Download the full playbook as a PDF: fifteen pages, no form, no email, no gate.

Where peak most often breaks

Foundations one to five decide whether the customer reaches your checkout and chooses to pay. Foundation six decides whether you can keep the promise you just made them. It is where peak most often breaks, because it breaks at a specific order volume, on a specific day, usually the one nobody modelled. The operators who survive it found their ceiling on a spreadsheet in September rather than on the warehouse floor in December. What follows is that spreadsheet, turned into checks.

Capacity, velocity and stock: find your ceiling, and check you have the product

The breaking point. Every fulfilment operation has a number: the daily order volume at which picking, packing, or dispatch can no longer keep pace with intake, and orders begin to age. Most retailers know their average day and hope. Capacity is two things, and retailers routinely measure only the first. Volume is how many orders you can hold and process. Velocity is how fast you move them out of the door. A warehouse that can hold thirty thousand orders but dispatch only four thousand a day has a four-thousand-a-day promise, and everything above that line is a delivery date slipping while the customer watches the tracking page.

Operator experience. I have seen the promise on the page quietly outrun what the floor could ship, and the gap only becomes visible when the tracking stops moving. Document the ceiling in September. None of the three fixes, more throughput, a longer operating window, or a softer promise, is available to you at nine on the busiest morning of the year.

Do you actually have the product. The most expensive campaign is the one shot for stock that never arrives. Before a pound of media is committed, the hero products behind it should be confirmed landed, in the warehouse, in sellable quantity, not on a purchase order with an optimistic ship date. Marketing spend paid against stock that does not physically exist manufactures demand you cannot fulfil, and pays for the disappointment twice.

Operator experience. I have had campaigns shot for products that sold out before the event, and for products that never arrived from the supplier at all. Confirm the stock is on the shelf before the shoot, not after the ad is live.

The customer test. The delivery date on the confirmation email is a promise made by your warehouse, not your marketing team, and it is only as real as the stock behind it. At your forecast peak volume, on the products you are advertising, can the operation keep it. If not, soften the promise or decline the sale.

Technology, traffic and the freeze: the stack has to hold at the volume you have not tested yet

Software paid and provisioned. Every platform, licence, plugin, and third-party service the operation depends on, confirmed paid, in-contract, and provisioned for peak load. The failure here is mundane and avoidable: a lapsed licence, a rate limit that was fine at normal volume and throttles at peak, a contract that auto-downgraded. None of it is dramatic. All of it is a revenue outage on the most expensive traffic day of the year. Confirm it now, in writing, service by service.

Traffic estimated and stress-tested. Put a real number on expected peak concurrent load, then prove the stack holds at it. An estimate nobody load-tested is a hope wearing a spreadsheet. Test at a multiple of your forecast, because the forecast is the floor of what could happen, not the ceiling. Speed is revenue at peak.

Code freeze booked. Agree the date past which nobody ships to production, and put it in diaries now. During peak, the most dangerous thing an engineering team can do is introduce a fresh change into a system already carrying its heaviest load of the year. The cost of a new bug is never higher, and the benefit of a new feature is never lower. Lock the code and run on what you have already tested.

The customer test. On the busiest hour of your year, the customer does not care why the site is slow or the payment page timed out. Every part of the stack not tested at peak load is a part you are asking the customer to test for you, live, with their basket full.

People, support and the promise: who is on call when it goes wrong at 6pm on Friday?

On-call cover and the danger window. A lot can go wrong between five on Black Friday evening and nine on Cyber Monday morning. A payment integration fails, a stock feed breaks, a delivery promise quietly becomes a lie, and the weekend rota has nobody watching. Document the on-call cover before peak: who holds the phone, which systems are monitored out of hours, what the escalation path is, and how fast a broken payment route actually gets fixed on a Saturday night. The weekend off is lovely. The unmonitored sixty-four hours across the biggest trading weekend of the year is where a good peak turns into a bad one.

Operator experience. The failures that hurt most are the ones that happen while the office is dark. Name the person, the number, and the escalation path in writing, and make sure they can actually reach the systems from home.

Customer service and marketing, staffed and locked. Support headcount scaled to forecast contact volume, not a normal week, with opening hours published and consistent between the site, the emails, and the phone line. The customer who cannot reach you during peak is a refund request forming, often with a chargeback behind it. Marketing executes a plan finalised before the freeze rather than building one live.

Cut-off dates clear and correct. The last order dates for guaranteed delivery, published, accurate, and consistent everywhere the customer reads them: product page, basket, checkout, confirmation email, and the carrier’s real capability. A customer who missed a cut-off but was told the truth is disappointed. A customer promised delivery that never had a chance of arriving is gone.

The customer test. At peak, the promise the customer believes is the sum of what marketing said, what the delivery dates implied, and whether a human answered when it went wrong. All three are set before peak begins.

On the checklist

  • Our daily breaking-point order volume is documented.
  • Forecast peak day sits below that line, or the promise is softened.
  • Hero products behind campaigns are confirmed landed and sellable.
  • Dispatch velocity, not just storage volume, is measured.
  • Every licence and third-party service is paid and provisioned for peak.
  • Peak concurrent load is estimated and load-tested at a multiple.
  • The code freeze date is agreed and in diaries.
  • On-call cover named for the full peak weekend, with escalation path.
  • Out-of-hours system monitoring is in place and reachable from home.
  • Customer service staffed to peak volume, opening hours published.
  • Marketing plan and creative locked before the freeze.
  • Delivery cut-off dates clear, correct and consistent everywhere.

Before peak arrives

This playbook is the front-end read a house can run on itself. The Barking Cat Review takes the whole house apart, shop front and back office, finds the binding constraint, and puts commercial numbers against it before peak arrives.

Back to Bulletproof for the Golden Quarter