When to turn paid on, and when to turn it off

turn paid on when three things are true: a retained cohort exists, organic has proven a message, and you know the CAC your economics can afford. turn it off when CAC drifts past payback with no creative answer in testing, or when the account has stopped scaling and started learning on your money. paid is a faucet. the skill is being unsentimental with the handle.

The on switch: three conditions, all of them

paid goes on when retention, message and math are all in place, and two out of three is a no.

1. a retained cohort exists. some group of users comes back without being reminded, whatever retention shape your category supports. paid pours users into the top of your product, and if the bucket leaks everything, spend just makes the leak more expensive. this is the honest pre-PMF answer, and it is why reddit and TikTok come before paid ads for early consumer AI apps.
2. organic proved a message. something specific already worked without media spend behind it: a hook, an angle, a demo that pulled real engagement from strangers. on ZuAI, paid earned its budget by scaling what TikTok UGC and reddit seeding had already proven, and that order was the entire trick. paid amplifies evidence. it does not manufacture it.
3. the CAC target is known. you have worked out what you can afford to pay from revenue per user and payback window, before the first campaign, so results land against a number instead of a feeling.

the common mistake is treating paid as the way to get the first two conditions rather than the reward for having them. spend can compress time once direction exists. it cannot buy direction.

The off switch: two signals that matter

switch off when CAC outruns payback with no answer in testing, or when spend has stopped scaling and started exploring. the first signal is a drift, not a bad day: blended CAC climbing past your payback threshold for a couple of readouts in a row, while nothing in the creative testing pipeline looks like the ad that reverses it. rising CAC with a promising new batch mid-test is normal fatigue turnover. rising CAC with an empty pipeline is the channel telling you it needs a pause, not a bigger budget.

the second signal is subtler: learning has replaced scaling. every readout ends in a new theory to test instead of a winner to feed, results swing week to week, and the account is functionally running experiments at scale prices. testing is supposed to be the small disciplined part of a working account. when it has quietly become the whole account, you are paying auction rates for lessons organic teaches free, and the honest move is to go get the answer cheaply and come back.

Pausing without panic

a pause is a decision, not a defeat, and the discipline is to run it like one. pause fully rather than trickling a token budget that keeps you emotionally invested while teaching nothing. write down, on the day you pause, what would turn the channel back on: a new proven message, a retention fix shipped and measured, a CAC target that changed because pricing did. then spend the paused budget’s attention where the actual blocker is, usually creative supply or retention, not the ad account. what makes this hard is that paid stops the moment spend stops, installs drop the same week and the graph looks like going backwards. it is not. it is the difference between a channel that was working and one that was expensively pretending to, made visible. founders who cannot pause paid end up run by it, defending last month’s spend with this month’s, and the sunk cost compounds quietly. the ones who pause cleanly come back to a channel that works, because they only turn it on when the three conditions hold.

the on switch and the off switch are the same skill: knowing what paid is for and refusing to let it be anything else. the full system is in the paid ads playbook. if you are staring at the handle unsure which way to turn it, the breakdown call is free. Let’s talk Growth.