Most startups are good at starting channels and bad at stopping them. A founder tries paid social, a newsletter sponsorship or a round of cold email, the results come back unclear, and the channel keeps running because nobody decided what clear would look like. Months later it is still on the card statement, defended by the person who set it up. The fix is not better judgment at the end. It is a rule written at the start.
Write the stop rule before the first dollar goes out.
Before a channel starts, the decision to end it is easy, because nothing has been spent and nobody has a stake in it. After the spend starts, every review becomes a defense of what has already gone out. The founder is no longer asking whether the channel works. They are asking whether they were wrong, and few people answer that question honestly about their own money.
So the rule comes first, on paper, agreed by whoever controls the budget. It has three parts: what is measured, for how long, and who decides. If any of the three is missing, the channel will outlive its usefulness.
Measure cost per qualified conversation.
Clicks, impressions, followers and open rates all move, and none of them pays the bills. For most early startups the measure that matters is the cost of a qualified conversation: a call, a demo or a reply from someone who fits the buyer the company has named. It is close enough to revenue to mean something, and early enough that a small test can produce a few of them.
Define qualified before the test, not after. If the definition is written once the results are in, it will drift toward whatever the channel produced. The company already needs the number its marketing is accountable to. The stop rule borrows that number for one channel.
Then set a ceiling. The ceiling is what the company can afford to pay for one of those conversations, given what a customer is worth and how many conversations it takes to close one. It does not need to be precise. It needs to be written down.
Fix the window and the spend.
A channel needs enough time and money to give an answer, and no more. Too short and the test ends before the channel has a fair chance. Too long and a failing channel becomes a habit.
Pick a window that matches how the channel works. Paid search can show something in weeks. Content and search optimization take months, and a stop rule that expects them to perform in a month will kill them unfairly. Whatever the window, set the spend that goes with it and do not add to it mid-test because the early numbers look close. Close is the most expensive result a channel can produce, because it always earns one more month.
Name the one person who decides.
When everyone reviews a channel, nobody ends it. A committee will find a reason to keep almost anything: the brand value, the learning, the possibility that next month is different. One person, named in advance, reads the numbers at the end of the window and makes the call against the rule.
That person should not be whoever runs the channel. An agency, a contractor or an employee who owns a channel has an honest reason to want it to continue. Their view matters, and they should present the case. The decision belongs to someone who is not paid by the channel's survival.
Stop, change one thing, or keep it.
At the end of the window there are three honest outcomes. The channel beat the ceiling, so it keeps its budget and gets a new window at a higher spend. The channel missed the ceiling by a wide margin, so it stops, and the money moves to the next channel on the list. Or the channel came close, and one specific change has a plausible reason to close the gap: a different audience, a different offer, a different page.
The third outcome is allowed once. Change one thing, run one more window with the same rule, and accept the answer. If the second window is also close, the channel stops. A startup does not have the runway to keep paying for almost.
This is why startups should run one channel at a time. With one channel live, the stop rule is easy to read. With five, every result is tangled with every other. The order to try them in is in Marketing channels a startup should try, in order.
Stopping a channel is not a failure. It is the test doing its job, and the money it frees up pays for the next one.
Write the measure, the window, the spend and the decider on one page before the next channel starts.
Tell us what you're building.