IVR and routing vocabulary isn't glamorous. Nobody brags about knowing what "barge-in" means. But the gap between operators who understand this layer and those who don't? It shows up in abandonment rates, agent utilization, and — ultimately — whether callers become customers or hang up frustrated.
The call tracking glossary covers attribution. This one covers the routing layer — the IVR menus, queuing logic, and agent distribution that sit between a caller dialing and a human picking up.
IVR Terminology: The Fundamentals
1. IVR (Interactive Voice Response)
The automated system callers interact with before reaching an agent. Press 1 for sales, press 2 for support. IVR handles caller identification, intent capture, and routing decisions.
Good IVR is short — 2-3 prompts max. Bad IVR asks 6 questions, plays 45 seconds of hold music, then transfers to the wrong department anyway. Keep it tight.
2. DTMF (Dual-Tone Multi-Frequency)
The technical name for touch-tone signals. When you press "1," your phone sends two simultaneous frequencies that the system decodes. DTMF is ancient (1960s) but nearly bulletproof — works on landlines, cell phones, VoIP, everywhere.
Every IVR supports DTMF. Not all support speech recognition. When in doubt, use DTMF for critical routing.
3. Speech Recognition
IVR that interprets spoken words instead of (or alongside) keypad input. "Say 'billing' or press 2."
Speech recognition feels more natural but introduces error. Accents, background noise, mumbling, speakerphone — all degrade accuracy. Real-world accuracy runs 85-92%, meaning 8-15% of callers get misrouted or need to repeat themselves. Use speech for simple inputs (yes/no, names) and DTMF for routing decisions where errors cost money.
4. Barge-In
Letting callers interrupt IVR prompts mid-sentence. The system says "For sales, press—" and the caller hits 1 before the prompt finishes. Good IVR supports barge-in. Bad IVR forces callers to wait through the entire prompt.
Barge-in shaves seconds per call. On high-volume systems, those seconds compound into hours of caller wait time recovered.
5. Timeout
What happens when a caller doesn't respond to an IVR prompt. Options: repeat the prompt, route to a default destination, or disconnect.
Two common mistakes: timeouts too short (caller was thinking), timeouts too long (silence feels broken). Most operators land on 5-8 seconds before repeating, 15-20 seconds before escalating to an agent. Test with real callers — not your office team who knows the menu.
6. Containment Rate
The percentage of calls that complete their task within the IVR without reaching an agent. Caller checks their appointment time via IVR and hangs up satisfied? That's containment.
High containment saves agent labor. But forced containment — burying the "talk to a human" option — drives abandonment and angry reviews. The goal isn't max containment. The goal is appropriate containment. See the call routing best practices guide for more on routing design.
7. Self-Service Rate
Similar to containment, but specifically measuring callers who chose to use IVR self-service rather than requesting an agent. A caller who wanted an agent but got stuck in IVR doesn't count as "self-service" — they were contained against their will.
8. Visual IVR Builder
A drag-and-drop interface for designing IVR flows without writing code. VeloCalls offers this — you draw prompts, branching logic, and routing rules visually. Marketing teams can modify flows without filing engineering tickets. Faster iteration, better testing.
9. IVR Tree
The branching structure of IVR prompts and paths. Each caller decision point is a node; each outcome is a branch. Complex IVR trees become unmanageable fast. If your tree has more than 3 levels deep, rethink it.
Call Distribution Methods
10. Hunt Group
A group of agents where calls are distributed in sequence. Call arrives, Agent A's phone rings. No answer in 15 seconds? Try Agent B. Still nothing? Agent C.
Hunt groups prevent calls from sitting unanswered, but can create uneven workloads if Agent A always picks up first.
11. Round-Robin Routing
Distributes calls evenly across agents in rotation. Call 1 goes to Agent A, call 2 to Agent B, call 3 to Agent C, call 4 back to Agent A.
Fairer than linear hunt groups. Agents can't cherry-pick easy calls by answering fast. Unpopular opinion: round-robin is overrated. Sounds fair, but it ignores that some agents are just better. Sending your closers more calls usually beats pretending everyone converts equally.
12. Ring Group (Simultaneous Ring)
All agents in the group ring at once. Whoever answers first takes the call. Minimizes wait time but can overwork the fastest agents.
Some platforms call this "blast ring" or "ring all."
13. Skills-Based Routing
Routes calls to agents based on expertise tags. Spanish-speaking caller? Route to bilingual agents. HVAC call? Route to agents trained on HVAC.
Skills-based routing is table stakes for multi-service contact centers. Running HVAC, plumbing, and electrical calls through the same number? You need skills-based routing or your HVAC callers end up talking to plumbers. And keep the tags current: a new agent nobody tagged quietly misroutes every call they take.
14. Weighted Routing
Assigns different call probabilities to agents or destinations. Agent A gets 50% of calls, Agent B gets 30%, Agent C gets 20%.
Useful for ramping new agents (give them fewer calls while training) or managing contractors who work part-time hours.
15. Geo Routing
Routes calls based on caller location — typically derived from area code, ZIP input, or GPS (on mobile). Phoenix caller routes to Arizona agents. Essential for local service businesses where crews can't drive two hours.
VeloCalls' Smart Call Routing can route by geography.
16. Time-Based Routing (Day-Parting)
Different routing rules for different hours. Business hours go to live agents; after-hours go to voicemail or an answering service. Monday-Friday vs. weekends. AEP (annual enrollment period) vs. off-season for Medicare.
Simple concept, but easy to neglect: forget to update time-based rules when hours change and calls can route to closed queues for weeks before anyone notices.
Queue Management
17. Call Queue
Holds callers in line when all agents are busy. Callers hear hold music or estimated wait time. Queues are better than busy signals — at least the caller knows they're waiting.
18. Queue Depth
The number of callers currently waiting. "Queue depth is 12" means 12 people on hold. High queue depth signals understaffing or a traffic spike.
19. Position Announcement
Telling callers where they are in line. "You are caller number 4." Reduces abandonment — callers tolerate waits better when they know progress is happening.
20. Estimated Wait Time (EWT)
Predicting how long the caller will wait based on queue depth, average handle time, and available agents. Your estimated wait time is 3 minutes.
EWT is tricky. Underestimate and callers get annoyed when they wait longer. Overestimate and they might hang up preemptively. Better to say "5 minutes" and deliver in 3 than vice versa. Hot take: EWT is overrated. Callers don't care about the estimate as much as they care about movement. "You're next" beats "estimated 90 seconds" every time.
21. Callback Queue
Instead of waiting on hold, callers leave their number and get called back when an agent frees up. Preserves leads that would otherwise abandon. Better experience than forcing a 20-minute hold.
The tradeoff: callback systems add complexity and some callers don't trust that you'll actually call back. (Honestly? Given how many businesses don't, I get it.)
22. Overflow Routing
When the primary queue exceeds thresholds (depth, wait time, agent availability), calls overflow to a backup destination. Secondary team, overflow vendor, voicemail.
Configure overflow triggers thoughtfully. Too aggressive and you pay for external agents unnecessarily. Too conservative and callers abandon before overflow kicks in.
23. Concurrency Cap
The maximum number of simultaneous calls an agent (or queue, or destination) can handle. Set it to 1 and agents take one call at a time. Set it to 3 and they can juggle multiple conversations (rare in voice, common in chat).
For buyer destinations in pay-per-call, concurrency caps prevent overload. If a buyer can only handle 5 simultaneous calls, set the cap at 5 — calls beyond that route elsewhere instead of ringing busy.
The reverse mistake is just as costly: a cap set below the number of agents you actually staff turns away callers your team could have answered.
Agent-Side Features
24. Whisper Message
A brief audio prompt played only to the agent before connecting the caller. "Google Ads lead, HVAC, Phoenix." The caller doesn't hear it.
Whispers give agents context without asking "how did you hear about us?" They can personalize the opener: "Thanks for calling about your AC — are you dealing with the heat right now?"
25. Barge (Supervisor Barge)
Lets a supervisor join an active call and speak to both parties. Training tool — supervisor can correct agent mistakes in real-time. Use sparingly; callers don't love finding out someone else was listening.
26. Monitor (Silent Monitor)
Supervisor listens to an active call without the agent or caller knowing. QA and training without interrupting. Usually lawful if you disclose monitoring and recording at call start (which you should be doing anyway), but consent rules vary by state, so check yours.
27. Coach (Whisper Coach)
Supervisor speaks to the agent during a call without the caller hearing. "Offer the 10% discount." Agent hears the coaching; caller doesn't. Less intrusive than barge.
28. After-Call Work (ACW)
The time an agent spends on post-call tasks — entering notes, updating CRM, filling disposition codes — before they're available for the next call. ACW inflates average handle time. Track it separately from talk time.
29. Disposition Code
A tag agents apply post-call to categorize outcomes: qualified, not interested, wrong number, callback requested. Disposition codes feed into QA reports and pay-per-call payout reconciliation. If your dispositions don't match your buyer's expectations, you'll have disputes. The call tracking glossary covers this too.
Technical Terms
30. AMD (Answering Machine Detection)
Technology that identifies whether a call was answered by a human or a voicemail system. In outbound dialing, AMD prevents wasting agent time on voicemails. In pay-per-call, AMD prevents paying for calls that hit voicemail instead of connecting. VeloCalls includes AMD in its AI Conversation Intelligence layer.
AMD accuracy runs 90-97% depending on the vendor and detection method. The remaining 3-10%? False positives (hanging up on real humans) or false negatives (connecting agents to voicemail). No AMD is perfect. False positives hurt most: when the system hangs up on a live person, you lose the lead AND look incompetent to your buyer.
Honorable Mentions
ACD (Automatic Call Distributor) — The system that routes calls to agents. Technically, ACD is the umbrella technology; IVR is the front-end menu, ACD is the back-end routing engine. In practice, people say "IVR" and mean both.
CTI (Computer Telephony Integration) — Connecting phone systems to computers. Screen pops, click-to-dial, CRM integration. CTI is what makes the "caller's record appears when the phone rings" magic happen.
SIP (Session Initiation Protocol) — The protocol that powers VoIP calls. Your calls ride SIP between carriers and platforms. Relevant for BYOC (bring your own carrier) setups where you're connecting your own SIP trunks to platforms like VeloCalls.
Quick Verdict
If you're building or managing IVR flows, these are the terms that matter most:
- Barge-in — callers will hang up if they can't skip prompts.
- Containment rate — track it, but don't force it.
- Concurrency cap — the silent killer of fill rates.
- Skills-based routing — non-negotiable if you handle multiple service types.
- Overflow routing — non-negotiable if you value the callers who arrive during peak volume.
Get those right and you'll avoid the most common IVR and routing mistakes.
And if you're buying paid traffic to drive those calls, ClickzProtect flags the bot traffic in it. No point optimizing IVR for calls that shouldn't have happened in the first place.
Frequently Asked Questions
What's the difference between DTMF and speech recognition in IVR?
DTMF (Dual-Tone Multi-Frequency) is the beep when you press a key — reliable, cheap, works on any phone. Speech recognition lets callers say "billing" instead of pressing 2. Speech feels modern but fails on accents, background noise, and mumbling. DTMF has near-100% accuracy. Speech recognition hovers around 85-92% in real deployments. Most operators use DTMF for critical routing decisions and reserve speech for low-stakes inputs like confirming a name.
What's a good IVR containment rate?
Depends on your intent. For simple tasks (check balance, confirm appointment), 60-80% containment is achievable. For complex inquiries, 30-40% is more realistic. The industry average across all IVR types is roughly 25-35% — meaning 65-75% of callers eventually reach an agent. Don't chase 90% containment if it means burying callers in menu hell. Forced containment drives abandonment and kills conversion.
How does overflow routing work?
When your primary agent queue is full or wait times exceed a threshold, overflow routing sends calls to a backup destination — a secondary team, an outsourced call center, or voicemail. Configure overflow triggers based on queue depth (more than 8 callers waiting), wait time (over 3 minutes), or agent availability (fewer than 2 agents logged in). The mistake operators make: setting overflow too late and losing callers who abandon before it kicks in.
What's the difference between a hunt group and a ring group?
In practice, these terms are often used interchangeably. Technically, a hunt group "hunts" through agents in a defined order (linear or round-robin) until someone answers. A ring group rings multiple agents simultaneously — whoever picks up first takes the call. Hunt groups distribute load more evenly; ring groups minimize wait time but can overload top performers. Some platforms call simultaneous ring a "blast group" instead.
Try VeloCalls for Your Vertical
Pay-per-call platform built for HVAC, plumbing, roofing, PI lawyers, Medicare brokers, and insurance. Smart routing, real-time bidding, visual IVR builder, AI conversation intelligence (transcription, sentiment, summarization, AMD). AI sales agents are on the roadmap — not shipping yet, but coming. Per-minute pricing: Managed starts at 4¢/min, BYOC at 2¢/min, both drop as you scale through lifetime usage tiers.