Most brands plan for peak as a traffic problem. Server capacity, inventory buffers, ad budget, discount architecture. The support plan, if it exists in writing at all, is usually one line that says all hands on deck.

Then the weekend arrives and the support inbox does something the traffic graph does not. It keeps climbing after the sale ends.

The cost of an understaffed peak is not the backlog. A backlog clears. The cost is what the backlog does to four numbers you will be reading in January: your refund rate, your chargeback rate, your repeat purchase rate, and the return on everything you spent to acquire that cohort in the first place.

This is a planning piece, not a pitch. It covers where peak support load actually comes from, what it charges you when it is not covered, and what the load looks like phase by phase so you can staff against it instead of reacting to it.

01Peak is a spike, not a busier week

The scale is easy to underestimate, because the headline figures get reported annually and peak is not an annual event. It is an hourly one.

Shopify merchants alone generated $14.6 billion across Black Friday Cyber Monday in 2025, a 27 percent increase year over year, from more than 81 million buyers. At the busiest minute of the weekend, sales ran at $5.1 million per minute.

That last number is the one worth sitting with. Peak is not a busier week. It is a quarter’s worth of order volume compressed into a window measured in hours, and every one of those orders carries a probability of generating a support contact.

The distinction changes how you staff. A busier week can be absorbed by asking your existing team to work harder. A spike cannot, because the constraint is not effort. It is simultaneity.

Twenty customers asking the same question across a week is a workload. Twenty asking it in the same minute is a queue, and a queue behaves differently.

02Where the load actually comes from

Peak support volume does not arrive evenly across question types. It concentrates, and it concentrates predictably.

Up to 30 percent of incoming e-commerce support tickets are shipping status requests, the category the industry shortens to WISMO, for where is my order. During peak that share climbs, because the gap between promise and delivery widens. Carriers congest, warehouses fall behind, and the customer who bought on Friday starts checking on Sunday.

The rest of the load clusters into a short list:

• Pre-purchase questions during the sale window. Sizing, compatibility, stock, delivery cutoffs. These are the highest value tickets you will receive all year, and the ones most likely to go unanswered, because they arrive at the same moment as everything else.
• Discount code failures. Stacking rules, expired codes, exclusions nobody read. Every one is a customer who wanted to buy and currently cannot.
• Address changes and cancellations. Time boxed by your fulfilment cutoff. Miss the window and a solvable ticket becomes a return.
• Duplicate orders from checkout errors during traffic surges.
• Then, after the weekend: delivery exceptions, damages, and the returns wave.

Notice what most of these have in common. They are not complaints. They are revenue events that have not closed yet, and they expire.

03The four line items an understaffed peak charges you

1. 01

   Revenue that never closed

   Pre-purchase questions during a 72 hour promotion have a shelf life measured in minutes. Best in class first response time is under one minute on live chat and under one hour on email. A brand running a 12 hour email response time during peak is not delivering slow service. It is delivering no service, because the sale ended before the answer arrived.

   These tickets do not turn into bad reviews. They turn into nothing, which is exactly why they never show up in a post mortem.

2. 02

   Preventable refunds and returns

   Address corrections, wrong variant catches and cancellation requests are cheap to resolve inside the fulfilment window and expensive after it. Once the parcel ships, a two minute reply becomes a return: reverse shipping, restocking, inspection, and often a unit that cannot be resold at full price.

   The NRF puts the online return rate at 19.3 percent of sales, with roughly 17 percent of holiday sales coming back. Your baseline is already high. Unanswered tickets during peak week simply move volume from the resolvable side of that number to the expensive side.

3. 03

   The cohort you burned

   This is the largest line, and the one that never appears on a Q4 profit and loss statement.

   The NRF found that 71 percent of consumers are less likely to shop with a retailer again after a poor experience, and four out of five will tell friends and family about it. For most brands, Black Friday is the single largest intake of first time buyers in the year. Those customers have no prior relationship with you to draw on. Their entire impression is formed by one transaction, and if the only human interaction in it was a ticket nobody answered, that is the impression.

4. 04

   Acquisition spend you have already committed

   Peak CPMs are the highest of the year. You paid a premium to buy that cohort. If a meaningful share of it does not return, your effective cost per retained customer is far higher than the number showing in your ads dashboard, and you will not find out until the second quarter when repeat rate lands flat.

A worked example, so the shape is visible

Illustrative only. Every assumption is named so you can replace it with your own.

05Why hiring in November does not fix it

The instinct, when the forecast looks heavy, is to add people close to the date. It rarely works, for three reasons that have nothing to do with the quality of the person hired.

Ramp time is real

An operator who does not know your products, your policy edge cases, your fulfilment cutoffs and your escalation thresholds is not neutral during peak. They generate rework. They escalate what they should resolve and resolve what they should escalate. Under normal volume, a senior teammate absorbs that. During peak, that senior teammate is the exact bottleneck you were trying to relieve.

Context has to exist before it can be transferred

This is the step most brands skip. An operator is only as effective as the documented brand knowledge behind them: tone, policy boundaries, the exceptions you make and the ones you never make. If that lives in a founder’s head, headcount does not help. It just adds another person waiting on the founder.

Quality drifts fastest when volume is highest

Without someone reviewing responses during the spike, drift is discovered in January, in the CSAT numbers and the refund rate, long after it can be corrected.

Capacity added eight weeks out arrives trained. Capacity added in week one of peak arrives as work.

None of this is an argument against adding capacity. It is an argument about timing.

If the question you are weighing is the underlying staffing model rather than the timing, we broke those numbers down separately in in-house versus outsourced customer service for DTC brands. And for how a well run peak looks day to day, what brands with a seamless Black Friday do differently.

06What a prepared peak actually looks like

Not a bigger team. A covered one.

• Coverage mapped to your actual order curve, not to office hours.
• Macros and help centre content updated against this season’s SKUs, promo rules and delivery cutoffs before the traffic arrives, not during it.
• A WISMO deflection path: proactive shipping notifications and a self serve order lookup, so your largest ticket category never becomes a ticket.
• Escalation thresholds written down, so nobody has to decide in the moment who owns a carrier failure.
• A named owner for the returns wave, resourced in December for work that lands in January.
• Someone reviewing a sample of responses daily through peak week, while it can still be corrected.

07Common questions