When should you kill a marketing channel?

Short answer: Kill a channel on a pre-registered rule: a fixed number of units, a minimum result, and a date. If the rule is met, keep going. If not, stop on the date - do not extend it after the fact.

Channels rarely fail loudly. They get killed in week one, when the numbers are small and scary, or kept alive for a year because "it is starting to work".

The kill rule, written before you start

One sentence with four parts:

If <channel> produces fewer than <N results>
after <M units> by <date>, we stop.

Example: "If LinkedIn outbound produces fewer than 3 replies after 100 sends by 15 October, we stop." That is a decision made while you are still calm, which is the only time decisions about volume are honest.

Signals that mean stop, all of them cheap

• Deliverability floor: bounce rate above ~3% or spam complaints - stop and fix the list, not the copy.
• Zero positive replies at volume: 100+ sends, no positive reply - the offer or the list is wrong.
• Cost per result above gross margin: the channel can never pay for itself; more volume makes it worse.
• You cannot produce the volume. If the channel needs 300 sends a month and you can send 40, you are not testing it.

Signals that mean keep going (and stop touching it)

• Positive replies exist and the rate is inside the 1-3% normal band.
• The cost per result is under your margin and the trend is flat or better.
• Results are consistent across two consecutive periods, not one lucky week.

The failure mode to avoid

Extending the deadline after a bad result converts a test into a subscription. Pick the date, then treat the date as the decision - the alternative is spending the next quarter arguing with your own data.

Source

The rules above come from marketing-mindset - an open-source (MIT) skill that gives AI agents decision rules for marketing instead of templates. Install with npx marketing-mindset or read the full skill file on the landing page.