Climbers on Everest pick a turnaround time before they leave base camp. If they haven't reached the summit by, say, 1pm, they turn around, however close the top looks. Annie Duke, the former poker player who wrote Quit, uses that example in an interview with David Epstein to make a point about deciding when to stop: the decision has to be made at base camp, because near the summit, tired and invested, nobody turns back.
Marketing campaigns have a base camp too. It's the week before launch, when you're writing the brief, setting the budget and choosing the target. Everyone agrees on what success looks like. Hardly anyone writes down what failure looks like, or when they'll call it.
So the campaign launches, the numbers come in a bit soft, and the conversation goes the way it always goes. Give it another two weeks. Let's try new creatives first. The algorithm is still learning. Each of those can be true. Without a line drawn in advance, there's no way to tell which one you're hearing.
01Stopping is the decision we're worst at
Starting a campaign is fun. There's a brief, a launch date, new creatives. Stopping one means admitting the money already spent didn't buy what we hoped, and that someone's idea, often your own, didn't work. Duke puts it simply: "Once you put something into something — lives, money, effort, time — it becomes very hard to let those things go." She adds that knowing about this bias doesn't protect you from it.
There's decent evidence that most of us wait too long. In 2016, economist Steven Levitt ran a study where people stuck on a big decision, like quitting a job, flipped a coin to decide. Those who made the change reported being substantially happier two and six months later. If changing course tends to feel better than expected, we're probably holding on longer than we should.
Marketing adds its own excuses on top. Ad platforms have a learning phase, so early results are noisy by design. Creative fatigue, seasonality and tracking gaps can all explain a bad week. Each excuse is sometimes right, which is exactly why it works so well as a reason to keep spending.
The way out, according to the research Duke cites by Barry Staw, is to commit in advance to the signals that would make you walk away. Decide while you're still at base camp, and the later decision becomes one you've already made.
02The stop sheet
Fill this in before any campaign goes live. It takes about ten minutes, and it's the most useful ten minutes of the launch.
Three parts carry most of the weight.
The stop line needs a date. Duke is precise about this: a kill criterion is a state of the world and a date. A line like "if cost per lead stays too high" will never actually get crossed, because there's always a reason it isn't quite too high yet. "Cost per lead above $120 on May 30" leaves nobody any room.
The gray zone keeps you from killing too early. Rajesh Dutta describes it on Medium for product teams: a success target and a lower kill threshold, with room in between to iterate. A campaign that misses its ambition but stays above the stop line has earned more tries. It just doesn't get unlimited ones.
Section 6 makes stopping easier. Stopping is much easier when the money has somewhere better to go. Name that place in advance and "stop" turns into "move."
03Pick the numbers from your economics
The stop line comes from what a customer is worth to you, worked backwards through your funnel.
Start with the most you can pay to win one customer and still make money. Then work backwards through your funnel. Say a new customer brings in enough margin that you can afford to spend $600 to win them, and about one lead in five becomes a customer. Your break-even cost per lead is $600 ÷ 5 = $120. That's your stop line, and it's why the example in the sheet says $120.
The minimum volume is the other half. Ten leads make an anecdote. Pick a volume where one extra good or bad lead wouldn't flip your conclusion, and don't judge the campaign before it gets there. If you're unsure how much data it takes to tell a real difference from noise, the math in my piece on A/B test sample sizes applies here too.
Two more rules keep you honest:
- When the real outcome is slow, set the line on an earlier step. If deals take three months to close, you can't wait for revenue. Put the stop line on qualified leads or booked meetings, and check later that they turned into customers.
- Write the line before you see any data, and don't move it mid-flight. If something outside the campaign changes (a price change, a tracking break, a competitor's launch), you can redraw it, but write down why. A line you move every time you get close to it isn't doing any work.
04On the day you cross the line
This is the moment the sheet was written for, and it will feel wrong. The latest numbers will look slightly better, someone will have a new creative idea, and the platform rep will mention the learning phase.
Stop anyway. Then take twenty minutes to write down what the campaign taught you: about the audience, the message, the channel. A stopped campaign with three clear lessons is worth more than a running one nobody can explain. Move the budget where section 6 said it would go, and start the next sheet.
If you hit the success line instead, the sheet still helps. It tells you the campaign earned more budget, and it reminds you to watch what the next dollars cost, since the first ones are usually the cheapest. I wrote more about that in the one-channel trap.
Duke has a poker statistic that stays with me: professional players play roughly 15 to 25% of the two-card hands they're dealt, while amateurs play more than half. Most of the skill is in the hands they fold.
The climbers who turn around at 1pm are the ones who get another go next season. Same goes for your budget.






