Pig butchering, flagged on day 9.
A “wrong person” text. Friendly chats. Screenshots of wins. Then an invite to a private signals group, and a $200 “start small.”
No single message looks wrong. Three weeks of them do.
From the outside, a scammer and a real user look the same.
The difference is in the pattern. SigWise scores every user against signals you define, so you can search your whole user base by intent:
is_scammer > 90
And find them before they cost you anything.
Fake support takeover, flagged at step 1.
A day-old “Support” account asks for a login code. Minutes later: new device, new bank account, full balance withdrawn.
Nothing gets hacked. The user is talked into it. That first message is where it should stop.
One user. Six events. Four signals scored live:
is_scammer: 91%
spam_risk: 52%
trust_score: 0.04
buyer_intent: browsing
You define the signals in plain English. SigWise scores every user against them in real time, each with a confidence score.
Got scammed? Watch for the second wave.
"Fund recovery" accounts reply to victims within hours, promising to get the money back for an upfront fee.
It's the same scam, aimed at people who just lost money.
Real recovery never starts with a DM and a fee.
Your coding agent now speaks SigWise.
Paste one line: Add SigWise to my app: https://t.co/dHsIZuQlRQ
Then ask: “Who are our riskiest users this week?” or “Set up scam detection with a Slack alert.”
Works with Claude Code, Codex and any coding agent.
Overpayment scam, flagged at message 2.
Buyer pays full price, moves payment off-platform, “accidentally” overpays, asks for a refund. The original payment never lands.
The pattern is there before the money moves. That’s where it should stop.
New account, 3 days old. One message: “Pay me by wire first and I’ll double it.” Then a $4,800 payout to a new bank account.
SigWise scores it as it happens. is_scammer: 97%. The rule fires, the payout is held, and trust and safety gets a Slack alert.
Write a rule in one line:
is_scammer >= 90 and trust_score < 2 → freeze the payout.
SigWise scores every user in real time. When a score crosses the line, the rule acts. Before the money moves.
We're SigWise.
We read what users say and do on your platform and flag the scammers before anyone loses money.
Marketplaces, trading apps, anywhere people talk and pay each other.
One API. No models to train.
https://t.co/pn2SnKm4I0
@ecomchasedimond Great tips! I'd only add:
1. A plain text email from the founder asking for reasons why they abandoned—this adds personalization.
2. If you have an Amazon store, promote that with a 'Rather shop on Amazon?' email—this builds more trust.
I've hired 133 people.
Sold millions in software.
0 funding.
Want to know what holds your company back from being high performers?
You treat your team like children.
Let me explain. 🧵