How OutCheese got a 30% reply rate in cold Telegram outreach for a user acquisition platform
Client Snapshot
Traffic & user acquisition platform
AudienceMart helps crypto and Web3 projects get users and scale traffic predictably - through Telegram channels, mini apps, bots, viral short-form videos, and influencers
Goal
Validate Telegram outreach as a lead generation channel and to road-test our own Telegram outreach service
Pain Points
Most of the target audience lives in Telegram, scattered across dozens of niche chats and communities. Traffic requests rarely sit on the surface: people write thousands of messages, only a handful turn out to be a signal, and sometimes the only hint of a need is hidden in the bio
Some leads sit in private chats with no visible member list at all
You could scroll through the chat manually, check every bio, and catch potential clients that way - but that would eat an enormous amount of time
Results
The pilot ran for one month. We parsed 5 target chats with a combined audience of about 7,000 people, ran every member through a prompt-based filter against AudienceMart's ICP, and flagged the signal - fit or not. For each target lead we built a unique custom variable based on the signal from their own messages
~7,000 people across 5 target chats - parsed and filtered
312 target leads matching the ICP
30% reply rate vs market averages of ~6-8% on LinkedIn and ~1% in cold email (roughly 4x LinkedIn and ~30x email)
11 chats created with potential clients for AudienceMart
What Made Our Approach Work
Access to audiences others can't reach
We parse members even from private Telegram chats - including the ones where the member list isn't visible at all, it's hidden
Then we pull every message the person has ever left and add profile data on top - including the bio (a project link, a role, "looking for traffic" right in the description)
Multi-layer filtering
Then we run each member and their messages through several layers of signal filters: direct requests ("looking for traffic", "who's running traffic to dating"), contractor search ("need an agency / KOL manager / media buyer"), and indirect signals (a token launch, an AMA, active ad buying).
In the end the message only goes to people whose need is real and current
Personalization from the lead's own words
After filtering, we rank all the signals by relevance and take the strongest ones into work. For each of those leads we build a custom message based on their own message - exactly what they asked for, in their own context.
It's inserted into the outreach automatically, so the person reads not a template but a message about their specific task, in their own style
Messaging Strategy
Here's how it looks in practice. The {{hook}} variable is built from what the lead wrote in the chat and gets inserted into the message.
The campaign ran in Russian, the leads' native language - the examples below are translated.
Template with the {{hook}} variable:
Hi! Saw your message about {{hook}}.
My team drives traffic through different sources: Telegram channels, mini apps, bots, organic content from viral short-form videos, and influencers.
Our website: audiencemart.com
If you're interested, we can put together a media plan based on your product's niche and geo.
What the lead wrote in the chat:
"Looking for traffic for my app. DM me)"
"I have an Android mobile app. Want to distribute it and pay on CPA."
The final message - the variable filled in from their own words:
Hi! Saw your message about traffic for an Android app on CPA.
My team drives traffic through different sources: Telegram channels, mini apps, bots, organic content from viral short-form videos, and influencers.
Our website: audiencemart.com
If you're interested, we can put together a media plan based on your product's niche and geo.
Tools Stack
OutCheese Telegram - our own platform, handles the Telegram sending
Bright Data - proxies
OpenAI API - mass row filtering, thousands of rows in hours
ICP We Chose (and Why)
01
Crypto / Web3 projects
that need traffic: token launches,
mini apps, bots, channels
02
Telegram-ecosystem products:
channels, bots, mini apps
03
Web services / SaaS
and consumer mobile apps
Stage:
Already buying ads or working with influencers, with confirmed demand and real budgets.
Target Personas:
Founders / Co-founders
Head of Marketing / Growth
Performance marketers / Media buyers
Influencer marketing managers
Buying Intent Signals
Direct traffic request - openly asking for traffic, users, or promotion in the chat
Looking for a contractor - searching for an agency, a KOL manager, a media buyer, or a Telegram specialist
Already spending - visible active paid acquisition (Meta, Google, influencers) and the intent to scale
Launch buzz - a token / product launch, an AMA, or a promo is in progress
Accessible decision-maker - a founder, growth lead, or media buyer is active in the chat
What Didn’t Work
We were running a pilot and had little time to generate leads, so we tried to connect interested leads with AudienceMart as fast as possible. As soon as someone replied positively ("interesting", "tell me more"), we'd immediately write "great, setting up a shared chat now - we'll go through the details" and create a group chat with the client.
For some leads this was too fast. A positive reply isn't readiness yet, and some reacted with irritation: "I didn't ask you to create a chat, and I don't have any questions".
Lesson: jumping straight into a shared chat isn't always the right move - you need to qualify the lead, answer all their questions, and wait for their consent before taking any action.
Key Takeaways
Contacts you won't find on Google are an unfair advantage - reaching audiences from private chats puts you in front of buyers your competitors don't even see
Telegram beats LinkedIn and email with crypto / Web3 traffic buyers - 30% replies vs 6-8% and ~1%. Message people where they're actually active
Need + personalization is the lever - writing only to confirmed demand, with a custom variable built from the lead's own words, is what takes the reply rate this far above the market