A LinkedIn connection used to mean something. In 2014 it meant a coffee, a previous job, an actual conversation. In 2026 it means one of you accepted a request you don't remember. The signal-to-noise has collapsed.
A mutual Telegram group is a different kind of signal — and a much stronger one. Let me explain why, and what to do with it.
What "warm" actually means
A warm signal is anything that says: this person is currently interested in the thing I sell, and there's a non-creepy way for me to start the conversation. Two parts. Most "warm" lists fail one or both.
LinkedIn second-degree connections fail the first part. Job-change alerts fail the second (the "congrats on the new role" opener is now a meme). Bombora intent surges fail both — the prospect doesn't know why you're calling and the topic is months stale by the time it lands in your CRM.
Mutual Telegram group membership passes both filters cleanly:
- Currently interested. The group exists because the topic is alive right now. People left groups they no longer care about. The remaining members are self-selected on the topic, today.
- Non-creepy opener. "I saw your message in [group] last Thursday about [thread]" is observably true, specific, and the prospect understands why you're reaching out. There's no pretext to manufacture.
That's the whole thesis: self-selection plus current activity equals a warm signal that holds up under scrutiny.
Why this is structurally true on Telegram and not on Twitter
The obvious counter is "you can do the same thing on Twitter or Discord." You can, but the structure is weaker.
Twitter follows are public, asymmetric, and noisy. People follow accounts they disagree with, accounts they followed five years ago, accounts they barely registered. A follow is not a topical commitment.
Discord servers are stronger but suffer from the mega-server problem — a 50,000-member general server isn't a topical signal, it's a noise signal. The good Discords are private and invite-only, which means you can't enumerate membership.
Telegram sits in a sweet spot: groups are small enough (typically 200–5,000 members) to be topical, public enough that membership is observable from inside, and active enough that recency is a strong filter. The ones that matter for B2B — niche founder Telegrams, market-making chats, regional dev communities, indie-hacker back-channels — are exactly the right granularity.
Scoring it
Scoring needs the mutual-group list as data first — the Telegram mutual groups export resolves shared groups for a handle and saves them alongside the member list, which is the input every formula below assumes.
The simplest scoring model that works:
warm_score = α * mutual_groups
+ β * topical_relevance(mutual_groups, ICP)
+ γ * recency(last_active_in_shared_group)
Tuning we've found works:
- α ≈ 0.4 — base weight on raw mutual count. Diminishing returns past 5–6 groups.
- β ≈ 0.4 — most of your lift is here. A founder Telegram with three of your buyers is worth ten generic crypto chats.
- γ ≈ 0.2 — recency multiplier. Active in last 7 days = 1.0; 30 days = 0.6; 90+ days = 0.2.
Cap the score, then bucket prospects into tiers (A/B/C). Send tier-A to your best AE manually, tier-B into a templated-but-personalized sequence, drop tier-C entirely. The discipline of not contacting tier-C is the part most teams skip; doing so is what protects reply rates as you scale volume.
Opener templates that hold up
Three patterns we've seen work, ranked by reply rate:
1. Specific-thread reference. "Saw your reply in [group] on Tuesday about [specific topic]. We hit the same wall last quarter — ended up [solution/observation]. Curious how you're thinking about it."
This works because it's observably true and trades information instead of asking for time.
2. Cross-group pattern. "Noticed we're both in [group A] and [group B] — pretty unusual overlap. If you're working on the intersection of [topic A] and [topic B], would love to compare notes."
Works for pattern-recognition prospects who appreciate that you noticed the overlap.
3. Ask for a tactical answer. "You mentioned [tool/technique] in [group] last week. Quick question — does it handle [specific edge case]? Trying to decide between that and [alternative]."
Works because it's a real question they can answer in 30 seconds, and gets you in the conversation.
What doesn't work: "We're both in [group], let's hop on a call." That's a LinkedIn opener with extra steps.
The boring discipline that makes this scale
Two operational notes that matter more than they sound:
- Cap volume per group per week. A group with 800 members can absorb 3–5 outbound touches a week before it starts feeling spammy. Beyond that you become that person and the warm signal inverts. Set the cap, enforce it.
- Track replies in the CRM, not the inbox. Reply rate by group, by topic, by opener type — instrument it. The model gets sharper every month if you measure it; it stays a vibe if you don't.
Where this stops working
Two failure modes to call out:
Selling outside community-native categories. If you sell payroll software to dental practices, Telegram is the wrong play. The warm signal exists where the buyer's identity is bound up in the community — crypto, dev tools, creator economy, indie SaaS, regional builder scenes. Outside those, this approach over-rotates on a signal that isn't there.
At very low ICP density per group. If you're getting one fit prospect per 500 members, the math gets bad fast. Warm-group outbound is for niches where the group is the ICP, or close to it.
Inside the niches where it works, mutual-group scoring is the most leveraged outbound signal we've found. The CRM has the contacts. The community has the activity. Bridging the two is the whole game.