Shipment delay impact
Gauge how much a delay matters to this customer, from a notification to urgent handling.
F@fProblem
Every delay gets the same templated email, whether the parcel is a spare phone case or the outfit for a wedding on Saturday.
Goal
Spend proactive contact and paid upgrades where a delay actually affects a customer's plans.
Input / state
The delay computed in code, the order, and anything the customer said about why they need it.
- delay_days
- Days past the promised date, computed in code.
- order_summary
- What was ordered.
- customer_context
- Anything the customer said about when or why they need it.
Decision question
delay_impactHow much does this delay affect the customer, given what they ordered and said?
Rubric levels
0Minor: no stated need and a small delay.1Noticeable: a longer delay, or the customer has asked where it is.2Affects a plan: the customer mentions a date, event or task the delay puts at risk.3Time-critical: perishable goods, or the delay will miss a stated date unless something changes now.
Allowed outcomes
Decided in your code from the answer and its confidence. Thresholds are yours to set and test.
- NotifyMost likely level is 0 and confidence meets your threshold.
- Proactive updateMost likely level is 1 and confidence meets your threshold.
- Offer optionsMost likely level is 2 and confidence meets your threshold.
- Urgent handlingMost likely level is 3 and confidence meets your threshold.
- Logistics agentConfidence is below your threshold: probability is spread across options.
Example scenarios
Predefined demonstrations. No model is called; the outcome shown is an illustrative judgment.
Example state
{
"delay_days": 1,
"order_summary": "Phone case",
"customer_context": ""
}Outcome
IllustrativeMost likely level is 0 and confidence meets your threshold.
Why
A one-day delay with nothing at stake.
Using it well
- Compute the delay from dates in code and pass the number. Date arithmetic is a known weak spot for language models.
- What counts as an option (upgrade, reship, refund of shipping) is your policy, applied in code.
Request definition
POST https://api.typesafe.ai/v1/systemone{
"model": "jev-latest",
"state": {
"delay_days": "<delay_days>",
"order_summary": "<order_summary>",
"customer_context": "<customer_context>"
},
"questions": {
"delay_impact": {
"type": "score",
"instructions": "How much does this delay affect the customer, given what they ordered and said?",
"criteria": [
"Minor: no stated need and a small delay.",
"Noticeable: a longer delay, or the customer has asked where it is.",
"Affects a plan: the customer mentions a date, event or task the delay puts at risk.",
"Time-critical: perishable goods, or the delay will miss a stated date unless something changes now."
]
}
}
}Request shape from the System One API reference; a Kev server accepts the same body. Replace the placeholder state with your data. Requests need your own API key or server.
Version History
Initial version