Four ways a kill-rule system quietly stops working

The symptom

A kill-rule system exists on paper, with a stated threshold and an age gate, and still seems to be killing good creative, keeping bad creative alive, or producing decisions that don't hold up when checked later.

The four ways this breaks

• Applying the age floor to the whole campaign instead of each individual ad, so a brand-new creative inherits an old ad set's clock and gets judged before it's had its own seventy-two hours.
• Killing on noise: visible when the median spend at the moment of killing sits well below the target multiple, which means the rule is being overridden by impatience, not actually being followed.
• Mistaking a learning reset for fatigue, when the real cause of a sudden drop is a budget or targeting change made inside the last seventy-two hours, which restarts the platform's learning phase and looks like decay.
• Pricing the kill threshold off lifetime value instead of day-one value, which produces a ceiling the business can't actually fund out of this week's cash.

The fix

Apply the age gate per ad, not per campaign. Record the spend multiple at the moment of every kill and check the distribution monthly; a median below roughly two times target means the rule is being overridden, not followed. Before reading any sudden drop as fatigue, check whether a budget change, a targeting change, or a conversion-event change happened in the last seventy-two hours. Build the kill threshold from trailing day-one order value, never lifetime value, so the number reflects what this week's cash can actually support.

How to spot it

Pull your last month of kill decisions and check the spend-at-kill multiple against your stated target for each one. If the median sits below roughly two times target, the stated rule and the actual behaviour have already diverged. Check your newest ad's age against the campaign's launch date, not its own; a mismatch here means the age gate is being applied at the wrong level.