Setting Up Call QC on Retreaver

You're running traffic through Retreaver, and you've already set up the basics: scrub lists, call caps, duplicate blocking, maybe some tag-based routing rules. Retreaver gives you solid caller-level controls. You can block numbers, filter by geography, and set duration thresholds before payouts fire.

But here's the gap. A caller passes every filter Retreaver has — valid number, right area code, not a duplicate, not on your scrub list — and the call connects to your buyer. The buyer's agent asks qualifying questions. The caller answers perfectly. Duration hits 90 seconds. Payout fires.

Three weeks later, the buyer calls you. None of those leads converted. The callers don't answer follow-ups. The applications are abandoned. The traffic looked clean in Retreaver because it was clean at the caller level. The problem was in the conversation — and Retreaver wasn't listening to it.

This is the difference between caller-level quality control and conversation-level quality control. If you're relying on Retreaver's built-in tools alone for Retreaver call quality control, you're covering half the problem.

What Retreaver Catches (and What It Doesn't)

Retreaver is good at what it does. Its fraud prevention tools operate at the caller and traffic level:

Caller-level controls. Duplicate detection, number scrubbing, block lists. If a number has been flagged or has already called, Retreaver can block it before it reaches a buyer. This stops the most obvious abuse — recycled callers, known bad numbers, robocall patterns.

Traffic-level rules. Call caps per publisher, geographic filtering, time-of-day routing, minimum duration thresholds. These let you shape the traffic before payouts fire. A publisher can't send 500 calls in an hour if you've capped them at 50.

Tag-based routing. Retreaver's tag system lets you route calls based on caller attributes and campaign criteria. You can segment traffic, prioritize sources, and control which buyers get which calls.

All of this is necessary. None of it tells you what happened on the call.

A coached call passes every one of these filters. The caller's number is clean. They're calling from the right state. They hit duration thresholds because they were scripted to stay on the line. They "qualify" because someone told them exactly what to say. Retreaver sees a successful call. Your buyer sees a lead that will never convert.

Similarly, compliance violations that happen during the conversation — a caller saying "stop calling me," a caller who clearly doesn't match the campaign vertical, an agent making unauthorized claims — don't generate any signal in Retreaver's system. These issues live in the audio, not the metadata. And without Retreaver call monitoring at the conversation level, they accumulate silently until a buyer complains or a regulator notices.

The Two Layers of Call QC

Think of it as two distinct layers:

Layer 1: Caller-level QC. Is this a legitimate caller? Have they called before? Are they from the right geography? Is their number on a block list? Retreaver handles this well.

Layer 2: Conversation-level QC. What did the caller actually say? Did they genuinely qualify, or were they coached? Were there DNC or TCPA indicators in the conversation? Did the call have real commercial intent?

Layer 1 prevents bad calls from connecting. Layer 2 evaluates what happened on calls that did connect. You need both. Most operations running Retreaver have Layer 1 covered and Layer 2 missing entirely.

The result is a blind spot that looks like this: your Retreaver dashboard shows healthy traffic — calls connecting, durations hitting thresholds, payouts firing. But your buyer conversion rates are declining, and you can't explain why because everything looks fine in the data you have.

The explanation is in the recordings nobody's listening to.

How Conversation-Level QC Works

Automated Retreaver QC automation at the conversation layer follows a straightforward process:

1. A call ends on Retreaver and a pixel fires, sending the recording URL and call metadata to the QC system.
2. The recording is transcribed — full text of everything said on the call.
3. AI analyzes the transcript for red flags: coached call indicators, compliance issues, DNC signals, TCPA concerns.
4. Results are stored: transcript, summary, disposition classification, and any detected flags.
5. QC results are posted back to Retreaver as tags on the call record.

Each call is analyzed independently on its own merits — the same judgment a human QC analyst would apply, but applied to every call instead of a sample. Processing is asynchronous; results appear within minutes of the call ending, not during the call. For QC purposes, same-day detection is what matters. You catch a bad publisher on day one instead of day fourteen.

The flags that matter in pay-per-call:

Flag	What It Catches
Coached Call	Caller is being fed fabricated information or scripted responses to fraudulently qualify
Compliance Issue	Caller doesn't match campaign qualifiers — wrong vertical, demographic, state, or misrepresentation
DNC Violation	Indicators in the conversation suggesting a Do Not Call registry issue
TCPA Violation	Potential Telephone Consumer Protection Act concerns

Setting Up QC Pixels on Retreaver

Retreaver uses per-campaign tracking pixels — unlike TrackDrive's single global webhook, you add the QC pixel to each campaign individually. This means a few extra minutes per campaign, but the process is the same each time.

Here's the setup using ConvoQC, which has a native Retreaver integration with postback support.

Step 1: Get Your API Key

Sign up at ConvoQC and grab your API key from the integrations page. You get $10 in free credit — enough to process roughly 10 hours of call recordings at $0.015/minute before committing.

Step 2: Add the Pixel to Your Campaign

In Retreaver, navigate to Campaigns → [Your Campaign] → Pixels. Create a new standard GET pixel.

Important: Do not use Retreaver's Webhook Configurator (the Ping URL / Headers / Data option). Retreaver's bot blocks requests to major platform domains when using that method. Use the standard GET Pixel URL instead — it works reliably and carries all the data you need as query parameters.

Set the pixel trigger to "If call reached a buyer." This fires after the call ends for any call that connected to a buyer, which gives you full QC coverage of connected calls. Retreaver also offers triggers like "If call converted" or "If call didn't convert" if you want to narrow the scope, but "If call reached a buyer" captures everything.

Step 3: Configure the Pixel URL

The pixel URL includes your API key and Retreaver's token placeholders. Retreaver replaces these tokens with actual call data when the pixel fires:

https://dash.convoqc.com/api/analyze?api_key=YOUR_API_KEY&caller_number=[caller_number]&duration_seconds=[call_duration]&recording_url=[call_recording_url]&retreaver_call_id=[call_uuid]&campaign_name=[campaign_name]&call_timestamp=[call_start_time]&buyer_name=[buyer_name]&postback_key=[call_key]

Here's what each token sends:

Retreaver Token	What It Sends	Why It Matters
[caller_number]	Caller's phone number (E.164)	Identifies the caller, cross-references with DNC indicators
[call_duration]	Call duration in seconds	Contextualizes flags — short calls are different from long calls
[call_recording_url]	URL to the call recording	The actual audio that gets transcribed and analyzed
[call_uuid]	Retreaver's unique call ID	Links the QC result back to the specific call in Retreaver
[campaign_name]	Campaign name	Groups QC results by campaign for reporting
[call_start_time]	Call start time (Unix timestamp)	Ties the QC result to the correct call in your timeline
[buyer_name]	Buyer who received the call	Useful for buyer-specific quality reporting
[call_key]	Per-call postback key	Enables writing QC results back to Retreaver (see below)

Note that [call_start_time] returns a Unix timestamp (seconds since epoch), not an ISO date string. The QC system handles the conversion automatically.

Step 4: Save and Test

Save the pixel. The next call that connects to a buyer on this campaign will fire the pixel. Within a few minutes, the call will appear in your QC dashboard with a full transcript, AI-generated summary, disposition label, and any detected flags.

Step 5: Repeat for Each Campaign

Since Retreaver pixels are per-campaign, you need to add this pixel to each campaign you want QC coverage on. Copy the same pixel URL — the campaign_name token ensures results are tagged with the correct campaign automatically.

If you're running 10 campaigns, this takes about 15 minutes total. The main risk is forgetting to add the pixel when launching a new campaign. Build it into your campaign launch checklist.

Postback: Writing Results Back to Retreaver

This is where the [call_key] token pays off. After analysis, QC results are posted back to Retreaver as tags on the original call record. Three tags are written:

• convoqc_disposition — the AI-classified disposition (Qualified, Not Interested, Voicemail, etc.)
• convoqc_flags — any detected flags, comma-separated (e.g., "coached_call, compliance_issue")
• convoqc_summary — a brief summary of what happened on the call

This means you don't have to switch between systems for day-to-day operations. The QC data lives in both your QC dashboard for detailed analysis and your Retreaver call records for quick reference. You can use Retreaver's tag system to filter, sort, and report on QC results alongside your existing data.

What Changes After Setup

The first week of running conversation-level QC on Retreaver traffic is usually revealing.

Your "clean" publishers might not be clean. A publisher that looks great in Retreaver — calls connecting, durations hitting thresholds, no scrub list hits — might have a 6% coached call flag rate when you actually listen to the conversations. That's not visible in caller-level data. It only shows up when you analyze what's being said.

Short calls get context. Every operation has a stack of 30-60 second calls that get ignored. They didn't hit duration thresholds, so no payout fired and nobody cares. But transcribing them often reveals useful signal. A 40-second call where the caller immediately asks "who is this? why are you calling me?" tells you something about that publisher's traffic generation methods.

Publisher conversations become specific. Instead of "we're seeing quality issues," you can say "14 of your 180 calls last week were flagged for coached call indicators — here are the call IDs and transcripts." That's a conversation backed by evidence. It's harder to dismiss than a vague complaint, and it gives the publisher a chance to actually investigate and fix the problem.

Buyer confidence improves. When you can proactively share quality data with buyers — "here's your traffic quality summary for the month, we caught and addressed two publisher issues" — the relationship changes. You're demonstrating that you're monitoring quality, not just routing calls and hoping for the best.

Retreaver-Specific Considerations

A few things to keep in mind that are specific to Retreaver's architecture:

Postback timeout. Retreaver has a configurable "Postback Timeout" (default 600 seconds). It waits for a postback or for the timeout to expire before firing post-call pixels. If your QC pixel fires on "If call reached a buyer," be aware that Retreaver may hold the pixel until the postback window closes. This doesn't affect QC processing — the call still gets analyzed — but it means there may be a slight delay between the call ending and the pixel firing.

Tag filters on pixels. You can add conditions to your pixel triggers, like call_duration >= 30, to skip very short calls that are unlikely to have meaningful conversation content. This saves processing costs on calls that are just hangups or wrong numbers. Any trigger type supports tag filters.

Duration-based triggers. If you use "If outbound call meets seconds timer" triggers for other purposes, be aware that only the highest matching timer fires. If you have timers at 30 seconds, 60 seconds, and 120 seconds, a 150-second call only fires the 120-second trigger. For QC purposes, "If call reached a buyer" is simpler and more reliable.

Caller-Level + Conversation-Level = Full Coverage

The strongest QC setup on Retreaver uses both layers together.

Keep using Retreaver's built-in tools for what they're good at: blocking bad numbers, enforcing call caps, filtering by geography, managing scrub lists. These prevent obviously bad calls from connecting in the first place.

Add conversation-level analysis for what Retreaver can't see: what's actually being said on the calls that do connect. Coached calls, compliance violations, caller intent, lead quality — all of this lives in the audio, and the only way to evaluate it at scale is to transcribe and analyze every call.

Neither layer alone is sufficient. A caller can pass every filter in Retreaver and still be reading from a script. A caller can fail a scrub list check and never connect, which is exactly what you want. The two systems handle different parts of the problem, and the combination closes the gap that either one leaves open on its own.

The Bottom Line

If you're running traffic through Retreaver and relying solely on caller-level controls for quality, you have visibility into who's calling but not what they're saying. Coached calls, compliance violations, and low-quality traffic pass through cleanly because the fraud is in the conversation, not the metadata.

Per-campaign pixel setup takes a few minutes per campaign. Processing runs at $0.015 per minute of audio. Results post back to Retreaver as tags so the data lives in both systems. ConvoQC handles the transcription, analysis, and postback automatically — create a free account and the $10 signup credit gives you enough runway to test with real traffic and see what your call recordings actually contain.