Do the arithmetic first. Take a hypothetical retailer with $400 million in annual sales, 30 percent of it booked in the roughly 30 days between Thanksgiving and Christmas Eve. A peak day is worth about $4 million. The other 335 days average a little over $800,000. One lost day in mid-December costs roughly what five days in March would.
The ratio understates the problem, because much of peak demand cannot be deferred. A customer who cannot check out in March often comes back on Thursday. A gift that misses the last shipping day is a lost sale, and sometimes a January return as well. Logistics providers face the same curve one step removed, with their customers' peaks stacked on top of each other.
That arithmetic should set the terms for everything that follows: how much change risk to accept, how much spare capacity to buy, and how quickly to pull the trigger on a contingency.
Start with a calendar, not a risk list
Peak is not one period. Mark the dates that carry disproportionate value or cannot be recovered: promotion launches, Black Friday through Cyber Monday, each carrier's last-ship dates for ground and express, the last day for in-store pickup, and the post-holiday surge in returns and gift card redemptions. For each, estimate the value of an hour and whether lost sales come back.
Two things fall out of that exercise. One is a short list of windows where tolerance for disruption is close to zero. The other is evidence that your recovery time objectives, almost certainly set for an ordinary week, are too loose for those windows. A temporary peak RTO for checkout, order management and warehouse systems, approved by the business and funded accordingly, is a reasonable response.
Change freezes, and what they miss
A change freeze is the cheapest risk reduction available in peak season. Its value depends on three details:
- Scope. Code and infrastructure are obvious. Configuration, pricing rules, promotion setup and network changes are where many peak incidents start, and they often sit outside the freeze.
- Exceptions. Define who approves an emergency change, what evidence they need (a tested rollback, an owner on call), and a fast route for security patches. A freeze with no exception path gets ignored.
- The thaw. A January release carrying eight weeks of deferred work is a risk of its own. Plan it.
The larger gap is vendors. Your freeze does not bind your software-as-a-service providers, payment gateway, content delivery network or security tools. The CrowdStrike Falcon outage of 19 July 2024 came from a vendor-pushed content update that customers' own change control never saw. Ask critical vendors, in writing, for their peak change windows and whether they offer staged rollouts or customer-level deferral. Our guide to mapping third-party and supply chain dependencies explains how to decide which vendors qualify.
Temporary staff
Seasonal labor solves a capacity problem and creates three continuity problems.
Concentration. If one staffing agency supplies several distribution centers, its recruiting shortfall becomes yours across the network. Ask the agency how much of its peak commitment it has filled, and check again a month later.
Access. Accounts, badges and handheld devices provisioned in the fall must be removed in January. Late provisioning slows the start of peak; late removal leaves open access for months.
Safety and accountability. New workers do not know the site, the evacuation routes or the equipment. OSHA treats host employers and staffing agencies as jointly responsible for temporary workers' safety. Put agency staff into the emergency procedures, the site accountability lists and, where appropriate, the notification system.
Watch key-person risk among permanent staff as well. The supervisor who knows the warehouse system workaround should teach two colleagues before peak, not during it.
Carrier capacity is allocated, not promised
Large parcel carriers manage peak by giving shippers daily volume allocations and adding peak surcharges. Volume above your allocation can sit on the dock. Know your allocation per day per shipping node, compare it with forecast, and decide in advance where the overflow goes.
Secondary carriers (regional parcel carriers, the Postal Service, local couriers) help only if the contract, label integration and pickup schedule are live before the first peak week. Onboarding a carrier in December rarely works. The same logic applies inbound. The port congestion of late 2021, with container ships queued off Southern California, showed how quickly arrival dates can slip past the selling window. Earlier purchase order deadlines and buffer stock of top sellers are cheaper than air freight.
If you are the logistics provider, the mirror image applies. Collect peak forecasts from each client by a fixed date, set allocations of your own, and write down which client's volume takes priority if capacity runs short. Your clients will ask that question eventually. It is better answered in a contract in September than on a phone call in December.
IT load
Load test against forecast peak with a margin (1.5 to 2 times the expected peak hour is common), and include the external calls checkout depends on: payment authorization, tax calculation, address validation, fraud screening. The slowest third-party service sets your real ceiling. Request cloud quota increases well ahead, since they are not always instant.
Then design the degraded modes. A waiting room that admits customers at a sustainable rate beats a site that fails for everyone. Feature flags that switch off recommendations, reviews or personalization under load free capacity for checkout. Move heavy reporting and batch jobs out of peak hours.
Payments: plan for the processor's bad day
An authorization outage at peak is the fastest way to lose a day's revenue. In stores, most card terminals can run offline in a store-and-forward mode, accepting transactions up to a floor limit and authorizing them later. Some of those will decline. Setting the floor limit is a credit-risk decision for finance, made in October, not by a store manager at noon on the Saturday before Christmas. Keep enough cash float to trade cash-only if it comes to that.
Online, the main defense is a second gateway or processor with routing logic, tested under load. Gift card platforms often run separately from card processing, so test them as a dependency in their own right.
Fraud attempts tend to rise with volume, and tightening fraud rules in the middle of peak raises false declines on good customers. Agree in advance who may change fraud thresholds, and how quickly a change can be reversed if approval rates fall.
Weather overlap
Peak sits squarely in winter storm season for much of the country. Winter Storm Elliott, which brought blizzard conditions and severe cold to large parts of the United States in the days before Christmas 2022, is the reference case: a storm in the final shipping week does damage out of proportion to its length.
Tie actions to forecasts. Treat National Weather Service winter storm watches and warnings for your distribution center and store regions as formal triggers to pull shifts forward, pre-ship orders, adjust delivery promises on the website and arrange lodging near the site for essential staff.
Store and DC contingencies
For each distribution center, decide where its orders go if it is down for a day: a second center, ship-from-store or vendor drop-ship. That works only if inventory is positioned so no single building holds all the stock of your top items. For stores, decide what happens to pickup orders when a store closes, and who tells the customer.
Keep records as you go. A peak-period business interruption claim is measured against what you would have earned, so forecasts, prior-year results and extra expenses such as expedited freight and overtime all matter. Our briefing on business interruption insurance covers the policy terms to confirm before you need them.
Agree the thresholds before peak starts
Most peak losses come from slow decisions rather than wrong ones. Agree triggers, actions and owners in advance. The values below are illustrations; set your own.
| Trigger | Pre-agreed action | Decision owner | Decide within |
|---|---|---|---|
| Card authorization rate 5 points below normal for 10 minutes | Route to secondary gateway; stores go offline within limits | Head of payments | 10 minutes |
| Checkout errors or page times past set limits for 15 minutes | Turn on waiting room; switch off non-essential features | Digital operations lead | 15 minutes |
| DC throughput under 80 percent of plan for two straight shifts | Reroute orders to alternate nodes | VP supply chain | 4 hours |
| Volume over carrier allocation two days running | Move overflow to secondary carrier; adjust delivery promises | Logistics director | Same day |
| Winter storm warning for a DC or store region | Pre-ship, pull shifts forward, update site promises | VP supply chain | 12 hours |
| Shift staffing under 85 percent of plan | Overtime, agency call-out, prioritize order types | Site general manager | 2 hours |
Peak readiness checklist
Twelve weeks out
- Peak calendar and value-per-hour estimates agreed
- Temporary peak RTOs approved for critical systems
- Carrier allocations confirmed; secondary carriers contracted and integrated
- Load test plan written and quota requests submitted
Eight weeks out
- Load test completed, third-party calls included
- Vendor peak change windows confirmed in writing
- Staffing agency fill rates reviewed
- Payment failover and terminal offline mode tested; floor limits signed off
Four weeks out
- Change freeze in force, exception path published
- Temporary staff briefed on safety and evacuation, and added to accountability lists
- Decision thresholds signed by each owner
- On-call and command roster covering through the first week of January
After peak
- Temporary access removed
- January release plan for the deferred backlog
- Short review of what nearly went wrong, while memories are fresh
- Loss and extra-expense records filed in one place
