Sample report
A run that reached the cart and still failed the customer.
This is the report Agent Journeys produces, shown here for a product-selection journey on a fictional store. The surfaces are the real ones; the run is a worked example.
Run #1042 · 7 steps · 15.2s · Alexa-style (simulated)
Step-by-step replay
Every action, and what the agent saw.
Select any step to expand what the agent actually did, what it observed, and the frame captured at that moment.
Example run. Sample data for a fictional store — nothing here was measured on your site.
Run #1042
The agent could not complete the job on your site.
The correct jacket reached the cart, but the delivery deadline the customer asked for could not be established anywhere before checkout.
WebsiteScore
Reached the cart, could not satisfy the delivery requirement.
Graded by claude-haiku-4.5 · grader v1 · prompt v1 · confidence 0.86
Steps
7 recorded
Actual action
No delivery estimate on the product page; checked the cart and shipping pages
Observation
Delivery dates are only calculated after a postcode is entered at checkout, which is past the point this journey is permitted to reach.
Simulated shopping-assistant profile styled after Amazon’s Alexa for Shopping (formerly Rufus). This is Omni Impact’s own agent configuration — not Alexa, and not affiliated with, endorsed by, or operated by Amazon.
Recommended fixes
Three changes, ranked by what they unblock.
Every finding arrives with the run that produced it and the dimensions it moves, so the work can be prioritised against something.
Expose delivery estimates before checkout
A postcode-free estimate on the product page would have let the run satisfy the customer’s deadline requirement instead of abandoning it.
Unblocks
Action readiness · Task completion
Publish temperature range in the specification table
Cold-weather suitability is implied by the copy but never stated as structured data, so it could not be verified.
Unblocks
Selection quality
Retire the outdated returns page
A legacy page still states a 14-day window against a configured 30-day policy, producing a factual-accuracy discrepancy.
Unblocks
Factual accuracy
