A Telegram inbox with four hundred open conversations has an ordering problem, not a volume problem. Somebody has to decide who gets messaged this morning. Done by feel, that decision favours whoever replied most recently and whoever sounded keenest, which are two of the weaker predictors available.
Score on what happened, not on what was said
Scalegram stores no message content. Conversations are never saved, only contact recency and the link back to the real Telegram chat. That constraint turns out to push scoring in the right direction, because it removes the tempting signal and leaves the useful ones.
What is left is behaviour. Which tracking link brought them in, and therefore which campaign and which UTMs. Whether they clicked. Whether they joined. Whether a registration event fired. Whether a purchase event fired and what value it carried. How recently they were in contact. What stage a human or the assistant last moved them to.
Every one of those is a thing the lead did. None of them can be faked by writing a friendly message, and none of them requires reading anyone's private conversation to compute.
The four inputs that carry most of the weight
Start simple and resist the urge to build a formula with nine terms. Four inputs get you most of the way.
| Input | What it tells you | Watch out for |
|---|---|---|
| Recency of contact | Whether the conversation is still warm at all | Decays fast; a reply last month is nearly cold |
| Source and campaign | Which traffic actually converts for you | Judge it after purchase data, not after click volume |
| Verified events | Click, join, registration, purchase with value | Only counts if the event is verified, not self-reported |
| Stage | Where a person or the assistant last placed them | Goes stale if nobody moves stages honestly |
Weight them for your own funnel. A signals community and a software business have different distances between joining and paying, and a score copied from someone else's business is a score tuned for someone else's funnel.
Verified beats reported
The single largest quality gain comes from insisting that events are verified rather than claimed. A lead who says they deposited is not the same as a deposit confirmed through an integration, and in trading niches the gap between those two is wide enough to run a whole fake pipeline through.
Scalegram can check the state of things through connected integrations rather than trusting a message, including a broker IB integration used to confirm a deposit before a VIP invite is sent. The general principle holds beyond that specific case: any score built partly on self-reported milestones will rank the confident above the committed. How confirmed purchases land in the pipeline is covered in purchase sync.
A score is a sort order, not a verdict. It decides who you talk to first on a Tuesday morning. It should never be the thing that grants access, refuses someone or triggers a charge.
Decay, or the score lies to you
Scores that only go up are useless within a month. Everyone who ever clicked anything accumulates points and the top of the list fills with people who stopped answering in spring.
Recency has to pull the score down as it ages, and it should pull hard. A lead who was in contact yesterday and has done nothing else is usually a better use of the next hour than a lead who did four things a quarter ago and vanished. Decay is what keeps the list honest, and it is the part most homemade scoring skips.
Who acts on it
A score with no owner is a column nobody reads. Assign the top of the list to specific people, with permissions matching what they will need to do, since team access is granted per area for read, add, write and delete. The point is that the person who opens the ranked list can act on it without asking someone else for rights halfway through.
Pair it with saved views rather than a single global ranking. "High score, no contact in seven days" and "purchased, no follow-up" are lists someone can actually work through, which a raw leaderboard is not. That structure sits alongside the stages and tags described in lead management.
Check it against outcomes
Once a month, take the leads that converted and look at what their score was a week before they converted. If the high scorers converted at a similar rate to everyone else, the score is decoration and the weights need work.
That check is the only thing separating a scoring model from a horoscope. It also tends to correct the most common mistake, which is over-weighting source. Campaigns that deliver cheap clicks look excellent in a score built on early signals and much less excellent when measured against purchases with values attached. Fix the weights, keep the mechanism, and let the list decide the morning.
"Someone who clicked your link, joined the channel and asked what the payment methods are outranks anyone who typed a paragraph about how excited they are."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- Score on behaviour the lead performed, not on how enthusiastic their messages sounded.
- Recency, source and campaign, verified events and stage carry most of the useful signal.
- Insist on verified events rather than self-reported milestones, or the score ranks confidence above commitment.
- Decay the score with time, and check monthly whether high scorers actually converted better than the rest.
Frequently Asked Questions
How can leads be scored if messages are not stored?
Because the useful inputs are not message content. Contact recency, the tracking link and campaign, verified click, join, registration and purchase events, and pipeline stage are all recorded without saving anything anyone wrote.
Should a score decide who gets access or a discount?
No. A score is a sort order for deciding who to contact first. Access, charges and approvals should go through explicit gates with a person involved.
Why do scores stop working after a few weeks?
Usually because nothing decays. Points accumulate forever and the top of the list fills with contacts who went quiet months ago. Recency has to pull the score down as it ages.
About the Author
Alex Onta is an Executive Director at SINGUARD. He built eTrader, the terminal, the mobile apps, eTrader Broker, Copytrading, Business and Community, along with the worldwide clustered-server infrastructure it all runs on, with his brother Roman Onta helping on the design, and he leads that division today. Together with Roman he builds the Prop Firm CRM, the Broker CRM, Scalegram and CopySignals, and the two of them carry worldwide compliance, payment processing and international business structuring side by side. He lives and works in Dubai for most of the year. Meet the executive duo leading Singuard's five divisions.