Define the promise before measuring it
A delivery SLA records an agreed service commitment, its measurement method, exclusions, responsibilities and remedies. An internal target is an operating goal; an ETA is an estimate that may change. Neither automatically becomes a contractual guarantee. Specify when the clock starts, which time zone applies and whether the promise ends at arrival, attempted handoff or confirmed delivery. Preserve the original promised window when the ETA changes.
For e-commerce parcel shipping, separate warehouse dispatch, carrier transit and final handoff commitments. For local food delivery, separate preparation, driver acceptance, waiting at the store and travel to the customer. Your platform availability commitment does not guarantee an individual delivery time.
Calculate on-time delivery consistently
On-time rate = eligible deliveries completed within the original promised window / eligible deliveries due in the period × 100. Define how cancellations, failed attempts, reschedules and missing timestamps enter the denominator. Report exclusions and data completeness alongside the rate. A delivery with no completion timestamp is unknown until investigated; silently dropping it can overstate performance.
Track acceptance delay, pickup waiting time, pickup-to-door duration, failed-delivery rate, reassignment and complaints. Use median and a high percentile as well as the mean so a small number of long delays stays visible. Compare equivalent dayparts, distance bands, stores and courier models rather than ranking every location on a raw average.
Assign a failure owner and escalation path
Store teams own readiness and packaging. Dispatch owns assignment and reassignment. The carrier owns the transport commitments in its agreement. Support owns customer communication and claim follow-up. Set your own alert thresholds and name a primary and backup for late preparation, no accepted courier, stale tracking, bad addresses, failed handoffs and proof disputes. Escalate an incident with the order ID, timestamps, current state and next decision; avoid several teams independently booking replacement drivers.
Scale the policy across stores
Maintain one metric definition and a city-level operating appendix for opening hours, coverage, handoff procedures and provider escalation. Pilot the workflow at representative stores, review exceptions daily during rollout and compare performance with matching periods. Capacity planning should use local demand and driver availability; software alone does not establish how many stores a team can support.
Ordering.co's delivery reports support historical investigation with selected filters, timing panels and order lookup. Returned panels and data depend on context. Calculate your contractual SLA from the agreed events and denominator; do not assume every KPI is a native report or that an empty chart means no failures. Use driver reports and your delivery model to prepare a review, then discuss the operating scope.