Coached Calls 101: How to Identify Scripted Traffic Before It Costs You

Every pay-per-call broker has had this experience: a publisher's numbers look great on paper. High qualification rate. Calls hitting duration thresholds. Dispositions coming back as "qualified" from the buyer. Then the buyer calls you — none of those leads are converting. The consumers don't answer callbacks. The applications get abandoned. The sales pipeline is full of ghosts.

You pull the recordings and start listening. The calls sound... fine. Callers say the right things. They answer the qualifying questions correctly. But something is off. The phrasing is too perfect. The answers come too smoothly. There's an unnatural rhythm to the conversation.

You're listening to coached calls. And they've already cost you money.

What Coached Calls Actually Are

A coached call is one where the caller has been fed fabricated information or scripted responses designed to fraudulently qualify them as a lead. The caller doesn't genuinely need the product or service. They're performing a role — saying what they've been told to say to trigger a payout.

This takes several forms:

Real-time coaching. The caller is on the phone with your buyer while simultaneously receiving instructions from a third party (in person, via text, or through a second phone line). When the buyer asks "Do you currently have health insurance?", someone tells the caller to say "No, I lost my coverage last month." The information is fabricated.

Pre-call scripting. Before the call, the caller is given a script or set of answers to memorize. "You live in [zip code]. You have [condition]. You're looking for [specific product]. Your household income is [amount]." The caller recites this information during the call.

Call farm operations. An organized operation where multiple callers work from the same location, making calls across multiple campaigns using provided scripts. This is coached calling at scale — a small call center dedicated to generating fraudulent leads.

The Critical Distinction: Coaching vs. Legitimate Transfer Assistance

This is the most important nuance in call quality control, and getting it wrong in either direction costs money.

A transfer agent helping a caller communicate their real situation is NOT coaching. In many pay-per-call flows, a publisher-side agent speaks with a consumer first, then transfers them to the buyer. During that transfer, the agent might brief the caller: "They're going to ask about your current coverage — just tell them what you told me." That's facilitation, not fraud. The underlying information is real.

Coaching is when the caller is told to lie. The caller doesn't have the condition. They don't live in that zip code. They aren't looking for the product. Someone has told them to claim otherwise because the payout depends on it.

The distinction matters because if you flag every assisted transfer as coaching, you'll burn good publisher relationships and reject legitimate leads. If you fail to flag actual coaching, you'll keep paying for fraudulent traffic. Your QC process needs to distinguish between the two — which requires analyzing what the caller is saying, not just how they're saying it.

Red Flags in the Transcript

Coached calls leave identifiable traces. No script is perfect, and callers who are reciting fabricated information behave differently from consumers with genuine intent. Here's what to listen (and look) for:

Unnatural pauses before qualifying answers. When someone asks your name, you answer immediately. When someone asks a qualifying question and the caller pauses for 3-4 seconds before responding with a precise, specific answer, they may be receiving instructions. This is especially telling for questions that should have instant answers — "What state do you live in?" shouldn't require thinking time.

Verbatim phrasing across multiple calls. One caller saying "I'm looking for a comprehensive plan that covers my pre-existing conditions" is normal. Twelve callers from the same publisher using that exact phrase is a script. Real consumers describe their needs in their own words, with natural variation. Coached callers repeat what they've been told.

Inconsistency under follow-up questions. A coached caller can handle the expected qualifying questions because those are scripted. When the buyer's agent goes off-script — asking for details, probing deeper, or approaching the same topic from a different angle — coached callers stumble. Their details don't add up, or they default to vague responses when specifics would be natural.

Mismatch between caller knowledge and claimed situation. A caller claims to have been dealing with a specific condition for years but can't name their current medication or their doctor. A caller says they own their home but doesn't know basic details about the property. The information is fabricated, and the gaps show.

Background audio cues. Multiple callers from the same publisher with identical background noise — the same ambient sounds, the same room acoustics — suggests a call farm. This is harder to catch from transcripts alone but obvious in recordings.

Mechanical cadence. Real conversations have natural rhythm — interruptions, self-corrections, "um" and "uh," tangents. Coached calls often have an unnaturally clean flow. The caller answers, waits, answers, waits. There's no conversational texture.

The Financial Impact

Coached calls are the most expensive form of pay-per-call fraud because they're designed to pass every surface-level quality check.

Consider the numbers. In a health insurance vertical paying $80 per qualified call, a publisher running a coached call operation can send 50 calls per day. If 70% qualify (because they're scripted to qualify), that's 35 payouts at $80 each — $2,800 per day in fraudulent payouts. Over a 30-day billing cycle, that's $84,000 from a single publisher.

The buyer sees none of these "leads" convert. They're paying for transfer time, agent hours, and follow-up attempts on consumers who never had genuine intent. When the buyer disputes or demands clawbacks, the broker is caught in the middle — money already paid to the publisher, revenue clawed back by the buyer.

Even if you catch the fraud after two weeks and cut the publisher, you've already lost $42,000 in payouts that will never be recovered. The publisher knew this was the timeline. They planned for it.

This is why detection speed matters more than detection accuracy. Catching coached calls on day two instead of day fourteen is the difference between a $5,600 loss and a $42,000 loss.

Why Manual QC Misses Coached Calls

The fundamental challenge with coached calls is that they're designed to fool listeners. A QC analyst reviewing a single coached call in isolation will often mark it as legitimate. The caller said the right things. The call met duration requirements. The disposition was correct.

Coached calls become detectable through patterns, not individual review:

• The same phrasing appearing across multiple calls from one publisher
• Qualification rates that are statistically improbable (90%+ when the network average is 40%)
• Caller response patterns that are too consistent — real consumers show natural variation
• Clusters of calls with similar duration, similar cadence, and similar answers

A human QC analyst reviewing calls sequentially might notice these patterns if they happen to review enough calls from the same publisher back-to-back. But in a typical QC workflow, calls are reviewed in random order across all publishers. The pattern never becomes visible because the data points are scattered.

Even dedicated analysts reviewing a single publisher's traffic can only listen to so many calls. At 10-15 minutes per call (listening, documenting, scoring), an analyst can thoroughly review 30-40 calls in a day. If the publisher is sending 100 calls daily, most go unreviewed.

Catching Coached Calls at Scale

Identifying coached calls reliably requires two things manual QC can't provide: full coverage and per-publisher visibility.

Full coverage means analyzing every call, not a sample. When every call is transcribed and evaluated, coached call patterns can't hide in the unreviewed majority. Even sophisticated operations that vary their scripts slightly across callers will produce detectable flag rates when the entire dataset is analyzed.

Per-publisher visibility means seeing flag rates, conversion rates, and call quality metrics broken down by traffic source. When you can see that Publisher X has a 15% coached call flag rate while your network average is 3%, the problem is obvious — even if each individual flagged call required judgment to evaluate.

ConvoQC was built specifically for this problem. Every call is transcribed and analyzed individually for coached call indicators, compliance issues, and other red flags. The dashboard then aggregates flag rates by publisher — so when one source's coached call rate starts climbing, you see it in the data before it becomes a six-figure problem.

The difference between catching coached calls quickly and catching them weeks later isn't a QC improvement — it's the difference between a manageable write-off and a financial hit that reshapes your quarter.