps identified as opportunities for Freshdesk:

Gaps identified as opportunities for Freshdesk:

Prioritise. Prototype. Progress

While our path from discovery to delivery focused on solving critical problems, we first prioritised quick wins that delivered immediate value

From sticky notes to screens. Here’s how we turned ideas into interactions.

The solution we delivered in our first release

Agent capabilities

Supervisor capabilities

Admin capability

Our first release on Sentiment Analysis gained traction fast. But real-world use revealed a subtler challenge: visibility alone wasn’t enough; teams needed sentiment they could act on with clarity and control.

The challenges uncovered post adoption

Admin challenges: A one-size-fits-all sentiment model didn’t work

• V1 used a fixed sentiment scale, but businesses defined sentiment differently.
• There was also no way to track how sentiment changed during a conversation.

Agent challenge: Negative overload, no clear prioritization

Most tickets showed up as “Negative,” making prioritization unclear and leaving agents demotivated.