Least Cost Routing in Telecom: How LCR Works

Least cost routing (LCR) is the practice of sending each call to the cheapest carrier that can complete it — matched per destination against a ranked table of supplier rates, on every call attempt. Payments processing uses the same term for routing card transactions to the cheapest network; this guide covers telecom LCR only. In voice, LCR is the economic engine of the wholesale market — and, misused, the single biggest reason “cheap minutes” turn into unanswered calls.

SIPNEX is an FCC-licensed carrier providing wholesale voice termination on its own network. We sit on the receiving end of other people’s LCR engines daily, so this is the carrier’s-side explanation: how the machinery works, and where it quietly destroys quality.

What least cost routing means in telecom

Every wholesale voice provider buys termination from multiple suppliers. Each supplier publishes a rate deck — a table of dialing prefixes and per-minute prices, often thousands of rows deep, with effective dates and billing increments attached. Every supplier prices the same destination differently, and prices change constantly as decks are reissued.

LCR answers the obvious question: for this specific call, right now, which supplier is cheapest? An LCR system holds every supplier’s deck, matches the called number against them, sorts the candidates by price, and hands the call to the top of the list. If that route fails, it hunts to the next one. Multiply that by millions of calls and fractions of a cent become real margin — which is why nearly everyone in the wholesale chain runs some version of it.

How an LCR engine works

An LCR engine — a module inside a class-4 softswitch or standalone least cost routing software feeding one — runs a four-stage pipeline.

1. Rate deck ingestion. Supplier decks arrive as spreadsheets or automated feeds: prefix, rate, effective date, billing increment, sometimes jurisdiction columns for interstate versus intrastate US traffic. The engine normalizes them into one table. This is where quiet billing disasters start — a deck loaded late or a prefix misparsed means the engine happily routes at a price that no longer exists.

2. Prefix matching. For each call, the engine finds the longest matching prefix in each supplier’s deck — the same longest-match logic IP routing uses. For US domestic traffic there is a critical wrinkle: number portability broke the assumption that dialed digits identify the serving network. Serious engines first dip the number’s LRN — the Location Routing Number of the switch that actually serves it — and match on the LRN’s prefix instead. Engines that skip the LRN dip misprice and misroute every ported number, rating it against the wrong carrier and the wrong jurisdiction.

3. Route ranking. The matched candidates are sorted — in a pure LCR engine, strictly by price. Most production systems layer rules on top: exclude suppliers over a loss threshold, pin certain prefixes to certain routes, cap the share of traffic any one supplier receives. The output is an ordered route list for that call.

4. Failover cascade. The call is offered to route one. If the supplier rejects it or fails to connect — a SIP 503, a timeout, no ringback — the engine releases and re-offers to route two, then route three. Failover keeps completion up when a supplier has an outage, but it also adds post-dial delay on every cascaded attempt, which is why a call that “completes fine” through three hops can still take eight seconds to ring.

The quality trade-off: cheapest is usually worst

Here is what the rate deck does not show you. In wholesale voice, price and quality correlate — negatively. The cheapest route is cheap for a reason, and the reasons are rarely good:

• Grey routes. Termination paths that dodge the destination’s licensing or interconnect fees — SIM boxes, unauthorized gateways, disguised traffic. Cheap until they get blocked without notice.
• Over-brokered chains. The “carrier” selling the low rate is an LCR engine reselling another LCR engine reselling a third. Every hop adds post-dial delay, another failure point, and another failover cascade stacked under yours.
• Call looping. When several resellers all buy from each other, a call can be handed around the same circle of switches — burning seconds of PDD before it escapes to a real terminating carrier or dies.
• Degraded signaling and media. Aggressive routes clip ringback, transcode media to save bandwidth, or return answer signals that do not reflect reality, so billing and dialer pacing both drift from the truth.

The symptoms are exactly the metrics operators live by: ASR sinks, ACD gets erratic, and PDD climbs, while audio quality and MOS degrade from transcoding and congested paths. Pure-LCR reselling — always take the cheapest route, whatever it is — is precisely how bargain minute providers destroy answer rates. You save two-tenths of a cent per minute and pay for it in wasted dial attempts and idle agents. Cheap minutes over bad routes are expensive minutes.

Quality-weighted routing: the grown-up version

Mature wholesale operations do not run pure LCR. They run quality-weighted routing: the ranking function scores each route on measured performance — ASR, ACD, PDD, completion stability per destination — alongside cost. A route half a cent cheaper but ten points worse on ASR loses the ranking. Routes get tested continuously, degrading suppliers get demoted automatically, and price only breaks ties between routes that meet the quality floor.

That is the honest defense of LCR: applied across good routes, it is rational purchasing. As a machine for always taking the worst route that technically completes, it is a slow-motion answer-rate fire.

Where LCR sits in a class-4 softswitch

A class-4 softswitch is transit infrastructure — it moves calls between carriers rather than serving end users. Inside it, the LCR engine sits between ingress and egress: a call arrives on a customer trunk group, the routing engine performs the LRN dip and prefix match, produces the ranked route list, and the switch signals out through the chosen egress trunk — cascading on failure while the billing engine writes CDRs against both the customer’s rate and the supplier’s. If you buy from a reseller, this machinery — someone else’s decks, failover logic, and quality thresholds — is what actually decides where your calls go each week.

The question to ask any provider: how do you route?

Every provider will tell you what they charge. Very few volunteer how they route, and that answer determines what the rate is worth. Ask directly: Do you terminate on your own network, or broker across rented routes? How many suppliers sit between your switch and the terminating carrier? Does my routing change week to week as decks reload?

SIPNEX’s answer is structural: we terminate US traffic on our own FCC-licensed network with direct routes, rather than LCR-brokering your calls across whichever reseller was cheapest this week. Sub-3-second average PDD, native G.711u pass-through, and A-level STIR/SHAKEN signed with our own SP-KI certificate are properties of routes we run — not promises inherited from an upstream we do not control. The published volume tiers run from $0.025–$0.030 per minute at entry down to as low as $0.005 at 10M+ minutes per month on 6-second billing — and the route behind the rate does not rotate.

Frequently asked questions

Least cost routing shows you the rate; only live traffic shows you the route. SIPNEX offers wholesale voice termination on direct routes with published tiers, unlimited concurrent channels, and trunks live in 24 hours — put real traffic on our network and compare the ASR, not just the rate.