This is the next post in our series on how real customers use the Zumigo platform to solve real problems: a neobank — a bank that exists entirely on a phone, with no branch and no in-person relationship to fall back on — that relies on a single composite score to answer a question that would otherwise take a whole stack of separate checks to resolve.
Rather than pulling several raw signals and deciding case by case how to weigh them, this bank asks for one number: how trustworthy does this phone look, right now. What makes that number useful isn’t that it simplifies the question — it’s that the simplicity is on the surface. Underneath, it’s drawing on a real spectrum of signals, from things Zumigo already knows about a line to things only visible the instant a check is actually run.
At its foundation, the score draws on intelligence Zumigo has already built up about a line, well before anything unusual happens on a given call. Has this number been ported recently — and how many times, since a number ported once is different from one that’s changed hands five times in a year. Has it been deactivated in the recent past. What kind of line is it: a real postpaid line with an established carrier, or a prepaid one running on a virtual network, the kind of line that’s cheap to acquire and easy to discard. And has an unusual number of separate parties been asking about this exact phone number recently — something only visible because Zumigo sees that pattern across its whole client base, not something any single bank could ever notice on its own.
None of these require anything special to happen in the moment. That’s intelligence Zumigo has already accumulated about the line, which is exactly why they form the foundation: available on every check, cheaply, and already enough to rule out a meaningful share of the riskiest lines before anything else even runs.
Layered on top, the score can pull in facts that are only true right now, at the moment of this exact check. Has the SIM in this phone been swapped in the last few days — the clearest single tell of an account takeover in progress. Has the device behind the number changed. How long has this account actually existed, since a line with a few weeks of history behaves very differently from one with years behind it. And, critically, is call forwarding active on this number right now — because a forwarded call defeats a voice one-time code completely, silently, without tripping any of the checks above it.
This is the layer that turns the score from a snapshot of reputation into a live read of what’s happening this instant. A number can have a clean history and still be in the middle of being taken over; only a real-time check catches that.
The deepest layer looks past the phone’s own state entirely, into Zumigo’s own broader intelligence on fraud history sourced from consortium data. Has this number shown up in a spoofing attempt recently — as the attacker’s number, the victim’s, or the one being impersonated. Has it requested an unusual number of one-time codes in a single day, the kind of pattern that looks like a bot working through a list rather than a person signing in. Is it tied to far more devices or email addresses than a typical consumer would ever have. There are even flags for cases that sound almost too specific to matter until you’ve seen one: a phone still active on an application after the person associated with it has died, which happens more often in fraud than anyone would like.
Together, these layers are what make this one score a credible account-takeover defense rather than a single blunt check. A takeover rarely announces itself as one clean signal — it shows up as a recent port, a call quietly forwarded, a device that no longer matches the one on file, or a number that’s already turned up somewhere else in Zumigo’s fraud history. Missing any one of those layers means missing a class of takeover the others were built to catch.
That’s also why the score itself, not any individual signal underneath it, is the thing worth relying on. As Zumigo keeps adding intelligence — a new pattern, a new source of fraud history, a new real-time read — it flows into the same one number instead of requiring every bank downstream to add another check of its own. The specific signals behind the score will keep changing and expanding. The score is the part a bank can keep building on.
Madhu Vudali is VP, Product Management at Zumigo. Comments or questions? Connect on LinkedIn: @madhuvudali