Case Study

Choosing a Fast Baseline for a Ticket Routing Project

Case context: A product team is starting a new support-ticket routing system for a software company. One engineer wants to classify tickets using keyword counts, another wants to rely on sender department metadata, and a third wants to build a more elaborate model that combines message embeddings with account history. The team is unsure which direction deserves the most effort.

Question: Using the idea of rapid prototyping, what should the team do first to settle the disagreement and learn which approach is most promising?

Sample answer: The team should avoid spending a long time debating the perfect design before seeing evidence. A better first step is to build a simple routing model quickly, even if it is clearly incomplete or inaccurate in some cases. After that baseline is running, the team can inspect its errors to learn where it fails and which improvements are most likely to matter. That evidence helps the team choose among the competing ideas instead of guessing.

Key points:

  • Do not spend a long time arguing for an ideal design before testing
  • Build a simple working baseline quickly, even if it is not very strong
  • Use error analysis on the baseline to identify the most promising next step

Rubric: The answer should recommend constructing a basic, likely imperfect system quickly rather than designing the final solution first, and it should explain that examining the baseline's errors will help decide which direction to pursue.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI

Related