The Influencer Marketing Math That Made Cal AI Profitable

Always ask influencers to cross-post. Offer the deal (e.g. '$400 for four posts'), then add 'just cross-post it on both platforms' (Instagram and TikTok) as if it were nothing. If they push back, downplay it: they're already making the video, so posting it twice costs them nothing. Making it sound like a big deal makes them treat it as one.

Set a base rate for a package (about $350-450 for four posts), then let creators unlock a couple of view bonuses so they are incentivized to go viral. Cap the bonuses, usually at the 1M-view tier, and pay nothing beyond it. Reasoning: most videos lose money, and the profit comes from the one or two a month that go viral, so you need to keep that upside; creators don't feel much difference between 1M and 2M views, a video that hits 1M often keeps growing to 1.3-5M, and very viral views reach a broad, less targeted audience that converts worse.

Ignore generic 'good CPM' benchmarks. For each influencer video, work out how much revenue it made relative to its views to get your RPM (revenue per thousand views, e.g. $5), then only pay creator rates that put your CPM under that number. Expect a lower affordable CPM than paid ads (no direct link, more friction), and lower still for UGC accounts with no audience trust. Apps with low order value and LTV (e.g. $30 / $50) can't match the CPMs that clothing or telehealth brands with far higher LTV pay.

Very early after launch, when there is little background noise, watch downloads against each influencer video going live: Cal AI had zero downloads on days with no video, and later a baseline of ~100 downloads an hour that jumped to ~200 when a video posted. Note which creator types cause spikes, what their comment sections look like, and the content format, then generalize those patterns ('this type of content converts like this; this kind of comment section is a good sign') to judge videos later, once millions of old videos keep getting views and per-video attribution is impossible.