Estimate · not measured
Can this model buy the item?
Read this when you are about to send an agent to check out. The runs are live purchase attempts. They tell you if the order finished, where it stopped, and what one success cost.
Did the order finish?
The percent is completed purchases on this store. That is the chance the agent actually buys the item.
Where did it stop?
Product page, contact, shipping, or payment. The chart shows the step that took the attempt.
What did one success cost?
Failed attempts stay in the price. Use that dollar amount when you pick the next model.
The answer on this stack
Loading…
AI mystery shopping
The same runs, read as a shopper. The mission is to buy the item. The report is the step that stopped the attempt, and what that attempt cost.
Which model to run next. Finish rate and spend per success.
All costs are OpenRouter prices. Share the table with your agent: results.md