49. Model A costs 2 units per attempt and succeeds 80% of the time; model B costs 3 and succeeds 95%. What does that tell you?
EasyAI products
Case
Show model answer
Under those simplified assumptions, attempt cost divided by success rate is 2.5 units for A and about 3.16 for B. That calculation compares expected cost per successful attempt across a volume of independent tasks.
It does not price human correction, failure severity, latency or repeated retries. Examine those factors and uncertainty before choosing a model for a particular customer workflow.
What interviewers listen for
- Calculate cost per successful attempt
- Include correction and failure costs
- State independence and scope assumptions