The Core Problem
DeFi projects face a unique marketing challenge: their users live in Telegram groups, not on websites. Traditional marketing channels — email newsletters, SEO blog posts, paid ads — underperform because crypto traders and degens congregate in Telegram communities where they expect real-time, interactive experiences.
The problem compounds when you operate multiple DeFi products: a trading bot, a slot game, a token launch service. Each product needs its own Telegram bot with its own onboarding flow, user preferences, and engagement cadence. Managing these independently creates silos where a user who loves the slot game never learns about the trading bot.
DeFiKit's BotMatrix solves this with a serverless orchestration layer that connects all Telegram bots through a shared message queue, enabling automated cross-promotion, unified user profiles, and event-driven marketing sequences.
The Solution: BotMatrix Orchestration Layer
BotMatrix is a NestJS monolith built on grammY and Prisma that sits above individual bot instances. Instead of each bot managing its own user database, every interaction flows through BotMatrix:
```
User → Bot A → SQS Queue → BotMatrix Processor → Database → Bot B Notification
```
This design enables four critical marketing automation capabilities:
1. **Unified user profiles**: A single user across multiple bots gets consolidated preferences and behavior history
2. **Event-driven cross-promotion**: When a user completes an action in Bot A, BotMatrix schedules a Bot B invite
3. **Automated job scheduling**: Time-based marketing sequences (welcome, re-engagement, referral prompts)
4. **Behavioral segmentation**: Users are tagged by activity level, token holdings, and response rates
Architecture Overview
The SQS Ingestion Layer
Every bot interaction is serialized as a JSON event and pushed to an AWS SQS queue. This decouples user-facing bots from the marketing processor:
```typescript
interface BotEvent {
userId: string;
botId: string;
eventType: 'command' | 'message' | 'game_result' | 'referral';
timestamp: Date;
data: Record<string, unknown>;
}
```
Because SQS handles buffering, the bots never block on database writes. Marketing sequences can process events asynchronously without affecting the user experience.
The Prisma Database Schema
BotMatrix uses a relational schema that connects users to bots, groups, commands, and jobs:
```prisma
model UserPrivilege {
userId String
botId String
role String // 'admin', 'user', 'vip'
createdAt DateTime @default(now())
@@id([userId, botId])
}
model Job {
id String @id @default(cuid())
botId String
userId String?
type String // 'welcome', 're_engagement', 'cross_promotion'
status String // 'pending', 'running', 'completed'
runAt DateTime
createdAt DateTime @default(now())
}
```
The Agent System
BotMatrix includes an agent abstraction for complex automation flows. Each agent has a purpose and can spawn sub-agents for parallel tasks:
```typescript
interface Agent {
id: string;
name: string;
purpose: string; // 'cross_promotion', 'retention', 'onboarding'
status: string;
}
```
For example, the cross-promotion agent monitors game completion events. When a user finishes 5 rounds in the slot game, it triggers an onboarding sequence for the trading bot — complete with a personalized message about how trading signals complement their gaming rewards.
Step 1: Automated User Onboarding
When a new user joins any DeFiKit bot, BotMatrix orchestrates a multi-step onboarding sequence:
```
T+0: Welcome message + feature overview
T+5min: Interactive demo (inline keyboard with test commands)
T+1hr: Value proposition for other DeFiKit products
T+24hr: Referral program invitation
T+72hr: Re-engagement if inactive
```
Each step is a configurable Job in the database. The sequence adapts based on which bot the user joined through — slot game users get different messaging than trading bot users.
Step 2: Cross-Promotion Automation
Cross-promotion is the most direct marketing automation win. DeFiKit's event matching rules connect user actions across bots:
```sql
-- When a trading signal generates a profitable trade
-- Send a promotional message for the slot game
INSERT INTO jobs (bot_id, user_id, type, run_at, status)
VALUES ('casino-slot-bot', @userId, 'cross_promotion',
DATEADD('hour', 1, NOW()), 'pending');
```
The 1-hour delay ensures the user isn't spammed immediately. The message itself is templated and personalized:
```
🎰 *Nice trade, {username}!*
While you wait for the next signal, why not try your luck?
Our slot game paid out 2.3 SOL in rewards this week alone.
👉 /play_casino
```
Step 3: Group-Level Topic Routing
BotMatrix supports Telegram supergroup topics for organized community management. Each product channel (trading signals, game updates, token alerts) gets its own topic within a master group:
```prisma
model TelegramGroupTopic {
groupId String
topicId String
name String
product String // 'trading', 'gaming', 'tokens'
}
```
This structure means a single Telegram group serves as the hub for all DeFiKit products, with automated topic assignment based on user activity.
Step 4: Serverless Deployment
The entire BotMatrix stack runs on AWS with minimal operational overhead:
| Component | Service | Cost |
|-----------|---------|------|
| Bot instances | EC2 t3.micro (via SQS listeners) | ~$8/mo |
| Message queue | SQS (1M requests free tier) | ~$0 |
| Database | RDS PostgreSQL db.t3.micro | ~$15/mo |
| User profiles | Prisma + PostgreSQL | Included |
| Total infrastructure | — | **~$25/mo** |
The serverless SQS layer ensures that even if all 6+ bots are hammered simultaneously during a market event, messages queue up and process in order without dropping data.
Results
DeFiKit's automated marketing pipeline shows measurable improvements in user progression:
| Metric | Before BotMatrix | After BotMatrix |
|--------|-----------------|-----------------|
| Cross-product adoption rate | 8% | 34% |
| 7-day user retention | 22% | 51% |
| Average onboarding completion | 40% | 78% |
| Automated promotions sent/day | 0 (manual) | 200+ |
| Cost per new user acquired | ~$0.50 (ads) | $0.02 (cross-promo) |
Key Takeaways
- **Serverless queues are the backbone of multi-bot marketing**: SQS decouples user-facing latency from marketing logic, letting both scale independently
- **Cross-promotion requires timing, not frequency**: A well-timed single message (1-hour delay after a positive experience) outperforms 5 immediate spam messages
- **Unified user profiles unlock behavioral segmentation**: Without a shared database, you can't tell that a slot game user is also a trading bot user
- **Telegram supergroup topics enable scalable community management**: One master group with topic-based routing replaces 6 fragmented groups
- **The 8% → 34% cross-product adoption lift validates the architecture**: Users don't discover secondary products naturally — they need automated, contextual invitations at the right moment