Termination is the part of a call nobody sees: the handoff onto the network that serves the number being dialled. Everything else in wholesale voice — the rate decks, the daily repricing, the arguments about answer rate — sits on top of that one transaction.
By Vedant Vhanyalkar
When a call leaves your platform, it has to arrive somewhere. That somewhere is the terminating network — the operator that serves the dialled number. Wholesale voice termination is the business of getting it there.
You hand traffic to a provider. The provider either interconnects with the terminating operator directly, or passes the call to another carrier who does, or to a carrier who passes it on again. Somewhere at the end of that chain the call rings a handset, and each party in the chain takes a margin on the minutes.
That chain is the whole subject. Its length determines your rate, your post-dial delay, whether your calling number survives, and how much you can find out when something breaks.
An A-Z deck quotes every destination a provider carries, alphabetically from first to last. It is a coverage statement and a price list. It is not a quality statement, and it is routinely read as one.
The same provider can offer a premium direct route to one country and a heavily transited route to its neighbour, on the same deck, at rates that look proportionate. Nothing in the format tells you which is which.
Rates are quoted against dial-code prefixes, not country names. One country typically appears many times: fixed-line separately from mobile, and often each mobile operator separately again, because termination costs differ per operator. A deck showing one blended rate for a whole country is either a simplification or a route you should test carefully.
Two decks quoting the same per-minute rate can bill differently depending on increments. Per-second billing after an initial period costs less on short calls than 60/60 billing, where every call rounds to a full minute. For dialer traffic with short average duration, the increment can matter more than the headline rate.
Read the increment before comparing rates. It is usually a footnote and it is frequently the difference between two decks that look identical.
A commercial and technical relationship with the terminating operator, no carrier in between. One hop means fewer places for the call to be altered: calling number usually survives, post-dial delay is lower, and when something fails there is one party who can explain why. It costs more, and no provider is direct everywhere.
The call crosses one or more intermediate carriers before it reaches the terminating network. This is the majority of international wholesale traffic and it is not inherently bad — it is how coverage gets built without every carrier interconnecting with every other. What it costs you is visibility. Each hop can rewrite signalling, and you are unlikely to know how many hops there are.
Least-cost routing reselects continuously. Your rate can hold steady while the carrier chain underneath changes several times in a month. This is normal market behaviour and it is also the most common reason a route that tested well in week one performs differently in week six.
The practical response is to monitor continuously rather than qualify once. See CLI vs NCLI routes for how the same effect shows up in caller ID delivery specifically.
Answered calls as a share of attempts. The headline metric, and the easiest to move without improving anything. ASR read alone tells you very little, because several mechanisms raise it without producing conversations.
The average length of answered calls. Read it against ASR, always. High ASR with very low ACD is the classic signature of calls being answered by something that is not a person, or of calls being cut shortly after connection.
The gap between dialling and ringing. Long PDD usually means the call is crossing more hops than the route description suggests. It also costs you answers directly: callers abandon during silence, so PDD depresses ASR on the same route it is diagnosing.
Like ASR, but counts calls the network delivered successfully even when nobody answered — busy, no reply, unavailable. It separates network failure from human non-answer. When ASR is poor but NER is healthy, the network is doing its job and the problem is on the campaign side.
Ask for all four, per destination, and compute them from your own CDRs. Supplier reports describe the supplier's leg of the call.
Wholesale voice is a thin-margin, high-volume business. Understanding where a provider's margin sits explains most of their behaviour toward you as a buyer.
Buy at one rate, sell at another, keep the difference. On competitive destinations that spread is small, which is why volume commitments and payment terms carry as much weight in negotiation as the rate itself.
Providers hold multiple upstream options per destination and select dynamically. Margin comes from routing to the cheapest acceptable path at any moment. Your quality experience is a consequence of where "acceptable" is set — which is a commercial decision, not a technical one.
Nobody is direct everywhere. A provider's real product is the combination: direct where it matters to their customers, transit elsewhere, assembled so the deck covers A to Z. When you evaluate a provider, you are evaluating that assembly, not a single route.
Termination carries risks that do not appear on a rate deck and often surface only after an invoice. Two matter more than the rest.
IRSF works by generating traffic to expensive number ranges where the fraudster shares in the termination revenue. It is a wholesale problem specifically because the traffic looks legitimate at every hop and the money is real. If your platform is compromised, or a customer's account is, you are liable for minutes you never intended to send.
Practical defences are unglamorous: destination allowlists rather than blocklists, spend velocity caps per account, and alerting on sudden shifts in destination mix. Ask a prospective provider what fraud controls they apply on their side and whether they will hold traffic on anomaly — the answer varies more than you would expect.
Artificially inflated traffic to particular ranges shows up as a jump in attempts with unusually short ACD. It distorts your quality metrics before it distorts your bill, which is one more reason to read ASR and ACD together rather than in sequence.
Obligations attach to where the call lands, not where it originated. Caller ID rules, authentication frameworks such as STIR/SHAKEN, consent requirements and calling-hour restrictions are all national. A configuration entirely lawful for one destination can be non-compliant for the next country on the same deck.
Build the destination list around what you can comply with, then price it. Doing it the other way round produces routes you cannot legally use.
The most common and most expensive mistake. A route 20% cheaper that answers 30% less is not cheaper. This is arithmetic rather than opinion, and it is skipped constantly because rate is visible on day one and answer rate is not.
Paths change underneath a stable rate. A route accepted in January on a good test can be a different route by March with no notification, because nothing in the commercial arrangement changed. Continuous measurement is the only way to see it.
Supplier reports describe the supplier's leg. They are not wrong so much as partial. When your CDRs and their report disagree, both can be accurate about different things — and only yours reflects what your customers experienced.
Single-sourcing termination means a route problem is an outage. Most operations at volume run at least two providers per key destination with the ability to shift traffic quickly, and accept a slightly worse blended rate for that option.
A deck negotiated against last year's traffic profile stops fitting as the profile moves. Re-examine the mix quarterly. The destinations that now carry your volume are frequently not the ones the agreement was priced around.
Route quality is destination-specific and often operator-specific inside a destination. A provider excellent on your top country can be mediocre on your second. Test the mix you actually send, weighted the way you actually send it.
Compare providers by sending identical traffic to identical destinations at the same times. Sequential testing measures the hour as much as the route — international voice quality varies by time of day, and a week apart is not a comparison.
A useful question: can you tell me the CLI delivery rate for this destination? A provider who can answer has visibility into the path. A provider who cannot usually has more hops than they have described.
It will drop; paths change. The question is what follows. Who is notified, how quickly can traffic be moved to an alternate route, and is there a commitment or only an intention. Settle this before signing, because afterwards it is a negotiation rather than a term.
Qualify on a subset of destinations and a fraction of volume. Wholesale relationships are easy to expand and awkward to unwind, and the first month of live traffic tells you more than any amount of pre-sales testing.
Related reading: CLI vs NCLI routes, CC Routes, What Is SIP Trunking?.
MCC provides wholesale voice termination and CC routes for carriers, ITSPs and platforms. Coverage and route type vary by destination — confirm yours.