An HVAC operator in Phoenix showed me his dashboard last month. 847 calls. 312 billable. Sub-threshold rate: 63%. He'd spent $11,000 driving that traffic and billed for less than $5,000 of it.
The immediate assumption — bad traffic. Junk publishers. Click farms. I've made that assumption myself, more times than I want to admit. But when we pulled recordings, the picture was different. Real people. Real intent. Callers who said "I need my AC fixed" and then hung up at 67 seconds.
Sixty-seven seconds. Twenty-three seconds short of billing.
The problem wasn't traffic quality. The problem was everything between "call connects" and "buyer picks up." Long greeting. Two-level IVR. Then a hold queue that averaged 38 seconds. By the time a human said hello, the caller had already waited 55 seconds. Another 35 seconds of discovery — name, address, what's the problem — and you're past 90. But a lot of callers didn't make it that far.
Short-duration calls are the top publisher complaint in pay-per-call. You drove the traffic. You paid for the click. The call connected. And then... nothing billable. This is one of several ways pay-per-call campaigns bleed money — but it's the most fixable once you diagnose it.
This guide walks through the four causes of pre-threshold hangups and what to actually do about each one.
What We're Building
By the end of this, you'll have a diagnostic framework for sub-threshold calls and a prioritized fix list. You'll know:
- How to categorize your sub-threshold calls by root cause
- Which fixes move the needle fastest for each cause
- How to measure improvement without waiting 30 days
The principles apply whether you're on VeloCalls, Ringba, CallRail, or a custom build. The buttons are different; the logic is the same. If you're running home services pay-per-call campaigns, this diagnostic framework maps directly to HVAC, plumbing, and roofing verticals.
Prerequisites
Before you start:
- Access to call recordings for your sub-threshold calls (you need to listen, not just look at dashboards)
- Time-to-agent and hold-time metrics broken out by stage
- Your current IVR flow documented (number of menus, prompt lengths)
- A list of your top 5 publishers by call volume
- 60-90 minutes to actually listen to calls. Nobody likes this. Do it anyway.
Step 1: Pull Your Sub-Threshold Call Distribution
Before you fix anything, you need to know where callers are dropping. Most platforms track disconnect stage:
- Dropped during greeting
- Dropped during IVR menu 1
- Dropped during IVR menu 2+
- Dropped during hold
- Dropped during agent conversation (sub-threshold)
Export this data for the last 30 days. If more than 40% of your sub-threshold calls are dropping during IVR, that's your first fix. If most are dropping during hold or agent conversation, the problem is downstream.
I've seen operators spend three weeks optimizing IVR when their real problem was a buyer who took 45 seconds to answer. I've been that operator. Check the data first. (Sounds obvious. Nobody does it.)
What you should see: A distribution that points to one or two dominant drop stages. If it's evenly spread across all stages, you have multiple problems — start with the earliest stage and work downstream.
Step 2: Categorize Sub-Threshold Calls by Root Cause
Pull 30-50 recordings of calls that ended between 30-89 seconds (assuming a 90-second threshold). Listen to each one. Yes, actually listen. Categorize them:
Category A: Wrong Intent The caller wanted something you don't offer. "Do you guys do commercial HVAC?" (You're residential-only.) "I'm looking for a price quote, not an appointment." "I thought this was [competitor name]."
Wrong-intent calls are a traffic problem, not a routing problem. You can't fix them with IVR optimization. You fix them by auditing the ad creative and landing pages that drove those calls. If a publisher's traffic is 50%+ wrong intent, cut them.
Category B: IVR Friction The caller got frustrated or confused during the menu. Long pauses before pressing buttons. Repeated "I just want to talk to someone." Pressing 0 repeatedly. Hanging up mid-prompt.
IVR friction calls are fixable with menu collapse and shorter prompts. Every menu level costs you ~6% of callers, per our IVR abandonment study.
Category C: Hold Abandonment The caller made it through the IVR, hit the queue, waited... and gave up. These are the painful ones. Qualified intent, correct routing, lost to a stopwatch.
Hold abandonment is a capacity problem. Not enough agents, buyer pickup too slow, or call routing to unavailable buyers. Fix it with callback offers, better availability detection, or simply more agent coverage during peak hours.
Category D: Offer Mismatch The caller reached an agent, started talking, and hung up when they realized the offer wasn't what they expected. "Wait, there's a $89 service fee?" "I thought this was free." "You don't service my area?"
Offer mismatch is a messaging problem. The ad or landing page promised something the agent couldn't deliver. Fix it upstream — tighten ad copy, add geographic qualifiers, clarify pricing before the call. Be aware of TCPA one-to-one consent requirements when adjusting your lead-gen flows.
Tally your categories. If 60% of your sub-threshold calls are Category B (IVR friction), you know where to start. If it's split evenly, you have multiple leaks — but at least now you can prioritize.
Step 3: Fix IVR Friction (If That's Your Dominant Cause)
Most pay-per-call IVRs are overengineered. They collect data the buyer never uses. They ask questions an agent could answer in 10 seconds. They feel thorough. They bleed calls.
Look, I get it — someone built that IVR with good intentions. "Let's pre-qualify so buyers get cleaner calls." Sounds smart. But the math doesn't care about intentions.
The 8-second rule: Your greeting should be under 8 seconds. Everything after 8 seconds is abandonment tax. Cut the company history, the award mentions, the "your call is important to us" filler.
Good: "Thanks for calling ABC Plumbing. For service, press 1. For billing, press 2." Bad: "Thank you for calling ABC Plumbing, serving the greater Phoenix area for over 25 years. Your call is important to us. Please listen carefully as our menu options have changed. For..."
I can already hear the objection: "But we have to say the compliance disclosure." Maybe. But most compliance language is shorter than people think, and most IVR compliance is satisfied by "this call may be recorded" — not a 20-second legal block.
Collapse to one menu. Every additional menu level adds ~6 points of abandonment. If you're running a 3-level IVR, you're losing 18-20% of callers before they hit the queue. That's not a rounding error. That's your margin.
The qualification signals you think you need from menu 2 and 3? Move them to backend routing logic (geo-match on caller ID) or agent discovery (first 15 seconds of conversation). The caller who reaches an agent with one button press beats the caller who hangs up at menu 3.
VeloCalls' visual IVR builder makes this drag-and-drop. On other platforms, you'll need to edit the flow config directly. Either way — simplify first, measure second, iterate.
Step 4: Fix Hold Abandonment (If That's Your Dominant Cause)
Callers who survive your IVR and abandon on hold are the most expensive losses. They were qualified. They were routed. They just... waited too long.
Benchmark: Median hold tolerance is 47 seconds, per our data. At 60 seconds, you've lost 28% of the queue. At 90 seconds, 41%.
If your average hold time exceeds 30 seconds, you have a capacity problem. Fixes, in order of speed-to-implement:
1. Callback offers at 30 seconds. "All agents are busy. Press 1 to receive a callback within 10 minutes, or continue holding."
Callback offers recover 15-25% of callers who would otherwise abandon. But don't offer them if you can't fulfill them — a callback at minute 45 kills trust worse than the original hold. (Ask me how I know. Actually, don't.) Build the callback infrastructure before you offer the option.
2. Buyer availability audits. Are you routing to buyers who are technically "available" but actually at lunch? In a meeting? Already on two calls with no overflow capacity?
Real-time availability checks beat static schedules. If your platform supports capacity-aware routing, use it. If not, shorten your ring timeout (20-25 seconds) and escalate to the next buyer faster.
3. Overflow routing to lower-payout buyers. Better a $35 call that bills than a $55 call that doesn't. If your primary buyer can't answer, route to a backup who can. The payout delta hurts less than a zero.
This feels counterintuitive. You set up that premium buyer relationship, negotiated the rate, and now I'm telling you to route around them? Yeah. Because a call that doesn't bill isn't a call — it's an expense.
For hold-time diagnostics by buyer, JustAnalytics can break down hold and answer performance by traffic source without third-party cookie dependencies.
Step 5: Fix Wrong Intent (If That's Your Dominant Cause)
Wrong-intent calls can't be fixed in the IVR. The caller wanted something you don't offer. They're going to hang up eventually — the only question is whether they do it before or after you've wasted agent time.
Audit your traffic sources. Pull sub-threshold rate by publisher. If Publisher A is running at 35% sub-threshold and Publisher B is at 18%, Publisher A's traffic is the problem. Either their targeting is off, their creative is misleading, or they're sending you calls you never agreed to take.
Check your ad creative. Does the landing page match the service you actually provide? If your ad says "Emergency AC Repair — Call Now" but you don't do same-day service, you're generating calls that will never convert. Be specific about what you offer, where you service, and what it costs. Specificity filters bad intent before the call.
Consider pre-call IVR qualification for high-risk traffic. For publishers with chronic wrong-intent issues, add a single qualification question at menu 1: "For residential service, press 1. For commercial, press 2." If you don't do commercial, route those calls to a "sorry, we don't offer that" recording and hang up. It's a 5-second filter that saves 90-second non-billables. For more advanced scenarios, explore AI voice qualification — though it's not always the right fit.
Traffic quality issues trace back to click quality. If you're running paid search alongside call campaigns, ClickzProtect catches fraudulent clicks before they generate garbage calls — bot traffic and click farms often drive wrong-intent volume.
Step 6: Fix Offer Mismatch (If That's Your Dominant Cause)
Offer mismatch calls are the hardest to diagnose because they feel like qualification failures. The caller reached an agent, talked for 45 seconds, and hung up. Seems like low intent. Usually isn't.
Pull recordings. Listen for the disconnect moment. "Wait, how much?" "You don't service my zip code?" "I thought there was a coupon." These are messaging problems, not intent problems.
Fixes:
- Add pricing signals to your landing page. "Service calls start at $89" filters callers who expect free estimates.
- Add geographic qualifiers. "Serving Phoenix, Scottsdale, and Mesa" prevents calls from Tucson.
- Remove expired offers. If you ran a "20% off" promo in Q1 and it's still on your landing page in July, callers expect a discount that doesn't exist.
This is slow to fix because it's upstream creative work, not routing optimization. But it's often the highest-ROI fix — a landing page change affects every call, not just the ones that hit a certain IVR branch.
Honestly? Most offer mismatch problems trace back to someone in marketing who didn't coordinate with ops. Not throwing stones — I've been that someone.
Common Patterns and Quick Diagnostics
Pattern: Sub-threshold rate is high, but only from one publisher. Diagnosis: Traffic source problem. Audit that publisher's targeting and creative. Consider pausing them while you investigate.
Pattern: Sub-threshold rate spiked suddenly with no routing changes. Diagnosis: Check if a buyer went offline or reduced capacity. Also check if a landing page changed (maybe someone updated copy without telling you).
Pattern: Sub-threshold calls are longest during lunch hours. Diagnosis: Hold time problem. Buyers have lunch coverage gaps. Implement callback offers during 11:30am-1:30pm or route to buyers with staggered coverage.
Pattern: Callers are hanging up at exactly 60 seconds. Diagnosis: This is almost always wrong intent. They waited through the IVR, got an agent, realized it wasn't what they wanted, and bailed. Pull those recordings — you'll hear the "oh, never mind" moment.
Measuring Improvement
Don't wait 30 days to know if your fixes worked. Thirty-day cycles are how you burn a quarter before admitting something isn't working. Track these weekly:
- Sub-threshold rate by stage (IVR vs. hold vs. agent)
- Average time-to-agent (target: under 30 seconds)
- Sub-threshold rate by publisher (catch traffic regressions early)
- Callback offer acceptance rate (if implemented)
Set alerts. If sub-threshold rate increases 10%+ week-over-week, something changed. Catch it before you lose a month of budget.
VeloCalls surfaces most of this in the analytics dashboard. On other platforms, you may need to export to a BI tool. The data exists — the question is whether you're looking at it.
Frequently Asked Questions
Why do callers hang up before the billable threshold?
Four main causes: wrong intent (they wanted something you don't offer), IVR friction (too many menus or long prompts), excessive hold time (buyer didn't pick up fast enough), or offer mismatch (the ad promised something the agent couldn't deliver). Pull recordings of your sub-threshold calls and categorize them. The pattern usually points to one dominant cause — fixing that one thing moves the needle more than optimizing everything at once.
What's a normal sub-threshold call rate in pay-per-call?
Industry ranges run 15-25% of total calls landing below the billable threshold. Under 15% is excellent. Over 30% signals a structural problem — usually a traffic source sending wrong-intent callers or an IVR that filters too aggressively. Track this metric by publisher and buyer; the aggregate number hides where the leak actually is.
Should I lower my billable threshold to capture more calls?
Sometimes, but it's a tradeoff. Lower thresholds increase fill rate but also increase disputes — calls that barely clear 60 seconds often lack qualification signals buyers expect. The better fix is usually improving time-to-agent so qualified callers reach a human faster, rather than lowering the bar on what counts as qualified.
How do I know if the problem is IVR length or buyer answer speed?
Check your time-to-agent metric. If callers reach the buyer queue in under 20 seconds but still hang up at 45-60 seconds, the problem is hold time or buyer pickup speed. If callers are abandoning during the IVR before ever hitting the queue, the problem is menu depth or prompt length. Pull stage-by-stage abandonment data — most platforms track where in the flow the disconnect happened.
That Phoenix HVAC operator? We got his sub-threshold rate from 63% down to 22% in three weeks. Collapsed his IVR from three menus to one, added a callback offer at 30 seconds, and paused two publishers who were sending commercial intent to a residential-only campaign. No magic. Just data and willingness to cut what wasn't working.
If you're bleeding calls before they bill, VeloCalls can help you see where. Smart routing, real-time bidding, visual IVR builder, AI conversation intelligence for HVAC, plumbing, roofing, PI, Medicare, insurance. Per-minute pricing: Managed starts at 4¢/min, BYOC at 2¢/min, drops as you scale.