The eleven minutes
Our restock drop sold out in eleven minutes. That should be a good story. Here is what actually happened inside those eleven minutes, and the two hours after.
Our ads kept running for another 118 minutes on a product that no longer existed. Nobody was watching stock and spend at the same time, because those were two people looking at two dashboards in two rooms.
Support answered fourteen tickets promising a dispatch date the warehouse had already told ops was impossible, but ops was in the room with the founder, and support was in a different channel.
And a customer who had spent four thousand euros with us over two years received an automated apology written for a first-time buyer, because nobody flagged her order as anything other than an order.
Every person did their job correctly. The information just never crossed the gaps between them. That is the actual failure mode of a flash sale.
Decide two things before the sale starts
Most of the chaos on the day comes from decisions being made under time pressure by whoever happens to be looking. Make these two in advance, in writing, where everyone can see them.
- What is the stock floor, and who is allowed to breach it? Pick a number of units at which you stop selling a variant. Decide now whether oversell is acceptable, and if so, by how much and who can authorise it. On the day, this becomes a yes-or-no question instead of a debate.
- What is the promise support is allowed to make? Agree the dispatch window with the warehouse before the sale, not during it. Write the exact sentence support should use. Fourteen tickets with the same wrong promise is not a support failure. It is a briefing failure.
Pin them at the top of whatever channel you are running the sale from. If a decision is not visible to everyone in the room, half the team will make a different one.
Who watches what
The mistake is assigning people to areas (ops, support, warehouse) and assuming coverage. Assign people to numbers instead, and say the number out loud on a cadence.
- Stock per variant. One person. Their only job is to call out when any variant crosses the floor. They do not answer tickets.
- Spend and traffic. One person, watching cost-per-acquisition and which creative is pointing at which SKU.
- Inbound volume. One person watching the ticket queue depth, not the tickets themselves. Depth is the early warning that your promise is wrong.
- The floor. Someone in the warehouse posting what is actually happening on the pick line, in a sentence, every thirty minutes.
Four numbers, four names, one room. If your team is three people, one person holds two numbers, but they still say both out loud.
The pause trigger
The single highest-value rule we have: spend pauses automatically at the stock floor, and it is nobody’s judgement call.
Pausing ads on a sold-out SKU seems obvious in retrospect and never happens in the moment, because the person watching stock is not the person with access to the ad account. Fix that in advance:
- Give the stock-watcher the ability to pause, or put the ads person in the room with an agreed trigger phrase.
- Write the trigger as a number, not a feeling:
Sand/M under 10 → pause set 4. - Time how long it takes from call-out to paused. If it is more than two minutes, the wrong person has the permissions.
The hour after is the sale
The revenue is banked in eleven minutes. The margin is decided in the next three days: in refunds, in replacement shipping, in the support hours spent on delivery-date questions you created yourself.
So the room does not close when the product sells out. Give it one more hour and do three things:
- Reconcile what you actually sold against what you can actually ship, per variant, before anyone goes home.
- Flag the exceptions now: oversells, high-value customers, anything going to an address the carrier has failed at before. These are cheap to handle today and expensive on Thursday.
- Write the recap while it is fresh. What sold, what ran out first, what the queue depth peaked at, which promise turned out to be wrong. Fifteen minutes now saves you rebuilding the plan from memory next quarter.
The short version
- Decide the stock floor and the support promise before the sale, in writing, pinned.
- Assign four numbers to four names. Say them out loud on a cadence.
- Make the pause trigger automatic and permission it to the person watching stock.
- Keep the room open for an hour after sell-out. Reconcile, flag exceptions, write the recap.
- Do it in one place where all four people can see all four numbers. This is the whole trick.
We built Store Huddle because that last point kept being the hard part. Everything else on this list is free. It just does not survive a group chat.
Deciding the triggers before you start
A flash sale compresses every decision into a few hours, which is exactly when nobody should be deciding anything from first principles.
Agree three numbers in advance and write them down with names attached. At what stock level a product comes out of the email and off the ads. At what point you stop promising delivery dates you may not hit. At what queue length you switch customer messages to a holding reply rather than answering individually.
Each of these will be needed at some point during the sale, and each is a bad decision to make at speed. Deciding them calmly beforehand costs fifteen minutes and removes most of the improvisation.
Who says what, in what order
The comms failure in a flash sale is rarely the customer-facing message. It is that internal information moves more slowly than the sale does.
When something sells out, three things have to happen, usually by three different people, and they are not equally urgent. Whoever runs paid advertising pauses within minutes, because ad spend continues at full rate while the page sells nothing. Whoever handles merchandising changes the site within the hour. Whoever answers customers needs to know before the next message arrives, not after.
Write those three roles down beforehand. During the sale nobody will work out the sequence from scratch, and the default is that everyone waits for someone else to notice.
What to tell customers when it goes wrong
Two situations account for most flash-sale complaints, and both are survivable with the right message.
Sold out faster than expected. Say so plainly and quickly. Do not leave a page technically purchasable that will not ship, which converts a disappointment into a refund plus a complaint. If a restock is genuinely coming, say when; if it is not, say that too.
Dispatch is slower than promised. Message before the promised date passes rather than after. A customer who hears from you on the morning of the date is dealing with a business on top of it; the same customer chasing two days later is not, on identical facts.
The debrief
Half an hour in the following week, while it is fresh.
Three questions. Which of your pre-agreed triggers fired, and did any fire too late to act on? What did you find out later than you should have? What would you decide differently before the next one?
The second question usually points at a monitoring gap rather than a planning gap, and those are the cheapest things to fix. A stock level that was true at eleven in the morning and reached the team at four in the afternoon is a routing problem, not a forecasting one.
Before the sale: the three-line brief
Everyone involved should be able to answer three questions without looking anything up. If they cannot, the brief was too long and nobody read it.
What is on offer, and until when. Including whether it ends at a time or when stock runs out, which are different promises and get confused constantly.
What to say when something sells out. Written in advance, in the words you actually want used, so nobody improvises a restock promise you cannot keep.
Who to tell, and how, when something changes. One named place, not "let someone know". During the busiest hours of the year, a message posted somewhere general will not be seen.
The mistake that costs the most
Not stock, and not the site. It is continuing to pay for traffic to something that can no longer convert.
Advertising keeps spending at full rate whether or not the page can fulfil an order. A product that sold out at eleven in the morning and stayed in the ad set until the end of the day has quietly spent a meaningful share of the sale's margin sending people to a dead end, and it produces disappointed customers at the same time.
Whoever can pause spend should be a named person, contactable, with the authority to act without checking. This is worth deciding explicitly, because it is the one decision where a delay of hours has a directly measurable cost.
Common questions
What should be decided before a flash sale starts?
The triggers: at what stock level a product leaves the email and the ads, when you stop promising delivery dates, and when customer replies switch to a holding message.
Who needs to know first when something sells out?
Whoever can pause the advertising, because spend continues at full rate against a page that cannot convert. Merchandising and customer support follow within the hour.
Should a sold-out product page stay live?
Usually yes, provided it is honest and offers an alternative, since it keeps traffic from emails already sent. What causes damage is a page that is purchasable but will not ship.
How do we handle dispatch delays during a sale?
Message before the promised date passes, give a conservative new date you can beat, and state what happens if the customer does nothing.
What should the debrief cover?
Which triggers fired and whether any fired too late, and what you learned later than you should have. The second usually reveals a routing problem rather than a planning one.