A playbook for tuning your verification gate without killing growth

A playbook for tuning your verification gate without killing growth

Every verification step is a tradeoff: fewer bots, fewer humans. The goal isn't zero friction or maximum friction — it's the point where marginal humans lost exceeds marginal bots blocked. Here's how to find it.

Step 1: baseline both costs
— Count gate abandonment: members who join but never clear verification (Discord's member screening logs or a reaction-role audit)
— Count bots that still slip through per week

Step 2: classify your threat
Reaction-role gates stop almost no determined spam but cost ~5-15% of legitimate joiners who never click. CAPTCHA gates stop scripted raids but barely dent human-driven spam. Match the gate to what's actually attacking you, not the worst case.

Step 3: stage it
— Tier 1: light gate always on (one reaction)
— Tier 2: CAPTCHA auto-triggered only when join velocity spikes above your normal baseline

This is the key move — most damage comes in raid bursts, so escalate by velocity rather than running heavy friction 24/7.

Discord vs Telegram
Telegram's join requests + a single question often outperform CAPTCHAs because the human cost is lower and the bot cost is similar. Discord's onboarding questions can do the same if you stop treating them as rule-reading.

The caveat
Abandonment data can't distinguish bored humans from blocked bots cleanly. Sample manually — read 30 abandoned profiles before concluding.

Open question: is velocity-triggered friction visible enough to deter raiders who probe first?