A VoIP route is a bulk pathway for completing calls between carrier networks over IP. Buying one is not like buying a phone line — you are buying a path, and paths differ in what they carry, what they drop, and how honestly they report on themselves.
By Vedant Vhanyalkar
A VoIP route is a wholesale pathway that carries voice traffic from one carrier network to another over IP, ending at the network that serves the number being dialled. It is infrastructure sold in bulk minutes, priced per destination, and bought by the organisation originating the traffic rather than by the person making the call.
Retail VoIP is a product you use: a number, an app, a desk phone, a monthly seat. A VoIP route is a component underneath products like that. The distinction matters because the two are bought on completely different criteria. A retail buyer asks about features and support. A route buyer asks about answer rates, post-dial delay and whether caller ID survives the journey.
If you are evaluating a business phone system rather than carrier capacity, see UCaaS vs VoIP instead — this page is about the wholesale layer.
Four groups, with different requirements. Carriers and ITSPs buying termination capacity for destinations they do not serve directly. Contact centres and outbound dialer operations sending concentrated volume to specific countries. Resellers building voice products on someone else's infrastructure. And platforms — CPaaS, CRM, dialer software — that need voice underneath a product they sell as something else.
Per minute, per destination prefix, quoted on a rate deck that is reissued frequently because the underlying costs move. Anyone quoting you a single figure for "VoIP routes" is quoting an average of things that are not comparable: fixed-line differs from mobile, one mobile operator differs from another in the same country, and a direct path differs from a transited one.
Ask for the deck against the destinations you actually send, with the billing increment stated. On dialer traffic with short average duration, 60/60 billing versus per-second billing changes the effective cost more than a difference in the headline rate does.
A CLI route delivers the calling number to the terminating network so it can be presented to the called party. That single property drives most of the price difference in wholesale voice, because it drives answer rate: a handset showing a number gets answered far more often than one showing "unknown" or nothing at all.
CLI is effectively mandatory for anything customer-facing. Outbound sales, contact centre callbacks, appointment reminders, verification calls, and anything in a regulated sector where the called party must be able to identify or return the call. It is also the precondition for call authentication frameworks — a call with no calling number gives STIR/SHAKEN nothing to sign, so authenticated delivery into markets that use it is not available on Non-CLI traffic.
Route type is only half of it. Whether your CLI actually arrives depends on every interconnect the call crosses. CLI vs NCLI routes covers the three separate reasons a number can go missing and how to diagnose which one you have.
A Non-CLI route completes the call without presenting a calling number. It costs less, and there are legitimate reasons to use it: capacity and failover testing, transit between your own platforms, automated notifications where no return call is expected, and destinations where the terminating network will not accept foreign-origin CLI under any arrangement.
Where it goes wrong is when Non-CLI is bought as a cheaper substitute for CLI on customer-facing traffic. It is not a discount on the same product — it is a different product with materially lower answer rates and no return-call path. If the campaign depends on someone picking up, Non-CLI is usually more expensive per outcome despite the lower rate.
CC routes are provisioned for call-centre and outbound dialer traffic. The distinguishing factor is the traffic profile rather than the signalling: very high call attempts per second, short average duration, heavy concentration on a small set of destinations, and a dialer rather than a human initiating each attempt.
That pattern stresses a network differently from general business voice, which is why it is quoted and provisioned separately. Sending dialer volume down a route built for business calling tends to produce congestion, and sometimes a suspended account. See CC Routes for how MCC handles this traffic type.
These three are not interchangeable. A provider quoting one rate for "routes" without asking about your traffic profile has not yet gathered enough information to quote.
The thresholds below are the ones commonly used in wholesale voice operations. They are working rules of thumb, not standards — the right number for your traffic depends on destination, route type and campaign, and should be established from your own baseline rather than adopted from any guide including this one.
The share of call attempts that connect. It is the headline metric and the easiest to distort, because several mechanisms raise it without producing a conversation. Sustained very low ASR usually indicates a genuinely broken path rather than a campaign problem; unusually high ASR on a cheap route is worth investigating rather than celebrating.
The average length of answered calls. Read it beside ASR, always. High ASR combined with very short ACD is the standard signature of calls being answered by something that is not a person, or of calls dropping immediately after connection. Either way the minutes bill and the outcome does not arrive.
The silence between dialling and ringing. Long PDD generally means the call is traversing more hops than the route description implies. It also destroys answer rate directly, because callers hang up during silence — so PDD both diagnoses a problem and causes one.
FAS is a carrier signalling that a call was answered when it was not, so billing starts before anybody picked up. It is the reason ACD and ASR must be read together: FAS inflates ASR and depresses ACD simultaneously.
Providers frequently describe routes as FAS-free. Treat that as a claim to verify, not a specification. Compare answer timestamps in your own CDRs against billed duration, and look for clusters of very short billed calls that never produced a conversation. A provider willing to discuss how they detect and act on FAS is more useful than one asserting it never occurs.
None of these four means anything alone, and every one of them can be presented favourably in isolation. Compute all four per destination from your own CDRs. Supplier reporting describes the supplier's leg of the call, which is a different question from the one you are asking.
Traffic leaves your platform — a softswitch, dialer, PBX or application — as SIP. Your session border controller authenticates the session and passes it to the provider's ingress. What you send here determines what is available downstream: the calling number, the codec offer, and the headers the next network will read. Anything omitted at this point cannot be recovered later.
The provider's switch decides which upstream carrier receives the call, matching the dialled prefix against available paths. Least-cost routing selects the cheapest option meeting whatever quality floor is configured. This is dynamic and continuous, which is why the carrier chain under a stable contracted rate can change several times in a month without any notification to you.
Before and during the call, the platform applies whatever controls exist: capacity limits, codec negotiation, and monitoring that can withdraw a path performing below threshold. How much of this actually happens varies enormously between providers, and it is worth asking specifically rather than assuming — the difference between monitoring a route and acting on the monitoring is the difference between knowing about a problem and not having one.
The final carrier in the chain hands the call to the network serving the dialled number, which rings the handset. Everything the called party experiences — whether a number displays, how long the silence lasted, whether the call connects at all — was determined by decisions made at the previous three steps.
Minutes are rated per destination against the applicable deck and increment, then invoiced, usually with a settlement cycle short enough to limit credit exposure. Reconcile against your own CDRs every cycle rather than periodically. Discrepancies in wholesale voice are common, frequently explicable, and almost never surfaced by the party they favour.
MCC provisions wholesale voice by traffic profile rather than selling a single undifferentiated product. CLI and Non-CLI presentation are both available, with availability varying by destination — which is a property of the interconnect rather than a preference, so it is confirmed per destination rather than promised in general.
Call-centre and outbound dialer traffic is handled separately through CC Routes, provisioned against dialer type, CPS requirement and destination mix. General outbound voice at network scale runs through Wholesale Voice.
What is worth asking any provider, including this one: which destinations are direct and which are transited, what the CLI delivery rate is for your specific destinations, and what happens operationally when a route degrades. Those three answers tell you more than a rate deck does.
Related reading: CLI vs NCLI routes, Wholesale voice termination, CC Routes.
Routes are quoted against dialer type, CPS requirement, destination mix and CLI or Non-CLI presentation. Confirm coverage for your destinations.