What is STIR/SHAKEN, in plain words?
STIR/SHAKEN is the caller ID authentication system US phone carriers use to digitally sign each call, stating who placed it and whether that caller has the right to use the number shown. The receiving carrier checks the signature. Spam labels come later, from analytics engines that combine this signature with how your numbers behave.
The two names are acronyms for technical standards. STIR is Secure Telephone Identity Revisited and SHAKEN is Signature-based Handling of Asserted information using toKENs. You do not need to remember either. What matters is that the FCC required large US voice providers to put the framework in place on their IP networks from June 2021, and smaller providers were brought in after that. Our STIR/SHAKEN glossary entry has the one-line version for new staff.
If your floor sits in Lahore, Karachi, Mohali or Noida and dials US numbers, this matters to you even though you never touch a US carrier directly. Every call you place enters the US phone network through some provider, and that provider signs it, or fails to, on your behalf. The result shows up on the consumer's screen as a clean number, a name, or a warning.
How do attestation levels A, B and C work?
When the originating provider signs a call, it gives it one of three attestation levels. The level is the provider's statement about how well it knows you and your number.
- A, full attestation: the provider knows the customer who placed the call and has confirmed that customer is allowed to use the calling number.
- B, partial attestation: the provider knows the customer but cannot confirm the customer is authorised to use that particular number.
- C, gateway attestation: the provider is only passing the call along, often from an international or upstream source, and cannot vouch for the customer or the number.
Why do offshore calls so often land at B or C?
Most offshore floors buy minutes and numbers from a VoIP reseller, who buys from a wholesaler, who hands traffic to a US carrier. Each hop is a chance for the link between you and your number to get lost. If the provider that finally signs the call did not issue your numbers and has no record of you, it can only give B or C.
The fix starts with your paperwork, not your dialer. Ask your carrier in writing which provider signs your calls, whether it can sign them with A attestation, and what it needs from you to do so. Usually that means buying your DIDs from the same provider that carries your outbound traffic, or proving to that provider that you have the right to use numbers you got elsewhere.
A carrier that cannot answer these questions clearly is telling you something. Cheap minutes that arrive with C attestation can cost more in lost answers than they save on the invoice.
Does an A attestation stop the Spam Likely label?
No, and this is the part most floors get wrong. Attestation proves identity. It says nothing about whether the calls are wanted. A perfectly signed call from a number that dials thousands of people a day, most of whom hang up in five seconds, can still be labelled.
The labels you see on US phones, such as Spam Likely, Scam Likely or Potential Spam, come from call analytics engines that the big carriers and handset makers work with. Each engine uses its own model, so the same number can look clean on one network and flagged on another. They look at attestation, but they also look at behaviour. That behaviour is what people mean by caller ID reputation.
- How many calls one number places per hour and per day
- How many of those calls are answered, and how long answered calls last
- How many recipients block the number or report it as spam
- Whether the number is new, recently ported or suddenly very busy
- Whether call volume follows normal business hours in the recipient's time zone
Why do fronting numbers get flagged so fast?
A typical lead generation floor creates exactly the pattern these engines look for. A predictive dialer pushes high volume through a small pool of numbers. Answering machine detection hangs up on voicemail within seconds. Stale lists produce wrong numbers and angry people. Each of those short calls and reports is a data point against the number.
Night shift timing adds to it. A floor in Pakistan working the US evening is fine, but a campaign that drifts outside the recipient's calling window, or that dials Pacific numbers on an Eastern schedule, creates complaints that hurt the whole pool. Volume spikes do the same. A number that was quiet yesterday and places three thousand calls today looks like a robocaller to any model.
What can an offshore floor actually do about it?
You cannot buy your way out of a bad label, but you can control most of the inputs. These are the steps that make a practical difference, roughly in the order we would do them.
- Get A attestation where you can. Buy numbers from the carrier that signs your calls, and confirm in writing that your traffic is signed A.
- Register your numbers. The main analytics providers accept registrations from legitimate businesses, and some accept them through a shared portal. Register the business name, the purpose of the calls and every outbound number.
- Set up CNAM. Registering a caller name (CNAM) for each number means some phones, mostly landlines, show your business name instead of only digits. Branded calling products for mobiles exist too, usually at extra cost.
- Size the pool to the volume. Keep calls per number per day at a level a real business would place. If your volume grows, add numbers and register them before you use them.
- Clean the list before the dial. Scrub against the DNC registry and your own suppression list, and drop data that is too old to be wanted.
- Test your own numbers. Keep a few handsets on different US carriers, or use a monitoring service, and check weekly how your numbers display.
Is number rotation ethical, or just a way to dodge labels?
Using a pool of numbers that matches your call volume is normal. Every large contact center does it. What carriers and regulators object to is rotation used to hide behaviour: burning through numbers as soon as they are flagged, cycling fresh numbers so complaints never stick, or showing local-looking numbers you do not own. Displaying a number you have no right to use, with intent to defraud or cause harm, is spoofing under the US Truth in Caller ID Act.
The honest test is simple. If a consumer calls the number back, do they reach your business or the campaign's business, and can they ask to be removed? If yes, your rotation is a capacity decision. If the number goes dead the day after it is flagged, it is evasion, and carriers increasingly block whole ranges and providers that carry that traffic.
Treat a Spam Likely label as feedback. It usually means the list, the script, the timing or the volume is wrong for that number. Swapping the number without fixing the cause only moves the problem.
How do consent and list quality change your reputation?
The single biggest lever is calling people who expect to hear from you. A lead who filled in a form yesterday answers more often, talks longer and reports less. That raises answer rates and average call length, and both feed straight into the analytics models. Old or bought lists with unclear consent do the opposite.
Consent also matters legally, separate from labels. The TCPA rules on autodialed, prerecorded and artificial voice calls to mobiles depend on what the consumer agreed to, and the FCC has said that AI-generated voices count as artificial voices for those rules. Our guides on TCPA basics for offshore call centers and DNC scrubbing before the dial cover the detail. This post is practical guidance, not legal advice, so check anything specific with a US compliance lawyer or your buyer.
Where do voice bots fit into caller ID health?
Bots do not change attestation. A bot that runs on your dialer places calls through your carrier with your numbers, so every point above applies in the same way. What a bot can change is the behaviour on the call itself: the same polite opener every time, a clear reason for calling, and an immediate, logged response when someone asks not to be called again.
B3 Voice bots run on the center's own dialer, such as ViciDial or Convoso. The Fronter bot qualifies on your own script and warm-transfers qualified people to a live agent, and the client dashboard shows the live transcript, transfers and dispositions for every call. That gives a manager a clear view of which lists and time slots produce short calls and removal requests, which are exactly the patterns that damage a number.
A weekly caller ID checklist for managers
Labels build up slowly and fade slowly, so a short weekly routine is worth more than a panic every few months. Put one person in charge of it and keep a simple sheet of numbers, carriers and status.
- Check how each outbound number displays on at least three US carriers
- Review calls per number per day and flag any number above your own ceiling
- Look at answer rate and average answered call length by number and by list
- Confirm every new number was registered and CNAM was set before first use
- Read a sample of calls that ended in under ten seconds to see why
- Confirm removal requests from the past week reached your suppression list
