Route Quality, Measured Obsessively and Shared
Carriers all claim great routes; the difference is whether they can show you the numbers. An AI system manages route quality at DIDHub: every route is scored continuously on live ASR, NER, ACD, PDD, MOS and CLI delivery, drift and FAS are caught in minutes, traffic shifts automatically, and the same KPIs appear in your Analytics reports, per destination, with trends and exports. Observability isn't an enterprise add-on here; it's the default.
1. The metrics that define a route
| Metric | What it measures | Healthy looks like | What bad means |
|---|---|---|---|
| ASR (Answer-Seizure Ratio) | % of attempts that get answered | 40–60%+ conversational; dialer traffic naturally lower | Sudden drops = route failing or numbers filtered |
| NER (Network Effectiveness Ratio) | % of attempts the network handled correctly (answer, busy, no-answer all count as success) | ≥95% | The purest signal of route failure; it strips out human behavior |
| ACD (Average Call Duration) | Mean billable duration of answered calls | Stable per traffic type | Collapsing ACD = audio problems or FAS contamination |
| PDD (Post-Dial Delay) | Seconds from dial to ringback | <3s; <2s excellent | Long PDD kills dialer pacing and AI-agent UX |
| MOS (Mean Opinion Score) | Computed audio quality (jitter/loss/codec) | ≥4.0 on G.711 | Degraded media path despite calls connecting |
| CLI delivery | % of probe calls where the A-number arrived intact | ~100% on CLI grades | A “CLI route” that isn't; re-graded automatically |
| FAS incidents | False-answer signatures detected | Zero | Route billing dead air; pulled and credited |
2. The AI quality manager
Routes don't degrade politely on schedule, so route quality at DIDHub is managed by an AI system rather than a shift roster. It consumes the live telemetry stream (every CDR, every probe result, every SIP failure code) and acts on it:
- Continuous scoring with fast decay: a route's rank reflects its last minutes, not its reputation.
- Anomaly detection tuned to wholesale-voice failure shapes: ASR cliff vs slow drift, ACD compression that fingerprints FAS, PDD inflation that precedes capacity exhaustion, CLI-delivery drops on a single breakout.
- Automatic remediation: de-rank, drain, shift traffic to the next-ranked route in the LCR portfolio; re-admit recovered routes gradually, on probation.
- Evidence-attached escalation: carrier tickets open automatically with the telemetry that proves the fault.
Probe calls complement passive monitoring: automated test calls through every CLI-graded route to in-country numbers we control verify the A-number, the audio path, and the answer signal honestly, including on routes currently carrying little of your traffic, so the portfolio is measured even where it's idle.
4. Why this is rare
Most wholesale voice runs on the opposite model: quality data is held close, problems are acknowledged only after escalation, and the NOC's job is managing perception. Sharing per-destination KPIs is uncomfortable unless the routes genuinely hold up, which is precisely why we share them. The portfolio is built on direct, licensed interconnects, managed by software that never sleeps, and audited by your own dashboard.
Send us your traffic profile
Destinations, monthly minutes, CLI requirements. We'll return a route plan and rate deck within one business day.
FAQ
What KPIs does DIDHub share with termination customers?
ASR, NER, ACD, and PDD per destination, traffic and spend analytics with hourly heatmaps, quality/reroute events affecting your destinations, and CSV/API export of rated CDRs and KPI aggregates, all in the customer dashboard's Analytics reports.
What is FAS (False Answer Supervision)?
A route fraudulently signaling 'answered' before the called party actually picks up (during ringback or to a dead-air prompt), so billing starts early. Our AI detects FAS signatures (instant answers, compressed uniform ACDs, silent media), pulls the route automatically, and credits affected calls.
How fast do you react to a degrading route?
The AI works on continuously-decaying scores over live telemetry; meaningful drift triggers de-ranking and automatic traffic shifting typically within minutes, before threshold-based alarms or customer complaints would fire.
What's the difference between ASR and NER?
ASR counts only answered calls as success, so it's polluted by human behavior (busy, no-answer). NER counts every correctly-handled attempt (answer, busy, ring-no-answer) as success, isolating pure network performance. A route with collapsing NER is broken; collapsing ASR alone might just be cold-call reality.
Do you monitor routes you're not currently using?
Yes: probe calls keep scoring the standby portfolio so failover targets are measured before they're needed, not discovered broken during an incident.
Route your outbound through carrier-grade A–Z termination
Direct interconnects with some of the world's largest carriers, quality-weighted LCR, AI-managed route health, and the KPIs to prove it, shared with you rather than hidden.