The on-time delivery KPI is the simplest formula in logistics and the easiest one to get wrong. Divide on-time deliveries by total deliveries, multiply by 100. Two minutes of arithmetic.
The trouble is that almost every term in that sentence needs a decision. What counts as on time? Does a delivery nobody was there to receive count as a failure? Do cancelled orders belong in the denominator? Businesses regularly discover their real on-time rate is eight or ten points below the one they’d been reporting, and nothing operational changed. Only the definition did.
This guide covers the formula, the definitional choices that determine whether your number means anything, the 2026 benchmarks, and the levers that actually move it.
The Bottom Line
- The formula is on-time deliveries / total deliveries x 100. The definition of “on time” is what determines whether the result is useful.
- 95% or better is good, with e-commerce and manufacturing leaders at 98%+. Between 85% and 94% is workable but leaky, and below 85% signals something structural (procurementaiagents, 2026).
- Failed deliveries cost roughly $17.20 per parcel in the US, and 70% of shoppers say they’re unlikely to return after one (SmartRoutes).
- Measure from the actual delivery timestamp, never the ship date. Shipping on time and arriving late is the most common way the number gets inflated.
Earn $1,200+/week delivering for local businesses
Driver benefits:
- Steady local business routes
- Weekly pay + tips
- Dedicated support team
- Flexible scheduling
The on-time delivery formula
On-time delivery rate = (on-time deliveries / total deliveries) x 100
If 950 of 1,000 deliveries arrived within the promised window, your OTD rate is 95%.
Three steps, each with a trap in it:
- Count on-time deliveries. Orders that reached the customer within the committed window, measured by actual delivery timestamp.
- Count total deliveries. Every delivery attempted in the period.
- Divide and multiply by 100.
That’s the whole calculation. Everything difficult about this KPI lives in step one and step two, not in the division. Before you report a number, settle the definitions below, because changing them later destroys your ability to compare quarters.
The on-time rate is one of six KPI families in a full operational scorecard, and the supply chain metrics and KPIs framework shows where it sits relative to the rest.
What counts as on time
This is the decision that determines whether your KPI is useful, and it has to be made explicitly, in writing, before you count anything.
The options, from loosest to strictest:
| Definition | What it means | Best for |
|---|---|---|
| Promised date | Arrived any time on the committed day | Freight, pallet deliveries, B2B restocking |
| Promised date with cutoff | Arrived on the day, before a stated hour | Most wholesale and B2B |
| Delivery window | Arrived inside a stated window, e.g. 8am to 11am | Local delivery, food, floral |
| Appointment time | Arrived within a tight tolerance of a slot | Catering, events, time-critical goods |
Pick the strictest definition your customers actually hold you to, not the loosest one you can defend. If a bakery customer needs product before they open at 7am, a 10am delivery on the right date is a failure no matter what your records say, and a KPI that scores it as a success is actively misleading you.
One subtlety worth settling: does early count as on time? For most freight, yes. For anything where the customer has to be present to receive it, or where the product needs refrigeration on arrival, a delivery two hours early can be as much of a problem as one two hours late. If early deliveries cause you trouble, define a window with both edges and count outside-either-edge as a miss.
Four ways the on-time number gets inflated
Nearly every surprisingly-good on-time rate has one of these in it.
1. Measuring the ship date instead of the delivery date. The single most common error. A shipment dispatched on schedule and delivered two days late counts as on time under this method, which measures your dock rather than your service. Always use the actual delivery timestamp.
2. Counting early-but-failed deliveries as successes. The driver arrived within the window, nobody was there, the goods came back. That’s a failed delivery that frequently gets recorded as attempted-on-time. First-attempt delivery success rates run about 97.2% in the US and 93.7% in the UK, which gives you a sense of how much volume this category represents (SmartRoutes).
3. Removing inconvenient orders from the denominator. Cancelled orders, weather exceptions, customer-caused delays. Some exclusions are legitimate, but every exclusion needs to be written into the definition in advance. Deciding after the fact which misses don’t count is how a KPI stops being a measurement.
4. Counting partial deliveries as complete. Half an order arriving on time is not an on-time delivery from the customer’s point of view. This is exactly what on-time in-full exists to catch, and so the gap between your OTD and OTIF rate is a number to track in its own right. OTIF sits in the order management KPI set alongside order accuracy.
Audit your own calculation against these four before you benchmark yourself against anybody. A business comparing an inflated 96% against an industry 95% is drawing the wrong conclusion twice.
On-time delivery benchmarks for 2026
| Performance band | On-time rate | What it means |
|---|---|---|
| World-class | 98%+ | E-commerce and manufacturing leaders |
| Good | 95% to 97% | The standard target for most operations |
| Acceptable | 85% to 94% | Working, with real room to improve |
| Problem | Below 85% | Signals systemic issues, not bad luck |
By sector, e-commerce targets 95% to 98%, manufacturing supply chains average 90% to 94% with leaders above 95%, and retail distribution treats 93% to 95% as good performance (procurementaiagents, 2026).
Segment before you conclude anything. A 95% average can comfortably hide a single route at 78% or one demanding account at 70%. Break the number down by route, by customer, by day of week, and by product category. The segment-level view is where the usable finding always is, and the aggregate is mostly for reporting.
What late deliveries actually cost
The business case for this KPI is better documented than for most.
Failed deliveries cost approximately $17.20 per parcel in the US and £11.60 in the UK, and 70% of shoppers say they’re unlikely to buy again from a retailer after experiencing one (SmartRoutes).
The retention effect is the expensive half. Research into late delivery consequences finds 23% of consumers refuse to reorder after a failed delivery and 21% lose trust in the retailer, and manufacturers running below 90% on-time lose an average of 15% of their customer base annually (eShipper).
That reframes the KPI usefully. The direct cost of a late delivery is a redelivery fee and a support conversation. The real cost is the share of affected customers who quietly start buying elsewhere, and it doesn’t appear on any invoice.
How to improve your on-time delivery rate
The levers, in rough order of how much they tend to move the number relative to effort.
Fix the promise before fixing the operation
The fastest improvement available to most businesses isn’t operational. If your actual average transit time is 1.9 days and you promise two days, you will miss constantly, because you’ve promised your average rather than your reliable case.
Look at your transit time distribution and promise the 90th percentile, not the mean. Promising three days and hitting 97% beats promising two days and hitting 82%, both in customer satisfaction and in the amount of support time you spend apologising. Customers forgive a longer stated lead time far more readily than a broken short one.
Measure on-time pickup as your early warning
A load collected late is almost always delivered late, and you know about it hours earlier. On-time pickup rate is the leading indicator that on-time delivery is the lagging indicator of, and most businesses track only the lagging one. It sits in the transport management KPI set along with transit time variance.
Attack the dock, not the road
When deliveries slip, the instinct is to look at routes and drivers. The cause is frequently upstream. If orders take 20 hours to get from receipt to the loading dock, no amount of route optimization recovers that, and a route planned for a 6am departure that actually leaves at 9am was never going to hit a morning window.
Order cycle time is the number to check, and it belongs to distribution. Inside the building, dock-to-stock time and picking throughput are measured as warehouse KPIs. Check both before touching your routing.
Stop promising stock you don’t have
A significant share of late deliveries aren’t logistics failures at all. The order was accepted, the stock wasn’t actually available, and the delay was baked in from the moment of confirmation. That failure originates in inventory accuracy, covered in the inventory management KPI set, and no delivery improvement will touch it.
The tell is an on-time rate that’s bad on specific SKUs rather than specific routes. If your misses cluster by product rather than by geography, you have an availability problem wearing a delivery costume.
Sequence routes by window, not by distance
Shortest-total-distance and most-windows-met are different optimizations, and they produce different routes. If you have hard morning windows, the route has to be built around them even at the cost of extra miles. This is also where stop density and delivery windows start to conflict, which is a real trade-off rather than a problem to solve. Several of the broader fleet optimization strategies that save time apply directly here, particularly sequencing and departure scheduling.
Communicate proactively when you’re going to miss
This doesn’t improve the KPI. It substantially reduces the damage a miss causes. A customer told at 7am that their 9am delivery will arrive at 11am can usually adapt. The same customer discovering it at 11:30 cannot, and that’s the experience that drives the churn numbers above.
Common on-time delivery mistakes
Reporting only the average. The aggregate is for the board. The segment breakdown is for fixing things.
Changing the definition to improve the number. If you loosen “on time” from a window to a date, your rate improves and your history becomes meaningless. Make definitional changes at quarter boundaries and annotate the chart.
Treating it as a driver scorecard. Most late deliveries are caused by decisions made before the vehicle loaded: late order release, overloaded routes, unrealistic windows. Measuring drivers on a number they only partly control produces defensive reporting, which is worse than no reporting.
Tracking on-time without tracking in-full. A 96% on-time rate on orders that are routinely short is not 96% service. Track both.
Frequently asked questions
How do you calculate the on-time delivery KPI?
Divide the number of deliveries that arrived within the promised window by total deliveries, then multiply by 100. Use the actual delivery timestamp rather than the ship date. If 950 of 1,000 deliveries arrived on time, your rate is 95%.
What is a good on-time delivery rate?
95% or higher is good for most operations, with e-commerce and manufacturing leaders targeting 98% and above. Between 85% and 94% is acceptable but indicates real room for improvement, and below 85% points at a systemic problem rather than bad luck.
Should on-time delivery be measured from the ship date or the delivery date?
Always the delivery date. Measuring from the ship date tells you whether your warehouse hit its schedule, not whether the customer got their order on time. It’s the most common reason a reported on-time rate is inflated.
Does an early delivery count as on time?
It depends on your product. For freight and pallet deliveries, early is generally fine. For anything requiring someone to be present, or needing refrigeration on arrival, early can be as disruptive as late. If that’s your situation, define a window with both edges and count anything outside it as a miss.
What is the difference between on-time delivery and OTIF?
On-time delivery only asks whether the order arrived in the promised window. On-time in-full requires that it arrived on time and complete. OTIF always reads lower, and the gap between the two tells you whether your problem is timing or completeness.
How can I improve on-time delivery quickly?
Start by checking whether you’re promising your average transit time rather than your reliable case, since repromising to the 90th percentile often adds several points immediately. Then measure on-time pickup rate as an early warning, and check order cycle time to see whether the delay is happening before the vehicle even loads.
Start with the definition
Before you try to improve this number, spend an hour writing down what it means. What window counts as on time, whether early counts, which exclusions are permitted, and which timestamp is authoritative.
Most businesses that do this discover their real rate is below what they’d been reporting. That’s an uncomfortable afternoon and the most valuable thing you can do with this KPI, because an honest 87% you can act on beats a comfortable 96% that doesn’t match what your customers are experiencing.
