Part 3 of 4 · 2 min read
Exercises
Easy
Easy 1: Run it (30 minutes)
Run python -m pytest -q and python run_all.py in code/mini-pricing.
Done when: you can explain the difference between cost per task and cost per resolved task.
Easy 2: Your task (30 minutes)
Replace Assump with the token counts, tool costs and rates of an agent you built.
Done when: you report your cost per task, per resolved task and a price floor.
Hint
use measured, not planned, token counts.
Easy 3: Recompute a figure (30 minutes)
Recompute the $56.39 example from the caution above (1,695,384 tokens at $0.15 and $0.60 per million) and write down what you get.
Done when: you can state the formula and your result.
Medium
Medium 1: Cap the tail (3 hours)
Add a per-customer monthly cap or quota to the seat model.
Done when: you show negative-margin customers fall and report what the cap does to heavy users.
Hint
caps should degrade gracefully (slower, cheaper model) rather than just refuse.
Medium 2: Churn and price (3 hours)
Add a simple demand model (usage falls as price rises) and find the revenue-maximising price per task.
Done when: you plot revenue and margin against price.
Hint
start with a constant elasticity; fit it later.
Medium 3: Tiered plans (3 hours)
Add three tiers with included usage and compare with the hybrid model.
Done when: you report margin, negative share and customer spread.
Hard
Hard 1: Real cost per resolved task (8 hours)
Compute it from real traces and billing for a month. (I did not run this, so your numbers are the ground truth.)
Done when: you report cost per task, per resolved task and margin per customer.
Hint
join traces to billing on a customer id and a task id.
Hard 2: Price experiment (8 hours)
Test two price structures with real prospects or an internal team. (I did not run this.)
Done when: you report conversion, usage and margin under each.
Hard 3: Model migration (6 hours)
Re-price after switching to a cheaper model with lower resolution. (I did not run this.)
Done when: you show whether cost per resolved task went up or down.