Black Friday 2026 falls on Friday, November 27, the day after US Thanksgiving, which federal law sets on the fourth Thursday of November (November 26 this year). Less than three weeks remain. It is too late to replace your ERP and too late to migrate a warehouse, but there is still plenty of time to stop the peak from exposing, in front of your customers, everything nobody has tested.
A peak almost never creates a new problem. It amplifies the ones you already have: a stock sync that misses one order in a hundred becomes ten incidents an hour, and a label template that takes three seconds to print stalls a dock. This article lays out a realistic path from T-21 to T+30 for an IT director and an e-commerce or logistics manager who share the same ERP.
It contains no market statistics. What matters is your own volume history, which is already in your systems. Take last year’s figures, apply the growth your sales team expects this year, and size from there.
What Breaks First During a Peak
Four areas account for most seasonal-peak incidents. Knowing them lets you spread your preparation effort instead of testing everything equally.
Inventory synchronization. Your web shop, marketplaces, stores and warehouse all read and write the same stock. When updates run as batch jobs every fifteen or thirty minutes, overselling appears within the first hour of a promotion: the customer orders an item the warehouse no longer has. It is the most visible failure because it turns into cancellations and negative reviews.
Integration queues. Orders arrive through APIs, connectors or files. Under load, queues grow, timeouts fire, and retries create duplicates. If nobody watches queue depth, the first alert is a call from customer service.
Printing and shipping. Carrier labels, waybills, customs documents: these outputs often depend on a third-party service, a print spooler or a single workstation. A slow carrier API is enough to block picking, even if the ERP is running perfectly.
Invoicing and postings. End-of-day jobs (invoicing, posting, payment reconciliation) are sized for normal volume. A peak multiplies their runtime, and they can spill over the backup window or into the start of the next business day.
For each area, ask your teams a simple question: “Who would know, and how quickly, if this stopped at 2 a.m. on Saturday?” Vague answers point to your priorities.
The T-21, T-7, T-0 Checklist
Preparation comes down to three milestones, each with a distinct goal: lock, verify, hold.
Change Freeze and Deployment Freeze
From T-21, open a freeze window: no version upgrades, no new custom development, no non-urgent configuration changes on sales, inventory and shipping flows. The freeze is not limited to the ERP. The web shop, connector, WMS, carrier integration and warehouse workstations are all in scope.
Define two rules in writing. First, what remains allowed: critical security patches and fixes for blocking bugs. Second, who rules on an exception: one named person, with a record of the decision. Without this, the freeze erodes through small exceptions that add up to a full deployment.
Load Testing the Flows
The peak is won or lost on flows, not on screens. Reuse the method from our guide on ERP performance testing before go-live and adapt it to three scenarios: gradual ramp-up, a sudden spike when a promotion opens, and sustained load over several hours.
A few principles to apply within two weeks:
- Start from your observed maximum volume, per hour, for orders, order lines, stock movements and shipments, then test at a reasoned multiple of it.
- Test the whole chain, from web shop to printed label, not each component in isolation. A fast component can be blocked by the one that follows.
- Test in a representative environment, or failing that, document the correction factor. A test on an undersized environment gives false comfort.
- Measure batch jobs (invoicing, replenishment, availability recalculation) under volume, because their runtime determines your overnight window.
If no test environment exists, a partial test in production, outside business hours, on a set of tagged orders, is better than no test at all. Warn the teams beforehand and plan the data cleanup.
Available-to-Promise Stock and Reservations
Available-to-promise (ATP) stock is not physical stock. It is physical stock minus reservations, quantities being picked, blocked stock (quality holds, uninspected returns) and, depending on your policy, a safety stock per channel.
Check three points before the peak:
- The availability formula: do all channels use the same one, or does each connector recalculate it its own way?
- Publication frequency to the web shop and marketplaces. During high-promotion hours, switch to near real time on the affected SKUs or, failing that, apply a safety margin on the most requested items.
- Reservation release time on abandoned carts and failed payments. Too long and you lock up sellable stock; too short and you oversell. This trade-off is covered in our guide to ERP for omnichannel retail and real-time inventory.
Promotion and Pricing Rules
Prices and promotions are the costliest source of error per minute. A wrong price published on the wrong price list is fixed in seconds, but the orders placed in the meantime still have to be honored or cancelled.
- Centralize the calendar: each promotion has a start and end date and time, an item scope, a channel and an owner.
- Test stacking rules: product promotion, discount code, loyalty offer, free shipping. Unexpected combinations produce discounts larger than the margin.
- Enforce price floors: an alert, or even a block, when the selling price falls below cost or below a defined minimum margin.
- Check the end-of-promotion switch: a promotion that does not stop at midnight is a Monday-morning classic.
Have the complete list reviewed by someone who did not enter the promotions: fresh eyes find what the author no longer sees.
Interface Monitoring and Alerts
A peak without monitoring is driving at night without headlights. Build on the principles in our article on ERP interface monitoring and set a reduced dashboard for the period:
- orders received by the web shop versus orders created in the ERP, with a tolerated gap close to zero;
- integration queue depth and the age of the oldest message;
- messages in error or awaiting reprocessing;
- time between order and release to picking;
- response times of the carrier APIs and payment gateway.
Every alert needs a recipient, a threshold and an associated action. An alert that fires with nobody knowing what to do is just extra noise at 3 a.m.
Go-Live Day and the Days Before
The T-7 milestone is for rehearsal. Run a one-hour dress rehearsal: a series of test orders, from the promotion opening to the printed label, with the on-call teams present. Use it to verify access, permissions and escalation phone numbers.
On T-0, the command cell is in place: a fixed status meeting at regular times, an incident log kept continuously, and one person dedicated to communicating with the sales team. Avoid ambitious technical decisions during the peak: you stabilize, you do not refactor.
Logistics Capacity: Reinforcements, Access Rights and Training
The best ERP does not pack boxes. Warehouse capacity depends on people, and the peak must also be prepared on the staffing side.
Temporary staff arrive with accounts to create, profiles to assign and permissions to limit. Prepare accounts in advance from standard profiles (picker, checker, dock) rather than copies of existing accounts. Each account should have an expiry date, and deactivation should be scheduled at creation. This is also a compliance point: excessive rights granted in a hurry are an audit-review classic.
Crash training should take under an hour and focus on three actions: pick, check, pack. A one-page sheet per workstation, with screenshots of the real screen, is a better substitute for a long manual. Pair new staff with an experienced operator on the first day.
The WMS must be configured for the peak: adapted wave picking, zone batching, replenishment rules for pick locations, and priority for orders with near delivery deadlines. If you have no WMS, the same principles apply to the ERP: pick lists sorted by zone, reserve locations close to the shipping docks, and replenishment scheduled for off-peak hours.
Finally, check the equipment: scanners charged, label printers stocked with consumables, backup workstations. A shortage of label rolls stops a dock as surely as a software outage.
Invoicing, Payment and Reconciliation at Volume
The order peak becomes an accounting peak a few hours or days later. Three processes deserve anticipation.
Invoicing. If invoices are generated at shipment, their volume follows warehouse dispatches. Check the job’s runtime under volume, its behavior on partial failure (restart, duplicates) and gapless invoice numbering, which tolerates no holes. If you are affected by e-invoicing mandates, also review how your transmission flows behave under load.
Cash collection. Payment providers pay out in batches, with commissions, fees and withholdings. Prepare clearing accounts and automatic matching rules so reconciliation does not depend on several days of manual entry. Partial refunds and installment payments are the cases that make automatic matching fail.
Taxes and postings. Cross-border sales add their share of VAT rates and place-of-supply rules. Check the configuration before the peak, not after: a wrong rate repeated across thousands of lines is expensive to fix.
On all three, agree with the finance department on the date of the first reconciliation and the acceptable variance threshold. This avoids discovering in December a gap that dates back to November 27.
Fallback Plan: Degraded Mode, Runbook and Escalation
Nobody wants to use it, and everybody is relieved to have it. The fallback plan answers one question: what do we do when a component fails, and who decides?
Define a degraded mode for each critical flow. Examples: if stock sync is down, stop selling the most exposed SKUs or apply a capped stock; if label printing fails, switch to a backup printer or the carrier’s portal; if the order interface is interrupted, import orders by file following a tested procedure.
Write a short runbook. For each foreseeable incident: symptom, check, action, person to notify. One page per scenario, accessible outside the ERP (a shared document, a printed copy). A runbook locked inside the system that is down is useless.
Organize escalation. Three levels are enough: the on-call team, the technical lead, and the business sponsor who can decide to suspend a promotion. Specify the timing: after fifteen minutes without resolution, escalate. Also plan vendor or integrator on-call support and check that the contract covers the period.
Decide the stop criterion in advance. At what stock variance or outage duration do you suspend an offer? This decision is far easier to make in cold blood on a Tuesday in November than at 9 p.m. on the day, under pressure from sales.
After the Peak: Returns, Credit Notes, Close and Lessons Learned
Black Friday does not end on Monday. The following weeks carry their own load, often underestimated.
Returns. Post-promotion and post-holiday returns arrive in waves. Prepare for receipt: a quarantine location, inspection rules (restock, downgrade, scrap) and a target processing time. A return sitting three weeks in a holding area is lost sellable stock and a customer waiting for a refund.
Credit notes and refunds. Automate what can be: credit note generation on inspected receipt, a link with the payment provider, correct allocation to the original invoice. Watch for duplicate refunds, which are common when customer service and the automation process the same request.
The close. Plan a specific month-end close with accounting: delivery cut-off, provisions for expected returns, cash reconciliation, checks on goods in transit. A cycle count on the best-selling SKUs lets you correct variances before year-end.
The retrospective. Within fifteen days, gather the teams for an hour with three questions: what held, what nearly failed, what do we change. Record the decisions in a dated action plan with an owner. The peak data (hourly volumes, incidents, processing times) becomes the sizing baseline for next year, and for the winter sales that are coming up.
The 5 Mistakes to Avoid
| Mistake | Consequence | Countermeasure |
|---|---|---|
| Testing components one by one, never the whole chain | Each piece passes, but the web-shop-to-label chain stalls at the slowest link (often a carrier API) and picking stops | Run an end-to-end test from order to printed label, at a reasoned multiple of your observed peak volume |
| Letting the freeze erode through “small” late changes | An untested change lands on the sales, stock or shipping flows days before the peak and cannot be rolled back in time | Name one exceptions arbiter, log every decision, and allow only security patches and blocking-bug fixes |
| Publishing stock in slow batches (every 15 to 30 minutes) | Overselling in the first hour of a promotion, then cancellations and negative reviews | Move hot SKUs to near real-time publication, or apply a safety margin on the most requested items |
| Having no written fallback plan | At 9 p.m. on the day, nobody knows who can suspend an offer or which degraded mode applies, so the outage drags on | Write a one-page runbook per scenario outside the ERP, with escalation after fifteen minutes and a stop criterion decided in advance |
| Ignoring the aftermath (returns, refunds, reconciliation) | Returns sit in a holding area, refunds are duplicated, and a November variance is discovered in December | Prepare a quarantine location, automate credit notes, and agree the first reconciliation date with finance |
Printable Readiness Checklist
| Milestone | Action | Owner | Done |
|---|---|---|---|
| T-21 | Open the freeze window and name the exceptions arbiter | IT director | ☐ |
| T-21 | Extract last year’s peak volumes per hour and per flow | E-commerce manager | ☐ |
| T-18 | Load test the full chain, from web shop to label | IT director, integrator | ☐ |
| T-18 | Measure batch job runtimes under volume | IT director | ☐ |
| T-14 | Validate the ATP stock formula across all channels | Logistics, e-commerce | ☐ |
| T-14 | Review the promotion calendar, stacking rules and price floors | Sales | ☐ |
| T-10 | Create temporary staff accounts with expiry dates | IT director, HR | ☐ |
| T-10 | Configure WMS picking waves for the peak | Logistics | ☐ |
| T-7 | One-hour dress rehearsal with on-call teams | Everyone | ☐ |
| T-7 | Set up the monitoring dashboard and alerts | IT director | ☐ |
| T-7 | Validate degraded modes and print the runbooks | IT director, logistics | ☐ |
| T-3 | Check equipment: scanners, label printers, consumables, backup workstations | Logistics | ☐ |
| T-1 | Confirm vendor, integrator and carrier on-call support | IT director | ☐ |
| T-0 | Fixed-time status meetings and incident log | Command cell | ☐ |
| T+1 | First check of orders, shipments and invoices reconciliation | Finance | ☐ |
| T+7 | Process the first wave of returns and credit notes | Logistics, finance | ☐ |
| T+15 | Retrospective and dated action plan | Everyone | ☐ |
Will your ERP and your flows hold up on November 27? If you want an outside view of your preparation plan, contact us: we will review the breaking points, thresholds and fallback plan with you. To structure your load tests, start with our guide on ERP performance testing before go-live.
