Voice

Outbound traffic

Outbound traffic is calls a business places: agents dialing customers, notifications, and transfers off-net. Carriers treat that direction as termination onto the PSTN or a peered IP network.

Updated August 27, 2026

Outbound is the direction you initiate. A user clicks to dial, an autodialer launches a campaign, a PBX follows a forward off-net, or an application places a two-leg call through CPaaS. The destination is usually an E.164 number you do not own. The network must accept your session, route it, and present an identity the far end will trust enough to answer.

That makes outbound a different operational problem from inbound traffic. Inbound is mostly a routing and staffing problem. Outbound is about identity, consent, pacing, and interconnect quality. A call that completes at the network level can still fail in practice if it is labeled spam or if it violates TCPA or equivalent rules.

What counts as outbound

If your system sends the SIP INVITE (or the WebRTC equivalent) toward a number that is not an on-net extension, that leg is outbound. Internal extension-to-extension calling is not outbound traffic in the carrier sense. A warm transfer that adds an external party creates a new outbound (termination) leg even if the original call was inbound.

  • Manual click-to-call from a CRM or softphone.
  • Preview, progressive, and predictive dialers in a contact center.
  • Transactional alerts (appointment reminders) that use a voice channel.
  • International calling, which is still termination, but at different rates and under different CLI rules.

CLI, CNAM, and STIR/SHAKEN

The number you present is the Calling Line Identity (CLI), a form of Caller ID. It should be a number you are authorized to use, typically a DID on the same account. Many networks rewrite or reject CLI that does not pass attestation. The display name the callee sees, if any, often comes from a CNAM database keyed by that number, not from a string you send in SIP.

In the United States, originating providers are expected to authenticate outbound calls with STIR/SHAKEN. Attestation levels describe how strongly the provider ties the call to the customer and the number:

Attestation A (Full)
The provider can authenticate the customer and confirm they are authorized to use the calling number.
Attestation B (Partial)
The customer is authenticated, but the provider cannot fully confirm number authorization (common on some wholesale paths).
Attestation C (Gateway)
The call entered from an unauthenticated gateway. This is the weakest signal and the most likely to be treated as untrusted.

Attestation is not a substitute for clean CLI hygiene. High-volume outbound on rotating numbers, or CLI that does not match the brand the callee expects, still attracts analytics labeling and blocking. Spoofing someone else's number is both a trust problem and, in many cases, unlawful.

In the United States, the Telephone Consumer Protection Act (TCPA) and FCC implementing rules constrain marketing calls, autodialed calls, and prerecorded messages, especially to mobile numbers. Prior express consent (and, for certain autodialed or prerecorded marketing, prior express written consent) is a legal requirement, not a carrier feature. National and state do-not-call lists apply on top. Other countries have analogous regimes (Ofcom in the UK, CRTC in Canada, ePrivacy and national laws in the EU).

Capacity and calls per second

Outbound load is often limited in calls per second (CPS) as well as concurrent sessions. A dialer that blasts 100 INVITEs in one second can trigger provider fraud heuristics, interconnect CPS caps, or downstream analytics even if the concurrent cap is not reached. Providers watch answer seizure ratio (ASR), average call duration (ACD), and the share of short-duration calls because those metrics correlate with robocalling.

International outbound brings its own considerations: least-cost routing, quality-based routing, and CLI policies that vary by destination country. Some nations require a local CLI; others reject foreign CLI. See E.164 for how destination numbers are formatted, and origination and termination for how that hop is billed.

Outbound use cases

The same [[sip-trunking|SIP trunk]] can carry several of these if CLI, CPS, and consent are separated by campaign.
Use caseTypical identity needsTypical constraints
Agent-assisted sales or supportStable CLI + CNAM the customer recognizes; A attestation where possibleTCPA / DNC if marketing; recording consent if recorded
Appointment remindersConsistent local or toll-free CLIConsent to call; low CPS; avoid silent calls
Click-to-call from a web appBusiness DID as CLI; often WebRTC on the agent sideNAT, microphone permission, and the same CLI rules as a desk phone
High-volume collections or outreachAuthorized numbers, traceback-ready recordsAnalytics labeling risk; CPS caps; sector-specific law
International customer callsCLI accepted by the destination; possible local presence numbersRate, quality, and country CLI policy

How outbound differs from inbound

You choose when outbound happens. You do not choose whether the callee's network accepts your identity. Downstream carriers, analytics firms, and handset spam lists sit on that path. Keeping outbound healthy is a reputation problem (number age, answer rates, complaint rates) as much as a call quality problem.

For the reverse direction, including IVR and queues that greet people who dial you, see inbound traffic.