Case context: A logistics startup is building a model to predict which delivery vans will need maintenance soon. The engineering lead wants to spend two months trying to prove, on paper, the single best model class before collecting any baseline results.

Question: As a machine learning consultant, what is the weakness in this plan, and what should the team do instead?

Sample answer: The weakness is assuming the best solution can be identified reliably before any experiments are run. For a new machine learning task, the right model choices are usually not obvious in advance, so the team should avoid long theoretical delays. A better plan is to build a simple baseline quickly, measure it on real data, and then improve it through repeated experiments.

Key points:

• The best approach for a new ML problem usually cannot be known in advance.
• Early experimentation is more useful than months of speculation.
• A baseline system should be built first, then refined with data-driven iteration.

Rubric: The response must identify the flaw as overconfidence in preselecting the optimal method and recommend starting with a baseline and iterating based on measured results.