Telegram welcome message: a sequence, not one greeting
A welcome funnel is steps with delays, each skippable by condition, stopping the moment someone pays. Per-step drop-off is the part that makes it improvable.
TL;DR. A welcome funnel is a sequence of messages with delays between them, not a single greeting. It fires on one event — an approved join request — so the channel needs join requests enabled. Each step can carry media and buttons, and can be skipped by condition: Premium or not, paid or not, one of four language rules. The funnel stops on its own when someone pays or leaves, if you let it. And every step reports its own drop-off, which is the only part that makes the thing improvable rather than decorative.
Most channels have a welcome message, and most welcome messages do one job badly: they greet, they explain, and they ask for money, all in the first thirty seconds of a relationship. The reason they do it all at once is that one message is all there is.
One trigger, and what it implies
The funnel starts when a join request is approved. That is the only trigger, and it is a precondition rather than a detail: a channel that anyone can join freely never produces the event, so there is nothing for the funnel to hang on.
If your channel is open, this feature asks you to close it behind join requests first. That is a real trade, not a formality — join requests add friction to joining and remove the “just walked in” audience.
Steps and delays
A funnel is steps in order, each with its own delay. That is the difference between a greeting and a sequence: the welcome lands immediately, the explanation an hour later, the offer on day two.
Each step carries text, media and inline buttons, and can disable link previews or protect its content from forwarding. There is also a preferred send hour, so a step scheduled for 3am waits until morning instead of arriving at 3am.
Conditions: one funnel, several audiences
A step can be skipped by condition, and there are eight: Premium only, non-Premium only, has paid, hasn’t paid, and four language rules.
This is what keeps a funnel from needing to be several funnels. The step that pitches a paid subscription carries “hasn’t paid”, so buyers never see it. The step that thanks people for buying carries “has paid”. Same sequence, two experiences, and nobody gets the wrong one.
Stopping is a feature
Two switches decide when the funnel gives up on someone: stop on payment and stop on unsubscribe.
Turn on stop-on-payment. A sequence that keeps pitching after the purchase is the fastest way to make a new customer regret being one, and it is the single most common way welcome funnels do damage. Those two switches are yours; the opt-out button is the reader’s own — it lets someone end the sequence without leaving the channel.
The number that matters
The analytics report per step: scheduled, sent, failed, cancelled, clicks and a drop-off rate — plus completion rate and average time to finish.
Per-step drop-off is the one to read. A funnel with a good completion rate and one step that loses a third of people is not a good funnel; it is a good funnel with one bad message, and you cannot tell those apart from the total. Because funnels are versioned and carry a traffic weight, you can run two at once and compare against the previous version rather than against a memory.
If… then…
| If… | Then… |
|---|---|
| Your channel is open to anyone | No trigger exists — join requests first |
| You have one long welcome message | Split it: greeting now, the ask on day two |
| You sell a subscription | Step with “hasn’t paid”, and stop-on-payment on |
| Your audience is mixed-language | Language conditions rather than a second funnel |
| One step loses everyone | Rewrite that step, not the funnel |
| You want to compare two versions | Traffic weight, and read the previous-version column |
Where to start
Take the welcome message you already have and split it in two: what someone needs in the first minute, and what they need on day two. Almost every existing welcome message contains both, compressed into one paragraph that does neither well.
Then turn on stop-on-payment before you write anything else, because that switch prevents the only mistake in this feature that costs you a customer you already won.
Read the per-step drop-off after the first fifty runs. The step that loses people is the one to rewrite, and until you have that number you are guessing.
The channel this runs on — the channel page. Buttons that go in each step — add buttons to a Telegram post. Writing to people who lapsed instead — win back customers.
What people usually ask
- What starts the funnel?
- One thing only: an approved join request. That is worth knowing before you build anything, because it means the funnel needs join requests turned on for the channel. A channel anyone can enter freely has no moment to trigger on.
- Is it one message or several?
- Several, in order, each with its own delay. The delay is the whole point — a greeting and a nudge two days later are different messages doing different jobs, and sending them together makes both worse.
- Does it keep selling to someone who already bought?
- Not if you turn on stop-on-payment, and you should. A funnel that keeps pitching after the purchase is the fastest way to make a new customer regret being one. There is a matching stop-on-unsubscribe for people who leave.
- Can a step be shown to only some people?
- Yes, with eight conditions: Premium or non-Premium, has paid or hasn't, and four language ones. A step with a condition is skipped for everyone it doesn't match, so one funnel can carry both a payment nudge and a thank-you without sending either to the wrong person.
- Can I test two versions?
- Yes. Funnels have a traffic weight, so several can run on the same channel with the split you set, and the analytics compares against the previous version.
- What can I actually measure?
- Per step: scheduled, sent, failed, cancelled, clicks and a drop-off rate — plus completion rate and average time to finish for the funnel as a whole. The per-step drop-off is the number worth looking at; it tells you which message loses people.