Four weeks before Black Friday, most enterprise retailers are in code freeze. The system you have now is the system you'll run peak on. However, you can still find out where this system will break, and decide what to do when it does.

In our last article, we argued that fulfillment networks don't fail because of traffic. They fail in small places at the same time: a store count that's off, a distribution center that's full by lunchtime, a carrier delay that never reaches the checkout. None of these needs new code. All of them need someone to have looked before the volume arrives.

So in this blog post, you'll find five checks in the order you need to do them. Each one has a deadline. Miss it, and the check turns from preparation into firefighting.

‍

1. Four weeks out: Do you know your stock accuracy per location?

Black Friday, 8 pm. Your largest store shows 40 units of your bestseller. There are 6 on the shelf. Online orders keep coming in, and your system keeps accepting them.

Most retailers can name their inventory accuracy. Fewer can name it per store. That difference matters, because a solid network average can hide the three stores that ship most of your online orders and sit well below it. At peak, every order accepted on stock that isn't there becomes a cancellation, a substitution or a customer who doesn't come back.

This is the first check because it takes the longest to fix. Counting takes store hours, and store hours get scarce from mid-November. What you can still do now: run cycle counts for your top peak SKUs in every store that ships or hands out online orders. Where you know accuracy is weak, raise the safety buffers. Showing "out of stock" online costs you a sale. Selling stock you don't have costs you the sale and the customer.

‍

2. Three weeks out: Who decides when a location is full?

Cyber Monday, 2 pm. Your main DC has hit its picking capacity for the day. Three stores nearby still have staff and stock. New orders keep routing to the DC.

At peak, the bottleneck is rarely the total capacity of your network. It's orders landing in the wrong place. Most routing rules know where the stock is. Far fewer know how much a location can still pick and pack today. So the DC keeps receiving orders it can't ship until tomorrow, while store capacity sits unused.

The real check isn't technical. It's this: is there a defined backlog threshold at which you reroute or split orders, and a named person who makes that call? If the answer is "the team will see it", you don't have one. On Cyber Monday afternoon, everyone on that team is busy with something else.

Three weeks out is the latest point to settle this, because it involves people outside your own team. Write down the daily capacity limit of every fulfilling location, even if your system can't use it yet. Agree the backlog level that triggers manual rerouting, and who pulls that trigger. Then brief the store teams on the ship-from-store volumes they should expect. They should hear it now, not on Black Friday morning.

‍

3. One week out: Can you change a delivery promise without a release?

Your main carrier announces a one-day delay for parts of the country. Your checkout still says "delivered tomorrow".

Customers accept a longer delivery time they were told about. They don't forgive a promise that breaks. The problem at peak is that promises are often hard-coded somewhere: in checkout logic, in a carrier integration, in a config file only one developer knows. When conditions change by the hour, a promise you can't adjust is a promise you'll break.

So the check is simple: can someone in operations change promised delivery dates without a code change? If not, you'll be running peak with the promise you set in October.

The week before Black Friday is the last moment to prepare for that. Update carrier cut-off times for peak days and fix your last guaranteed delivery date before Christmas. Write the delay messages for email and SMS now, so nobody drafts them under pressure. And on your highest-volume days, promise conservatively. One day more on the screen beats one day late at the door.

‍

4. Peak day: Who owns which exception?

Cyber Monday, 10 am. Order export to your ERP is running 45 minutes behind. Nobody is sure whether that's a problem, or whose problem it is.

The most expensive question on a peak day isn't technical. It's "whose call is this?" A 45-minute delay is harmless at 10 am and serious by noon, because at peak, small delays compound. What decides the outcome is how fast someone notices, and whether that person is allowed to act.

The check: is there a war room with a named owner for each exception type: stock, routing, carrier, system? "IT handles system issues" doesn't count. A name counts, along with a deputy and a clear line on what that person may decide alone.

This check belongs to peak day itself, but it only works if it's set up before. List the five exceptions you most expect, and give each one an owner. Agree on peak-day contacts and response times with every system and carrier partner. And set up a simple view of order backlog per status. A manual one beats none, because you can't own an exception you can't see.

‍

5. December 27: How long until a return is sellable again?

December 27. More parcels are coming back than going out. Returned items pile up in stores and the DC, and your online shop shows them as out of stock.

Peak doesn't end on Christmas Eve. It turns around. The returns wave decides how much of your peak revenue you keep, and how fast returned items earn money again in January. The volume is not small: the returns research group at the University of Bamberg expects around 560 million return parcels in Germany this year, with the returns rate stable at 24.1% (etailment).

The check: do you know how long it takes from return receipt to sellable stock? Not as a feeling, as a number. Most retailers treat returns as a cost centre at the end of the process. In practice, they're part of the order lifecycle, and stock that sits in a returns cage for two weeks is stock you're not selling.

Returns also bring risk. In a bevh/CRIF survey of 76 German online shops, 47% were affected by returns fraud between June and August 2026 (bevh/CRIF via Presseportal). Volume makes that harder to catch, not easier.

What you can still do: set a target time from receipt to sellable stock and measure it from day one. Define which returns stores may put straight back on the shelf. Review your fraud rules before volumes rise. And plan staff for late December and January now, while you're still planning staff for November.

‍

What these five checks have in common

Look at the five checks again. None of them is about traffic, and none needs a new release. Each one describes a decision that, at peak, depends on a person: someone who adjusts a buffer, reroutes orders, changes a promise, owns an exception or releases a return back into stock. And on Black Friday, that person is usually busy doing something else.

For this season, the answer is ownership, thresholds and workarounds, agreed before the volume arrives. For the next one, the better question is which of these decisions should still depend on a person at all. That's what an Agentic OMS is built for: moving decisions like these out of spreadsheets and war-room calls and into the system, where they're made continuously, with the person in charge deciding where the limits are.

‍

[cg_inject-el="card"]