A referral bonus is an instruction to create fake accounts. Every referral system is a fraud problem wearing a growth mechanic, and the ones that fail do so because that was treated as an afterthought rather than the core requirement.
How attribution works in Telegram
Telegram gives you deep links: t.me/yourbot?start=CODE. When someone opens that link, your bot receives the code with their first message, and you attribute the join. It is clean, instant and requires no cookies or tracking pixels — which is why referral mechanics work better here than on the web.
The complexity is not attribution. It is deciding which joins deserve payment.
What separates a working system from a drained treasury
| Control | What it does | Without it |
|---|---|---|
| Account age check | Rejects brand-new throwaway accounts | One person farms hundreds of invites in an evening |
| Joining velocity | Flags bursts from one source | A script invites faster than you can review |
| Device and IP signals | Detects self-invites | Users refer themselves indefinitely |
| Qualifying action | Pays only after the invitee does something real | You pay for joins that never become users |
| Held-for-review queue | Suspicious joins wait rather than pay | Fraud is irreversible once paid |
| Lifetime caps | Bounds worst-case exposure per user | One exploited account can drain the budget |
Multi-level, and when it is worth it
A multi-level tree pays the referrer and their referrer, sometimes deeper. It increases the incentive to recruit recruiters rather than users, which is powerful and also exactly how these systems drift toward looking like a pyramid.
Two levels with a hard cap is usually the sensible ceiling: enough to reward genuine network effects, shallow enough that the economics stay legible and the optics stay defensible.
Where the effort goes
Making people share more than once
| 1 | Show progress toward the next milestone A visible counter with a target outperforms a flat per-invite rate. |
| 2 | Notify on every successful referral The reward has to feel immediate, or the loop does not reinforce. |
| 3 | Run a leaderboard Competition drives the top decile, who bring most of the volume. |
| 4 | Explain rejections A flagged referral with no reason reads as theft. Say why, and allow appeal. |
| 5 | Pay something small early The first payout is what makes the rest of the scheme credible. |
What ShipWeb3 sells here
The Telegram Referral Bot is the complete system: multi-level tree with per-level rates, instant deep-link attribution, fraud scoring on account age, device fingerprint and joining velocity, a held-for-review queue with a reason on every flag, milestones, leaderboards, segmented broadcasts and admin payout controls with CSV export.
Referrals are also built into the play-to-earn app and the USDT-paying bot if you want them as part of a larger product rather than standalone.
Common questions
Growth without the drain Multi-level attribution with fraud scoring, review queue and payout controls. See the referral bot |
