Scoring API

Connection intelligence in one call.

Send the phone number, the IP address and your context. The Scoring API scores them against intelligence reported by the operators that carry the traffic, and returns a risk ranking, a band, a recommended action and the evidence behind it.

POST /v1/score Illustrative, abbreviated
Scenario
Tier
Request

        
Response 200 OK

        

Three tiers. Each one adds evidence.

The tiers are cumulative. Each starts from the one below and can only raise a call up the bands as new evidence arrives. Nothing moves a call down.

01

Standard

Phone number

Can this number exist, and how likely is it to be used for fraud?

caller_id_integrity
Is this caller ID even possible, or does it sit in an unallocated range?
reputation_ledger
Has this exact number, its range or a nearby block been reported?
ndc_range_risk
How much fraud has built up against this numbering range?
range_prediction
How likely is this range to carry fraud next, before any report arrives?
premium_tariff
Is it a premium-rate range?
02

Premium

+ IP address

Where does the session really come from, and is that infrastructure used to hide?

ip_reputation
This exact IP has been reported. High on its own.
Hard anonymisers
Tor, proxy or bogon traffic, routed to be untraceable. High on its own.
Evasion infrastructure
Data centre, cloud or VPN rather than a consumer line. Circumstantial.
Geography
Origin on a different continent from your expected corridors. Circumstantial.
03

Bureau

+ Behaviour

Does this subject behave like fraud the network has already seen?

Infrastructure clustering
Numbers and IPs working as one network, with a stable cluster_id.
Velocity
Bursts of attempts from one number, IP or cluster.
Impossible travel
One number appearing from origins no journey could connect.
Origin spread
One number arriving from many different networks.
Account linkage
Caller ID not matching the account, or a run of failed logins.

Number and IP intelligence describe the asset. Bureau describes how the asset is being used.

The most severe verdict wins.

Each layer assigns its own severity and the request takes the highest. A strong signal is never diluted by clean results elsewhere, and weak signals only escalate when independent sources agree.

Two rules no configuration can override

1. Only a confirmed report in operator intelligence can reach block.

2. Proximity to reported numbers is capped. It never blocks and never counts as corroboration.

  1. CriticalConfirmed report: exact number or reported rangeBlock
  2. CriticalImpossible caller ID plus at least one independent signalChallenge
  3. HighOne hard signal: impossible caller ID, proximity, reported IP, Tor, proxy or bogonStep-up auth
  4. MediumTwo independent circumstantial signalsInvestigate
  5. LowOne circumstantial signalWatch
  6. MinimalNothing adverseAllow
Standard · range_prediction

Flag the range before the report.

Our model learns where fraud accumulates across numbering ranges, so a number can be flagged before it, or anything near it, has been reported.

ReportedPredictedNo signalIllustrative
Bureau · infrastructure clustering

Many caller IDs. A few addresses.

Caller IDs are free and unlimited. Hosting addresses are finite and cost money. When many identities share a few addresses, the operation shows itself.

coverage: strong · partial · thin

A low score with thin coverage means unwatched, not clean.

Every response says how much operator intelligence stands behind the destination country. It never changes the score. Wire it into your workflow anyway.

Your policy decides what happens.

The recommended action is friction, not a verdict on your customer.

action_policyMap each recommended action to what your flow actually does.
risk_postureTilt the whole model: recall_weighted, balanced or precision_weighted.
score0 to 100 for thresholds and analytics. A ranking: 78 means riskier than 78% of scored ranges.

Anywhere a stranger arrives.

What the Scoring API never returns.

You get risk and reasoning. You never get a subscriber's identity, and you never learn which operator reported what.

  • subscriber_name
  • address
  • location
  • date_of_birth
  • device_owner
  • reported_by
  • member_id

Priced by tier and usage.

A base annual licence plus usage that scales with your volumes. Organisations that contribute to the Exchange pay less to consume.

Standard

Phone number intelligence

  • Reputation ledger and range risk
  • Range prediction
  • Caller ID integrity
  • Sandbox on registration
Start in sandbox

Premium

Phone + IP

  • Everything in Standard
  • IP reputation and anonymiser detection
  • Geography against your corridors
  • Named technical contact
Talk to us

Bureau

Phone + IP + behaviour

  • Everything in Premium
  • Clustering with stable cluster_id
  • Velocity, travel and origin spread
  • Account linkage signals
Talk to us

Score your first connection this afternoon.

Register, get sandbox keys straight away, and integrate before any commercial conversation.