Launch expert: 3 years of Telegram traffic, $167,175 ad spend, 277% ROMI

Context and task

The client is a direct-marketing expert: sells courses and mentoring on launches through a Telegram channel and a network of bot funnels. The flagship program starts once every few months; between launches the audience is warmed up with content and tripwires at ₽2–5K — audience research, lead magnets, mini-trainings.

The pain is standard for the niche, but in an acute form. First, Meta does not see the event “subscribed to a Telegram channel” — without stitching the click to the subscription the algorithm has nothing to learn from, campaigns optimize for clicks and pour in bots. Second, online education burns through creatives in weeks: the combo that brought a subscriber for a dollar yesterday brings one for three tomorrow. The client kept a strict KPI: first a subscriber at up to ₽250, then tightened to ₽200 — “anything more expensive gets optimized or switched off”.

What we did

2023: stitched the click to the subscription — and found the first combo. Connected tgtrack: the service catches the click-through and the subscription and sends a server-side event to Meta, so campaigns learn on real subscribers, not clicks. The first month gave a subscriber at ~$2.4 against a KPI of ~$2.7. Then the funnel built on the expert’s loudest case kicked in — it brought the subscriber down to $1 and carried 90% of traffic for months.

2024: a funnel conveyor instead of “the best combo”. At first we thought like everyone else: find the best funnel and scale it. It turned out that in this niche combos live for weeks, and scaling kills them faster. We rebuilt into a conveyor of 8–10 parallel funnels: today some bring the leads, tomorrow others. The team called it diversification: the failure of any single funnel stopped being the failure of the month. Burned-out creatives were not thrown away — the offer library came back into rotation a couple of months later and worked again.

Experiments ran continuously — some became the system:

• Traffic to Telegram funnels instead of the channel: a separate circuit of campaigns led not to the channel but to bots with tripwires — a funnel lead cost from $1.4, the best combos produced a payment the same day (“3 leads, 2 payments — spent 9K, earned 10” — a day from the chat).
• The “channel ↔ funnels” swing: before every flagship launch the entire budget was switched from funnels to subscriptions — a warmed-up base for the launch matters more than leads today.
• Geo splits: Russia/Belarus/Kazakhstan separately, then cities versus regions, finally a split into T1/T2 countries. Kazakhstan and Belarus were switched off for part of the funnels due to lead quality.
• Campaign structures: CBO “broad + LAL on the buyer base + interests”, super-objectives against traffic cannibalization between post and bot, the rule “working campaigns are not touched — we scale only with new ones”.
• Instagram post promotion at 5–7 cents per click-through — a cheap warm-up of the base.
• A subscription-check button on the landing post — the subscriber immediately became an entry into the funnel.
• What did not work, we also know for sure: a post without a call to action — zero subscribers; delayed conversions drove the price to ₽600 — rolled back within a week.

December 2024: ad account ban — the most expensive month of the project. Meta banned the main ad account with ~$100K of accumulated spend. The loss was not money but trust: a new ad account passes moderation more slowly, ads get rejected more often. In parallel the client closed the channel (an experiment with a closed community) — and the campaigns trained on subscriptions collapsed. We relaunched the whole structure from scratch in a backup Business Manager; of the sagging combos one survived — and that is what we used to ramp up the new ad account.

2025: the video breakthrough. The rule became: launch on statics, within a week remake into video whatever took off. Clips of ~25 seconds — voice-over, subtitles, the message delivered through the expert’s personal case. One of these videos “shot through” to a subscriber at $0.5; it stabilized at $0.64–0.7 — three times below KPI. The client later retold this himself as the best result in the niche: competitors pay ₽200–300 per subscriber, he pays under ₽100.

2026: traffic to payment instead of subscription. For the client’s new product (a bot service) we moved optimization from subscription to events inside the bot: activation → paywall → purchase. Meta started learning on money, not on entries: 2,995 activations → 403 paywalls → 16 payments in the test combo — and further iterations on this funnel from there.

Results

• Ad spend in Meta Ads — $167,175 over 34 months without a break (09.2023–06.2026)
• 48,593 target events — channel subscriptions and funnel leads (09.2023–02.2026)
• Another 11,365 subscriptions to the client’s new channel for 03–08.2026 per the subscription tracker
• The channel grew to 55,000+ subscribers at its peak
• Cost per subscriber: ~$2.4 at the start → $1.5–1.8 at scale → $0.5–0.7 on video creatives against a KPI of ~$2.2
• Traffic ROMI — 277%

Insights

• The message matters more than the visual. The best creative is not the prettiest but the one where the offer is clear in seconds; mentioning Telegram in the creative had no effect on conversion at all.
• Optimizing for subscription and for a lead are different machines. When all traffic was moved to the channel, the funnel lead got more expensive: the algorithm stopped seeing the entry event. Two circuits are not a hedge but the structure.
• In a niche where combos burn out in weeks, stability comes not from the best funnel but from the number of them. Diversification across 8–10 funnels + a library of burned-out creatives for relaunch.
• An ad account ban is a question of “when”, not “if”. A backup Business Manager with its own history must exist before the ban, otherwise the relaunch eats a month.
• Three years with one client in a niche where contractors are changed every quarter is a consequence of the mechanics: the client saw not click reports but cost per subscriber against his KPI every day.