PPPick Path Notes
Issue 21  dock to promise · eight sequences2026-11-20
CUTOFF  /  Note 08

There are three cutoffs, and only one of them is on the website

The cutoff published on the product page is a promise. The cutoff the warehouse works to is a schedule. The cutoff the carrier enforces is a fact. A fulfilment operation that manages only one of the three will miss the other two.

LaneCutoff
Sections10
Checks12
Filed2026-11-20
The cutoff published on the product page is a promise. The cutoff the warehouse works to
The cutoff published on the product page is a promise. The cutoff the warehouse works to is a schedule. The cutoff the carrier enforces is a fact. A fulfilment operation that manages only one of the three will miss the other two.

Three clocks, three owners

The first clock is order entry: the time by which an order must be placed to be included in today's dispatch. The second is pick and pack completion: the time by which the parcel must be on the outbound dock. The third is carrier tender: the time by which the parcel must be scanned by the carrier.

Each clock has a different owner. The merchant owns the first because it appears on the storefront. The warehouse owns the second because it is a labour plan. The carrier owns the third and enforces it without discussion.

The gaps between them are the buffer, and the buffer has to be explicit. A storefront promising a two o'clock cutoff, a warehouse working to noon and a carrier tendering at four is a coherent design. A storefront at two, a warehouse at two and a carrier at two thirty is a design that will fail on any busy day.

The most common error is publishing the warehouse's cutoff as the customer cutoff. That removes the buffer entirely and converts every minute of warehouse delay into a missed promise, which is the opposite of what the customer-facing number is for.

The carrier's tender window is the real constraint

A carrier does not arrive at an arbitrary time. It arrives within a window, and the parcel must be scanned by the time the vehicle leaves. The vehicle leaves when its route requires it to leave, which depends on the volume already loaded and the distance to the next stop.

This means the tender time moves. A high volume day pushes the departure earlier because loading takes longer, and a day with a late vehicle pushes it later. A warehouse that treats tender time as a constant will be correct on average and wrong on the days that matter.

The carrier's scan is the event that defines on time dispatch, because it is the timestamp the carrier will use in any service dispute. An internal completion time of three fifty eight with a carrier scan at four twelve is a late dispatch, regardless of what the warehouse system recorded.

Which is why the internal target should be set against the scan rather than against the dock. A parcel on the dock at four with no scan until four fifteen has not been dispatched, it has been staged.

Time zones are a promise calculation, not a display setting

The storefront shows a cutoff in the customer's time zone. The warehouse works in local time. The carrier operates in its own network time. Three representations of the same instant, and every conversion is an opportunity for an error that produces a late order.

The conversion should happen once, at the point where the promise is computed, and the resulting deadline should be stored as an instant rather than as a local time. A deadline stored as two o'clock local is a deadline that becomes wrong when the clocks change.

Daylight saving transitions are the specific hazard. A cutoff that falls inside the transition hour is either repeated or skipped, and a warehouse that computes deadlines by subtracting hours from local time will produce a two hour error twice a year.

The practical rule is that all internal deadlines are stored in coordinated universal time and displayed in local time. The display is a convenience and the storage is the truth, and mixing the two is how a cutoff silently moves by an hour.

Same day and next day are a policy, not a clock

A same day promise is not defined by a time of day, it is defined by the dispatch event. An order placed at eleven in the morning that ships at nine in the evening is a next day dispatch wearing a same day label, because the carrier will not move it until the following day.

The correct definition is in terms of the carrier's first movement. A same day order must be scanned into the carrier network before the last tender of the day. Anything after that is next day, however fast the warehouse processed it.

The consequence is that same day promises are constrained by the carrier schedule rather than by the warehouse's speed. A warehouse that can pick in twenty minutes still cannot deliver same day if the carrier's last tender was two hours ago.

The honest presentation is therefore a promise based on the next tender rather than on a fixed hour. A storefront that says order within two hours for dispatch today is more accurate than one that says order by two o'clock, because the former responds to the actual schedule.

Weekends and holidays belong in the calendar, not in a policy note

A carrier that does not operate on a public holiday does not tender on that day, which means the last tender before the holiday becomes the effective cutoff for orders that would otherwise have shipped during it. That is a calendar fact rather than a policy choice.

The calendar has two layers: carrier operating days and warehouse operating days. They differ, because a warehouse can work a day the carrier does not, and the work it does on that day is staging rather than dispatch.

The presentation consequence is that the promise on a Friday before a two day carrier closure is materially different from the promise on an ordinary Friday, and a storefront that shows the same number on both is misleading the customer in a way that produces complaints rather than cancellations.

The design answer is a promise calendar that is queried rather than written into a policy document. The query returns the next tender for a given destination and service, and every displayed promise is derived from that answer.

Peak periods need a temporary cutoff with an expiry date

During a peak, the warehouse cannot hold the normal cutoff because the volume exceeds the labour the cutoff was designed around. The honest response is to publish a temporary earlier cutoff, clearly dated, rather than to keep the promise and fail it.

A temporary cutoff has to expire automatically. A cutoff moved earlier for a peak and never restored becomes a permanent degradation of the promise, and nobody remembers why it moved. The expiry date is the mechanism that prevents drift.

The temporary cutoff should also be communicated as a range rather than as a single day. A peak that runs from the middle of November to the middle of December needs a cutoff that applies to the range, with the restoration date stated.

The internal version of the same idea is a capacity gate on order intake. When the day's volume exceeds what can be dispatched, the intake should slow rather than the dispatch promise should break, because a slowed intake is visible to the customer as a longer estimate while a broken promise is visible as a late parcel.

The service level starts at tender, not at order entry

A dispatch service level measured from order entry includes the customer's own delay, which the warehouse cannot control. A level measured from the cutoff includes only the warehouse's performance, which it can. The second number is the one that belongs in an agreement.

The measurement point should be the carrier scan, because that is the event both parties can observe and neither can dispute. A warehouse system timestamp is an internal claim, and an internal claim is not evidence in a service conversation.

The reporting period matters as well. A monthly average hides the peak days, and the peak days are the ones the customer remembers. A weekly report with the peak week called out separately is more useful than a monthly number that averages the failure away.

The target itself should be expressed as a percentage of orders scanned before tender, with the tender time per service stated alongside. A ninety eight percent target against a tender that moves is a target whose meaning changes weekly.

Late orders: downgrade the promise rather than ship it late

When an order has missed the cutoff, the choice is between shipping it late against the original promise and shipping it on time against a revised promise. The second is almost always better, because the customer can act on a revised promise and cannot act on a broken one.

The revision should be proactive rather than reactive. A notification sent when the cutoff is missed, with the new estimated delivery, produces fewer contacts than a notification sent after the customer notices the tracking has not moved.

The internal decision that makes this possible is a rule about which orders get pulled forward. A late order that can still make the next tender should jump the queue; a late order that cannot should be rescheduled rather than expedited, because expediting it produces cost without producing an earlier delivery.

The cost of the downgrade is a goodwill cost that should be tracked separately from the freight cost. Operations that track only freight will systematically prefer to expedite, because the expedite cost is visible and the goodwill cost is not.

System time and human confirmation

The system decides the cutoff by comparing timestamps, which is reliable and consistent. A human deciding the cutoff applies judgement about whether a specific order can still make it, which is sometimes correct and always inconsistent.

The override has to be possible, because a genuinely urgent order is a real case. The override also has to be recorded, because unrecorded overrides convert a measurable promise into an anecdote. The record is the reason code and the approver, and both should be mandatory.

The dangerous override is the batch one at the end of the day, where a supervisor releases everything in the exception queue to clear the screen. That action makes the dashboard clean and makes the promise false, and it is invisible in the data because the release looks like a decision rather than a bypass.

The control is to compare the number of overrides against the number of late orders. A warehouse with many overrides and few late orders is one where the system is being corrected. A warehouse with many overrides and many late orders is one where the overrides are not working, and the fix is upstream rather than in the override policy.

Publishing the calendar as a table rather than as a page

A promise calendar published as prose is a document that will be out of date within a month. Published as a table keyed by date, service and destination, it is a data source that the storefront can query.

The table needs four columns to be useful: the date, the service, the destination group, and the last tender time. From those, the promise for any order can be computed rather than looked up, and the computation can be automated across a catalogue.

The table should be maintained by the party that owns the carrier relationship rather than by the marketing team, because the carrier is the source of the underlying facts. Marketing owns the presentation of the promise, not its computation.

The final discipline is a monthly review against the actual scans. If the published last tender is four o'clock and the actual scans show a median of three forty, the table is wrong and the promise is being missed by a buffer that nobody knew existed. The review is what keeps the table honest.

The published form of these three clocks is the order cutoff rules set by carrier tender, and a cutoff defined by the tender rather than by the clock is the one that survives a busy day.

Checklist12 lines
  • 01Separate the storefront cutoff, the warehouse completion target and the carrier tender time.
  • 02Set the internal target against the carrier scan rather than against the dock.
  • 03Store every deadline as an instant in coordinated universal time.
  • 04Convert time zones once, at the point where the promise is computed.
  • 05Model the daylight saving transition explicitly instead of subtracting hours.
  • 06Define same day by the first carrier movement rather than by a clock time.
  • 07Derive the promise from the next tender rather than from a fixed hour.
  • 08Put carrier holidays into the promise calendar rather than into a policy note.
  • 09Publish a temporary cutoff for peak periods and date its expiry.
  • 10Measure the service level from the tender time rather than from order entry.
  • 11Downgrade the promise on a missed cutoff instead of shipping it late.
  • 12Log every manual override with its reason and its approver.