From 2k to 47k TG members with 18% engagement: the playbook.
A DeFi protocol asked us to "make their community grow." We grew it 23× — but the more interesting number is engagement holding at 18% vs the 4–6% industry baseline. Here's exactly what we did, week by week.
Telegram community growth in crypto is mostly fake. The standard agency playbook — paid bot networks, paid airdrops to wallet farms, "engagement contests" that filter for nothing — produces big member counts and dead chats.
The protocol that came to us in early 2025 had been burned by exactly that pattern. They'd paid two prior agencies to "grow community" and ended up with 14,000 members and 280 daily messages, half of which were generic emojis. Most of the user count had no on-chain footprint with the protocol.
We started over with a 2,000-member core. Twelve weeks later: 47,000 members at 18% engagement rate. Below is the playbook — none of it is novel, but the discipline is.
Why most TG growth tactics produce dead members
The standard tactics — Telegram ads, paid follow campaigns, raid-style cross-promotions, bounty programs — all optimize for joins, not for retention. They produce:
- Bot accounts that ghost after the join
- Airdrop farmers waiting for the next bounty
- Real humans who joined for a contest and never engaged
- Self-promoting "community managers" looking for their next gig
What survives the first 30 days from this growth is typically 8–15% of the joiners. The rest sit as dead weight, dragging your engagement rate to 4–6% and giving you no actionable audience.
The 5-layer playbook
Layer 1 · Moderation script + bot filtering
Before adding a single new member, we rewrote the moderation stack:
- Combot configured for keyword/link filters tuned to common scam patterns
- Custom verification gate — on-chain check that the joining wallet has any interaction with the protocol's contracts (light, optional)
- Captcha + Q-style entry quiz (3 questions about the project)
- Auto-warning + escalation tree for spam patterns
- Manual review of ambassador and ranked-role applications
This dropped bot joins by ~95% from week 1. The trade-off: organic growth slowed temporarily because real humans also had to pass the gate. We accepted the friction.
Layer 2 · Ambassador structure
Most crypto ambassador programs are token-incentivized and attract token farmers. We rebuilt as a reputation system:
- Tier 1 — Contributor. Active in chat 14+ days. Helping new members. No token reward; reputation badge in chat.
- Tier 2 — Ambassador. Demonstrated technical knowledge of the protocol. Hosts beginner help threads. Modest monthly token allocation tied to verified contribution.
- Tier 3 — Lead Ambassador. Multi-region presence. Translates content. Hosts AMAs. Recurring stipend + visibility.
The key was that Tier 1 had no token reward. Reputation seekers, not yield farmers, filled the pipeline. By week 6 we had 40 active Tier-1s carrying most of the help conversations organically.
Layer 3 · AMA cadence
Two AMAs per week, structured differently:
- Tuesday — Technical AMA. Hosted by a protocol engineer. 30 minutes. Pre-submitted questions only. Focus on roadmap, integrations, on-chain mechanics. Recorded and posted to Mirror.
- Friday — Open AMA. Hosted by founder or community lead. 45 minutes. Live questions. Anything goes. More conversational.
What we noticed: Tuesday AMAs drove deeper engagement; Friday AMAs drove more joins. Both compounded. Week 8 onwards, AMAs were the single biggest source of new high-quality joins (people coming via shared clips from the Tuesday session).
Layer 4 · Contests that filter
Contests built around holders, not joiners:
- "Show your portfolio composition" — for verified protocol token holders only
- "Best protocol use-case writeup" — judged by ambassadors, prize in token allocation vesting over 30 days
- "Refer a holder" — referral reward only triggered when the referred wallet interacted with the protocol
Standard "tag 3 friends" engagement bait was banned. Same with vote-based contests where bot farms could swing outcomes.
Layer 5 · Founder presence
The founder showed up in chat 4–5 times per week — not as a marketing function but as a participant. Answered technical questions. Argued politely with critics. Posted half-formed ideas for community feedback.
This single behavior — founder being visibly present and accessible — is the highest-leverage thing a project can do for community quality, and the rarest. Most founders delegate community and never come back. The ones who don't see compounding loyalty.
The numbers, week by week
| Week | Members | Daily messages | Engagement rate |
|---|---|---|---|
| 0 | 2,000 | 180 | 9% |
| 2 | 2,800 | 340 | 12% |
| 4 | 5,200 | 820 | 15.8% |
| 6 | 9,400 | 1,640 | 17.4% |
| 8 | 16,800 | 2,910 | 17.3% |
| 10 | 27,200 | 4,750 | 17.5% |
| 12 | 47,000 | 8,500 | 18.1% |
The notable line is engagement rate. As size scaled 23×, engagement stayed in a tight 15–18% band. That's the signature of organic, anchored growth — not paid acquisition.
What we didn't do
The temptation in a community engagement is to optimize for vanity metrics. Specifically we refused:
- Telegram-promoted member campaigns (we tested briefly; quality was bad, killed at week 3)
- Cross-promotion swaps with other Telegram communities of unclear quality
- "Free token for every join" airdrops
- Outsourcing community management to non-native English speakers (the founder vetted every Tier-2+ ambassador personally)
Tools we used
- Combot — moderation, anti-spam, stats
- Custom verification bot — on-chain protocol interaction check (built by the protocol's dev team in week 2)
- Notion-based ambassador tracking — contribution log, reputation tier visible to all ambassadors
- Discord bridge — selective relay between TG and Discord for cross-platform consistency
- Mirror — long-form AMAs, ambassador writeups, founder posts archived
Founder being visibly present in community chat 4–5× per week. No agency can substitute for this. If your founder won't do it, your community will plateau regardless of the playbook quality.
If you've got a community that's stuck — high member count, low engagement, or growing but not converting into on-chain users — we'll audit it free. We'll show you what's drag, what's organic, and what you can change in the first 30 days.
We map your engagement quality and tell you what's salvageable. Most of the time, plenty.