Friday night, the second weekend of November. The line at register four is six deep, every card terminal is stuck on processing, and the shift lead is telling customers “one more minute” without knowing whether that’s true. Four hundred miles away, a distribution center’s sorter is still moving, but label printing has stopped and three trailers are waiting at the dock. Nobody on either floor is calling it a network incident yet. They just know it picked the worst week of the year to happen.
Peak season doesn’t ask whether your network is ready. El Buen Fin runs November 13–17 in Mexico this year; two weeks later, the Black Friday-to-Christmas stretch begins in the U.S. Together they put every circuit, failover rule, and support contract under real load. That test happens whether or not anyone prepared for it. The one thing you control is whether you run it first, in October.
Q4 volume is a load test you didn't schedule
The numbers behind those weeks keep growing. El Buen Fin 2025 closed with a record 219 billion pesos in sales, according to Concanaco-Servytur. In U.S. parcel networks, ShipMatrix projects that Cyber Monday volume can run about 70 percent above an average day. And when the systems that take payment go down, the losses are direct: research from Dynatrace, FreedomPay, and Retail Economics puts lost sales from payment outages at $44.4 billion a year for U.S. retail and hospitality, with nearly two-thirds of those outages hitting during peak trading hours.
For an operations leader, all of that reduces to something concrete. Payments, inventory, order routing, and the applications your teams work in all ride on connectivity, and in Q4 they all peak at once. October matters because it still leaves runway: time to add capacity where a site is thin, fix a single point of failure, and test a failover before holiday change freezes lock the calendar.
The review has four parts. None of them require an engineering background, and every one is a question you can put to your IT team or your provider this month.
Capacity: plan for the worst hour, not the average day
Average utilization tells you almost nothing about November. The number that matters is the worst hour of the busiest day, when every register is running, curbside orders are queuing, inventory lookups are constant, and the cameras and guest Wi-Fi are still consuming the same connection. Distribution centers (DCs) feel the same squeeze: every additional package means more scans, more warehouse-management calls to the cloud, more label and tracking requests hitting carrier systems.
Ask two questions per site. Does the circuit have headroom above last year’s peak, plus this year’s growth? And is that bandwidth actually yours? Sites on shared broadband slow down when the surrounding area gets busy, and in a shopping district that means exactly when your own traffic peaks. Dedicated Internet Access (DIA) avoids that trap: the bandwidth is exclusive to your site, symmetric in both directions, and can be scaled up before peak weeks instead of in the middle of them.
Run the exercise for every location that sells or ships, not only the flagships. A DC that loses its warehouse management connection stops shipping just as thoroughly as a store that loses its point-of-sale (POS) systems.
Redundancy: two invoices are not two paths
Plenty of operations discovered in November that their backup link was never really a backup. Two contracts with two providers can still enter the building through the same conduit, ride the same pole line, or share last-mile infrastructure down the street. One cut takes out both, and the redundancy you paid for all year does nothing.
The same payment-outage research found that only 44 percent of businesses keep a secondary internet connection at all. Having one puts you ahead. Having one that takes a genuinely independent route is what protects the site.
So ask the blunt question: can each provider show that its physical route into the building is independent of the other, all the way through the last mile? If the answer is a diagram, good. If the answer is “they’re different companies,” keep pushing. Our guide to redundancy and the last mile includes a short self-assessment for exactly this. In October, those questions cost you a meeting. In November, they can cost you a day of sales.
Failover: if it needs a human, it's too slow
Connections degrade more often than they die. A circuit dropping half its packets can hurt more than a clean failure, because nothing switches over while checkout crawls. The clock is unforgiving either way: shoppers give a frozen payment terminal about seven minutes before abandoning the purchase, while the average payment outage lasts close to two hours. A backup circuit that requires someone to notice the problem, find the runbook, and move cables is not a plan for that gap.
Failover should be automatic, triggered by degradation as well as outage. Just as important, the systems that sell should come first on the reduced capacity: application prioritization, known as quality of service (QoS), keeps payments and inventory responding even if back-office traffic slows for an hour.
Then prove it works. Pick a quiet weekday morning in October, fail the primary link at a pilot store, and time what happens. A failover you have never tested is a hypothesis.
Support: know what your SLA permits, and who picks up
The last part of the review is paperwork, and it’s the part most operations skip. Reread your service-level agreement (SLA), the uptime commitment in your connectivity contract, before peak. Even a strong 99.95 percent SLA permits around 21 minutes of downtime a month; our breakdown of what an SLA really means walks through the math. Decide whether your number matches what a November Saturday actually costs you.
Then get specific about people. Who answers at 2 a.m. on Black Friday weekend, and how fast? What is the escalation path when the first answer doesn’t resolve it: a name and a number, or a ticket queue? Are your site list and after-hours contacts current on the provider’s side? Does the provider freeze planned maintenance during peak weeks, and does its support calendar even recognize Buen Fin as a peak event? A ticket opened at 9 p.m. on November 14 that waits for Monday’s queue is the real service level, whatever the contract says.
Peak season crosses the border. Your provider should, too.
For retailers and logistics operators running in Mexico and the U.S., Q4 is a single event with two names. Buen Fin and Black Friday land within the same two weeks, and the traffic behind them rarely respects the border: stores in Guadalajara checking stock against a DC in Texas, freight data moving in both directions all season.
Each carrier handoff in that chain is a place where a ticket can stall while two providers point at each other. When something degrades during peak week, you want one provider to call and one team accountable, in either language. Flō operates its own network across both countries, with multiple international crossings and route diversity instead of a patchwork of intermediaries, service levels of up to 99.95 percent depending on topology, and a network operations center (NOC) staffed 24/7 in both Mexico and the U.S. That network has run for 25 years and serves more than 2,000 organizations, Fortune 500 companies among them.
Run the test on your own terms
Q4 will stress-test your network. That part was never optional. What you can still choose is whether you find out the results live, with a line at every register and trailers waiting at the dock. Put the four questions on this month’s calendar: capacity per site, true last-mile redundancy, automatic and tested failover, and a support path you have verified before you need it. Stores keep selling, DCs keep shipping, and the network stays out of the story.
If stores and distribution sites are your responsibility this peak season, see how Flō keeps POS, payments, and inventory online for multi-site retail.