From feedback to action: Recap of our customer insights workshop with Thematic

This week, we hosted a live workshop with Alyona Medelyan, PhD, CEO and co-founder of Thematic, YC S17, on using AI to supercharge customer support through better feedback analysis. The conversation explored what it takes to centralize feedback at scale, break down silos, and turn messy input into insights that drive real change across an organization.

---

Founding Thematic

Alyona shared that Thematic was born out of repeated requests from companies struggling to understand what drives their Net Promoter Score (NPS).
While NPS surveys provide a number, the accompanying text fields:“Why did you give us this score?” hold the real insights.

Three different companies asked her to help decode these open-text responses.
Her early prototype combined AI-driven analysis with visualization to reveal recurring drivers of satisfaction and frustration. This blend of qualitative and quantitative data became the foundation of Thematic’s approach.

Centralizing feedback at scale

For large organizations, the hardest part is still connecting feedback across systems.

• Feedback lives in many places: survey platforms, support channels, social media, inbound complaints, in-app feedback, and reviews.
• Each tool is owned by a different team, creating silos and raising data privacy hurdles.
• Every channel has its quirks. Support is often dominated by complaints, while review sites skew heavily positive with just a few negatives.

Too often, reporting relies only on survey data, which reflects a small, self-selecting group of customers. As Alyona put it, this means companies risk missing a representative view of what people really want.

Keeping support quality consistent across languages and markets

Traditionally, global support centers relied on transactional metrics like CSAT, NPS, or customer effort scores. But as Alyona noted, these surveys suffer from low response rates and biased samples, as hardly anyone actually fills them out.

AI now makes it possible to auto-score every single interaction. Instead of relying solely on post-call surveys, companies can generate synthetic ratings and effort scores by analyzing conversations directly: Was the issue resolved? How much effort did it take for the customer? What drove the positive or negative outcome?

Breaking down silos

To prevent customer insights from getting trapped within individual teams, companies need shared alignment:

• Agree on a single source of truth: one technology or platform for analyzing feedback.
• Centralize in a warehouse (Snowflake or similar) for transparency and access.
• Adopt a “newsroom” model: customer insights teams showcase how an issue was surfaced, what actions were taken, and the business impact.

This creates positive reinforcement loops. When teams act on feedback, they see measurable results, which motivates further collaboration.

From feedback to organization-wide change

Feedback isn’t just for minor fixes. Alyona highlighted examples where companies used Thematic to drive major process changes:

• Prioritizing by impact: quantifying how much fixing a single recurring issue (like billing errors) could boost NPS.
• Comparing regions: revealing why processes worked in San Francisco but failed in Chicago, or why credit card acceptance broke down in a new country.
• Iterating repeatedly: some processes take several tries to truly fix, requiring measurement before and after each change.

Companies need everyone aligned not just on taking action, but on measuring whether those actions are effective. The key is making sure feedback is not just anecdotes or opinions, but real data that can be measured and acted on.

Ensuring insights lead to action

Too often, insights end up in a dashboard that no one acts on. One practical solution:

• Generate a monthly shareable report.
• Tag owners directly: “This is your area, what have you done?”
• Close the loop with customers once action is taken.

This creates accountability, builds trust with customers, and encourages them to keep giving feedback because they see it leads to real change.

Rethinking metrics in an AI-first world

An audience member asked if classical support metrics like CSAT and NPS are still relevant. Alyona’s view:

• Metrics aren’t going away because they provide essential benchmarks.
• But in an AI-first world, companies can create personalized metrics aligned to their mission.
  • For a social media platform, it could be engagement.
  • For a delivery service, it could be on-time fulfillment.
• AI makes it possible to derive these mission-specific metrics from every interaction, assess whether they are being supported or dragged down, and convert qualitative feedback into quantitative signals that can be tracked.

Key takeaways

1. Messy feedback is still valuable: the challenge is connecting it across systems.
2. AI-powered analysis enables full coverage of customer interactions, not just survey snippets.
3. Silos kill insights: centralization and transparency are essential.
4. Feedback should drive process-level change, not just incremental fixes.
5. Action matters more than dashboards: closing the loop builds trust.
6. Metrics are evolving: AI enables more company-specific, mission-aligned measurements.

---

A big thank you to Alyona for sharing her journey and strategies, and to everyone who joined us live.