Why your ViciDial contact rate is dropping in 2026

If you’re running a ViciDial predictive dialer in 2026 and your contact rate has been sliding, you’re probably looking in the wrong places. Agents aren’t slacking. The list isn’t worse. The problem is three structural issues in the call delivery chain — and they compound each other.

Based on AMDY’s production network across billions of calls, only 12.5% of answered outbound calls reach a live human. Another 14% are carrier false-answers — the network signals “connected” when nobody actually picked up. The rest are voicemail, IVR, intercepts, and spam traps. If your AMD can’t see all of that clearly, you’re not just wasting calls — you’re actively hurting yourself.

Problem 1: Carrier labeling is killing your answer rate before AMD runs

The most upstream problem is one that AMD can’t fix: your caller ID is labeled “Spam Likely” or similar before the phone even rings, and prospects simply don’t answer. STIR/SHAKEN attestation, carrier analytics platforms like TNS and Hiya, and community-sourced spam reports all feed into this. The result is that a DID with a poor reputation can have answer rates 30–60% lower than a clean number — and no amount of AMD tuning recovers calls that were never answered.

The STIR/SHAKEN framework requires that calls carry a signed attestation level (A, B, or C) that indicates how well the originating carrier can verify the caller ID. Many VoIP-originated calls get B or C attestation, which carriers on the terminating side use as a risk signal. Combine a low attestation with high call volume from a number and it gets flagged fast.

AMD is connected to this problem in a way most operators don’t expect: aggressive AMD that drops real humans without connecting an agent creates high “ghost abandonment” — calls that answered but nobody spoke. Carriers track this and it accelerates flagging. How bad AMD wrecks your caller-ID reputation →

What this means practically: you need to monitor your DID reputation actively, rotate flagged numbers, and reduce the behaviors that trigger flagging. That last part — reducing ghost abandonment — is where accurate AMD helps.

Problem 2: Carrier false-answer supervision (FAS) is burning your agent capacity

Carrier false-answer supervision, or FAS, is what happens when the carrier network signals “call answered” but no human or machine actually picked up. The call is connected to your dialer. Your AMD sees some audio — often silence, a tone, or a brief carrier recording — and doesn’t know what to do with it. If it calls it a machine, you may leave a pointless voicemail. If it calls it a human, an agent gets connected to dead air.

On the AMDY network, 14% of answered calls are carrier false-answers. There are five distinct FAS patterns — different carriers produce different audio signatures. Stock heuristic AMD, which relies on timing and syllable counting, can’t reliably tell the difference between a real voicemail and a carrier intercept. It misclassifies both, and the errors flow downstream. Carrier false answers: paying for connections that never happened →

The cost is real. A 14% FAS rate on a dialer running 500 calls an hour means 70 phantom connections per hour. If each wastes 15 seconds of an agent’s time waiting for someone to speak, you’re burning more than 17 agent-minutes every hour on calls that were never real. Multiply that across a shift and a team, and it shows up as lower contact rate and lower talk time per agent-hour, with no obvious cause in the data.

Problem 3: Poor AMD accuracy creates “ghost abandonment” that compounds everything

Heuristic AMD in Asterisk — the engine behind ViciDial’s built-in detection — works by measuring silence duration, counting syllables, and watching for long greetings. When it works, it works. The problem is that real-world call audio in 2026 breaks its assumptions constantly.

A homeowner who picks up and immediately says “Hello? Who’s this?” without a pause triggers the wrong classification — the AMD sees a short, fast utterance and calls it a machine. The call is dropped. From the prospect’s perspective, somebody called, they answered, and heard silence. They hang up and that number is done. You never knew they were real.

This problem is invisible in most ViciDial reporting. The call shows as “answered — AMD machine” and gets archived. There is no count of live humans who were dropped. The TCPA angle makes it worse: the FTC’s 3% cap on abandoned calls is measured against answered calls. When AMD mis-classifies live humans as machines and drops them, those are counted as abandoned calls if they were truly live — eroding your compliance headroom without showing up anywhere obvious. How the 3% rule works in practice →

How the three problems compound each other

Each problem alone is manageable. All three together create a loop that gradually destroys your operation.

• Heuristic AMD drops live humans → agent doesn’t speak → call abandoned → carrier flags your DID faster.
• DID gets flagged → fewer prospects answer → the proportion of FAS and machine calls in your answered pool rises → AMD errors on humans matter more.
• FAS burns agent capacity → fewer real connects per agent-hour → managers push higher dial rates → more calls per DID → DID gets flagged faster.

The downstream signal you see is a falling contact rate and declining agent productivity. The upstream cause is three interacting problems, each feeding the others.

What AI AMD with honeypot detection fixes

Accurate AI AMD breaks the loop at two points. First, by classifying carrier false-answers correctly (AMDY recognizes all five FAS patterns by their acoustic signature), it stops burning agent time on phantom connections. Second, by correctly identifying live humans who answer quickly or quietly, it eliminates ghost abandonment — the dropped real prospects that accelerate DID flagging.

The honeypot detection piece addresses the reputation problem directly. Honeypot and spam-trap numbers are seeded into call lists to catch dialers making unsolicited contact. Calling them flags your DID with the carriers and analytics platforms that power the “Spam Likely” label. AMDY flags these numbers in real time before your dialer burns a caller ID on them, and reports which numbers triggered the flag so you can scrub your list.

None of this replaces list hygiene, STIR/SHAKEN compliance, or call center discipline — but it removes the AMD layer as a compounding factor and gives you accurate data on what your calls are actually reaching. When you can see the real breakdown (12.5% human, 14% FAS, the rest machines and intercepts), you can make rational decisions about dial rates, DID rotation, and compliance posture. Without that visibility, you’re flying blind. The state of AMD 2026: real numbers from 2.3 billion calls →