IVR Abandonment Rate Study: Menu Depth and Hold Time Data (1.8M Calls)

Every additional IVR menu level costs you 6 percentage points of callers. One menu: 6.1% abandonment. Two menus: 12.4%. Three menus: 18.7%. Four levels: 27.3%. The math is brutal and it doesn't care how clever your routing logic is.

We ran this analysis on 1.8M calls through VeloCalls IVR flows between January and May 2026. Anonymized, aggregated, spanning home services, legal intake, insurance, and Medicare verticals. The goal was to answer a question we kept hearing from operators: "Where exactly are we losing callers before they reach an agent?"

The answer is everywhere. But mostly in the first 15 seconds.

Limitation to acknowledge upfront: our data skews toward pay-per-call operators running $10K-200K/month in call volume. Enterprise contact centers above that threshold operate differently. Consumer-facing IVRs (banks, airlines, utilities) are a different animal entirely — our data doesn't touch those.

The Baseline: Where Abandonment Actually Happens

Before we get into the drivers, here's where callers drop across the entire flow.

The greeting and first menu together account for 8.1% of total abandonment. That's before the caller has even expressed intent. If your IVR front-loads a long company introduction or a compliance disclosure nobody asked for, you're bleeding leads before they start.

Hold queue is the biggest single-stage killer — 7.7 points of incremental abandonment. But here's the thing: you control the IVR. You can't always control hold time (that's a capacity problem). Which means the IVR stages are where the highest-impact changes live.

For attribution on which sources drive the best calls into your IVR, JustAnalytics tracks source-level conversion without the third-party cookie problems.

Menu Depth: Every Level Costs You 6 Points

We grouped IVR trees by depth and measured abandonment rate before reaching an agent.

The pattern is almost linear through three levels, then compounds. And median time to agent tracks the same curve — every menu adds 15-25 seconds of caller effort.

Direct-to-agent routing (no IVR menu at all) performs best on raw abandonment. But it has trade-offs: you lose routing intelligence, you can't filter by intent, and your agents handle calls they shouldn't. The sweet spot for most operators we work with is one menu — collect intent, route accordingly, skip the second layer.

Here's an opinion we'll defend: if you have a three-level IVR, you're over-engineering for internal convenience at the cost of caller experience. The routing complexity should live in your backend logic, not in the caller's button presses. Build smarter routing rules. Don't make callers navigate your org chart.

(I'll admit we've built some ugly IVR trees ourselves over the years — the temptation to "just add one more option" is real. It's always easier to add a branch than to rethink the whole flow. That's how you end up with a 4-level monster nobody remembers building.)

Prompt Length: The 8-Second Rule

We measured abandonment rate against greeting prompt duration.

Abandonment rate nearly triples between the 6-8 second range and the 13-18 second range. Past 30 seconds, you're losing one in five callers before they hear a single menu option.

The worst greeting we found in the data: 34 seconds. It opened with the company name, explained their founding year and service areas, mentioned they were award-winning (they didn't say which award), reminded callers they were being recorded, explained the recording would be used for training and quality purposes, then finally got to "Press 1 for sales."

31% abandonment rate on that tree.

The operator couldn't figure out why their call-to-close rate was so bad. They blamed the sales team. They blamed the ad creative. They even blamed their call buyers. The problem was upstream of everyone — it was the IVR prompt their marketing director had recorded three years ago and nobody ever questioned.

Individual menu prompts show the same pattern. Options listed at 2-3 seconds each hold attention. Options that take 6+ seconds (because someone decided to explain what each department does) bleed callers at every step.

The rule we give operators: 8 seconds max for greeting, 4 seconds per menu option. If you can't say it in that window, you're saying too much.

Hold Time: The 47-Second Cliff

Once callers clear the IVR and hit the hold queue, how long will they wait?

Median tolerance is 47 seconds. Which means half your callers who hit a queue will hang up before the one-minute mark.

The curve isn't linear — it accelerates. The first 30 seconds cost you 12 points. The next 30 seconds cost you 16 points. Past 60 seconds, you're losing callers at nearly 1% per 3-4 seconds.

But here's where it gets interesting — and honestly, where we spent way too many hours arguing about methodology. We segmented by vertical and intent signal (captured from the IVR menu selection).

Emergency and high-urgency callers tolerate longer holds — they need help now and they know it. Price-shopping callers (insurance quotes, routine HVAC) bail faster. They'll just call the next number.

This has direct implications for capacity planning. If your mix is 60% quote traffic, your hold tolerance is lower than industry averages suggest. Plan staffing against your actual intent distribution, not a blended benchmark.

For the pay-per-call economics behind these verticals, see our pay-per-call benchmarks report.

The Interaction Effect: Depth × Prompt × Hold

The three variables don't just add — they multiply. We ran a regression on abandonment rate against menu depth, total prompt duration, and hold time.

The simplified finding: a two-level IVR with 18-second total prompts and 45-second hold produces 26% abandonment. The same hold time with a one-level IVR and 10-second prompts produces 14% abandonment. Nearly half the drop-off, same queue wait.

Operators who focus on hold time reduction while ignoring IVR design are optimizing the wrong variable first. Hold time matters — but IVR design is cheaper to fix and often has higher impact.

This drives me nuts, actually. We see operators spending $15K on workforce management software to shave 10 seconds off hold time, meanwhile their greeting prompt is 22 seconds long. Fix the free thing first.

The three interventions in priority order:

1. Cut prompt length. This is free. Record shorter prompts today.
2. Reduce menu depth. Move routing logic to the backend. Let callers press one button.
3. Then attack hold time. Staff to capacity, add overflow routing, implement callback offers.

We watched an insurance call buyer cut abandonment from 19% to 11% in one week by re-recording their greeting (from 22 seconds to 7 seconds) and collapsing two menu levels into one. No staffing changes. No new tech. Just shorter audio files and smarter routing rules in VeloCalls.

What "Good" Actually Looks Like

Based on this data, here are the targets we recommend:

If you're running above the red flag thresholds on any of these, that's where your attention should go first.

An observation that surprised us: operators running BYOC (bring your own carrier) setups through Twilio or Telnyx had slightly better abandonment rates than Managed Carrier users at the same IVR complexity level. The delta was only 1.2 points, but it was consistent.

Our theory: BYOC operators tend to be more technically sophisticated and build tighter IVR flows. The tooling doesn't make the difference — the operator does. (We'd like to think our platform is the magic, but no. It's the people who actually read documentation and test their flows.)

The Hidden Costs of Bad IVR Design

Abandonment rate is the visible metric. The invisible costs are worse.

Caller perception. Callers who sit through a 25-second greeting and two menu levels arrive at the agent annoyed. We don't have conversion data downstream (that's buyer-side), but operator interviews consistently report that callers from high-abandonment IVR flows close at lower rates even when they do connect. The IVR trains them to expect a bad experience.

Attribution distortion. If your IVR abandonment rate is 20%, your source-level conversion data is 20% wrong. The "bad" publisher that converts at 8% might actually convert at 10% — you're just losing a fifth of their calls before intake. Fix the IVR before you blame the traffic. For click-side attribution issues, our sister product ClickzProtect handles fraud detection on the paid search side.

Agent efficiency. Long IVRs don't just lose callers — they slow down the ones who stay. A caller who pressed four buttons and waited 90 seconds is going to spend the first 30 seconds of the agent call re-explaining why they called. Your average handle time goes up. Your cost per resolved call goes up. You've essentially trained the caller to distrust the system before they ever reach a human.

TCPA exposure. This one's non-obvious. If your IVR collects consent ("Press 1 to agree to receive calls from us"), longer IVRs mean more callers abandon before consenting. Which means your follow-up pool shrinks — and the callers who did consent are a biased sample (they tolerated the experience, which may correlate with other behaviors). For more on the consent mess in pay-per-call, see our TCPA one-to-one consent breakdown.

What to Do With This Data

Three actions for this week:

1. Measure your current state. Pull your IVR abandonment rate by stage. If your platform doesn't break it out by stage, switch to one that does (yes, VeloCalls does this). You can't fix what you can't see.

2. Re-record your greeting. Time it. If it's over 8 seconds, cut it. Remove the company history. Remove the unnecessary compliance language (check with counsel — most of what operators include isn't legally required). Get to the first option within 8 seconds.

3. Collapse menu depth. If you have three levels, go to two. If you have two, consider whether you can go to one with smarter backend routing. Every level you remove saves you 6 points of abandonment.

The operators in the bottom quartile of abandonment aren't there because they have better callers. They're not smarter. They don't have fancier tech. They just respect the caller's time in the first 15 seconds. That's a design choice. Make it.

Methodology Appendix

For those who want to verify or replicate.

Data source: 1.8M calls through VeloCalls IVR infrastructure, January 1 - May 31, 2026. Anonymized and aggregated. No individual caller data retained for this analysis.

Verticals represented: Home services (HVAC, plumbing, roofing): 34%. Legal (PI, mass tort): 26%. Insurance (auto, home, Medicare): 28%. Other: 12%.

Abandonment definition: Caller disconnected before reaching a live agent or leaving a voicemail. Does not include callers who reached voicemail and hung up — those are counted as "connected."

Menu depth calculation: Counted by maximum button presses required to reach an agent in the IVR tree. Parallel branches counted at their deepest path.

Prompt length measurement: Automated extraction from IVR flow definitions, verified against a 5% sample of actual audio files. Discrepancy was under 0.4 seconds average.

Limitations: Data skews toward mid-market pay-per-call operators ($10K-200K/month). Enterprise contact centers and consumer-facing IVRs (banking, airlines, utilities) not represented. Vertical weighting reflects VeloCalls user base, not industry distribution.

Frequently Asked Questions

What is a good IVR abandonment rate?

Industry benchmarks put "acceptable" at 5-8%. Our data shows the median across 1.8M calls is 11.3%, but that average hides massive variance. Single-prompt IVRs with under 8-second greetings hit 4.2% abandonment. Three-level menus with 15+ second prompts hit 23%. The "good" rate depends entirely on your IVR design — not some industry standard.

How does menu depth affect IVR abandonment?

Every additional menu level adds roughly 6 percentage points to abandonment rate. One menu: 6.1% abandonment. Two menus: 12.4%. Three menus: 18.7%. Past three levels, abandonment compounds faster — four levels hit 27.3%. If your caller has to press more than two buttons to reach an agent, you're losing a quarter of your leads before anyone picks up.

What's the ideal length for an IVR prompt?

Under 8 seconds for the greeting, under 5 seconds per menu option. Our data shows abandonment rate nearly doubles when greeting prompts exceed 12 seconds. The worst offender we found: a 34-second greeting that explained the company history before listing options. 31% abandonment on that tree.

How long will callers wait on hold before abandoning?

Median tolerance is 47 seconds. At 30 seconds, 12% have already hung up. At 60 seconds, 28% are gone. At 90 seconds, 41%. High-intent verticals like emergency plumbing and PI auto have slightly better tolerance (median 58 seconds), but the curve is still steep. Every 15 seconds past 30 costs you roughly 8 points of abandonment.

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, summaries, AMD. AI sales agents coming soon. Per-minute pricing: Managed starts at 4¢/min, BYOC at 2¢/min, both tier down as you scale.

See pricing → · Book a demo