~/talk

Beyond Surveys: Real-Time User Satisfaction Prediction

slides:none on file

A system called DNPS predicts session-level Net Promoter Score from telemetry, behavior and text using a hybrid stack of gradient-boosted trees, sequence models and BERT embeddings, fused with an explainability layer (SHAP/LIME). It claims 85.9% accuracy with MAE ≈ 1.21 and live inference at ~80 ms latency. Ablations show removing text drops accuracy to ≈55% and removing behavioral/telemetry drops it to ≈78%, and dashboards surface root causes like high back‑clicks and short sessions.

AI-written summary · about

  1. 0:00Intro: Beyond Surveys—Real‑Time NPS Prediction with LLMs + Telemetry
  2. 0:39NPS 101: What Net Promoter Score Measures
  3. 1:16How NPS Is Calculated (Detractors, Passives, Promoters)
  4. 2:10How Companies Collect NPS Today (and Why It’s Slow)
  5. 2:49The 4 Big Problems with Traditional NPS Surveys
  6. 3:52The Big Idea: Predict Sentiment Without Surveys
  7. 4:14Introducing DNPS: Dynamic Net Promoter Score
  8. 4:45DNPS Architecture Overview: Ingestion → AI Engine → Insights
  9. 4:56Data Collection: Behavior, Telemetry, and Text Feedback
  10. 6:06Data Processing Pipeline: Cleaning, Alignment, Feature Extraction
  11. 6:57Hybrid Modeling: Structured + Sequence + Text Models
  12. 8:34Fusion Layer + Explainable AI (SHAP/LIME)
  13. 10:01Results: Accuracy, Ablation Study, and Real‑Time Latency
  14. 10:59Key Drivers & Business Value from Explainability Insights
  15. 11:45Limitations, Future Work, and Closing
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.