compliance
The Abandonment Rate Is a Number With a Rule Behind It
A predictive dialler that connects more calls than it has agents produces silence on the other end. The permitted rate and its calculation are specified.
A predictive dialler places more calls than there are agents, on the expectation that most will not be answered. When the prediction is wrong, somebody answers and nobody is there.
That is an abandoned call, it is the single most complained-about behaviour in outbound calling, and it is regulated with a specific permitted rate and a specific method of calculation.
What the rule requires
The Telemarketing Sales Rule at 16 CFR part 310 addresses abandoned calls directly, and the FCC's rules at 47 CFR 64.1200 cover closely related ground. The requirements, in outline:
A call is abandoned if a person answers and is not connected to a representative within a short specified interval after the greeting.
The permitted rate is capped, measured over a defined period rather than instantaneously, which matters because a short burst above the cap is not automatically a violation and a sustained level above it is.
When a call is abandoned, a recorded message must play promptly, identifying the seller and stating that the call was for telemarketing. Silence is not acceptable, and a disconnect is worse.
The phone must ring for a minimum period before being treated as unanswered.
Records must be kept sufficient to demonstrate the rate.
The FTC's guide to complying with the Telemarketing Sales Rule sets these out in plainer language, and the precise figures and intervals should be read from the current rule rather than remembered from an article.
The measurement is the compliance
The practical consequence for anybody running a dialler: you have to be able to produce the number, calculated the way the rule defines it, for the period the rule defines.
That is a reporting requirement masquerading as an operational metric, and most diallers report something adjacent. Common mismatches: reporting abandonment against all calls placed rather than against calls answered by a person, excluding answering machines inconsistently, or computing it per agent rather than per campaign.
A dialler that cannot produce the figure on the rule's definition cannot demonstrate compliance, regardless of how well it is actually behaving.
Why the pacing goes wrong
Predictive pacing is a forecast of how many of the next calls will connect, and forecasts degrade when the inputs change.
Changed list quality. A list with a higher live-answer rate than the algorithm expects produces more connections than agents, immediately.
Agent count changes. Breaks, meetings, somebody logging off. A pacing algorithm tuned for twelve agents and running with eight will over-dial unless it reacts quickly.
Time of day. Answer rates vary enormously across the day, and a model that averages across the whole shift is wrong at both ends of it.
Short calls. When calls resolve faster than expected, agents become free sooner, and a conservative pacer under-dials while an aggressive one has already over-committed.
The defence is a hard ceiling independent of the predictive model: never more than a set number of simultaneous attempts per available agent, enforced regardless of what the forecast says. That caps the worst case, which is what a cap is for.
Progressive and preview dialling
The simplest way to have no abandonment problem is not to have a predictive dialler.
Progressive dialling places one call per available agent. Abandonment is structurally close to zero. Throughput is lower.
Preview dialling shows the agent the record before placing the call, and the agent initiates. Lowest throughput, highest quality conversation, and no abandonment at all.
For a business whose lists are small and whose conversations are valuable, which describes most field service and home improvement sales, preview or progressive is usually the better economics as well as the simpler compliance position. Predictive pacing earns its complexity at volumes most of these businesses do not run.
What to watch weekly
Three numbers, and the first is the one with a rule attached.
Abandonment rate on the rule's definition, over the rule's measurement period.
The abandoned-call message, verified actually playing. This is the control most likely to be silently broken by a configuration change, and nobody notices because the people who hear it are not your customers yet.
Ring duration before treating as unanswered, verified against the current minimum.
All three are configuration rather than judgment, which means all three can drift during a change nobody connected to compliance, and that is exactly how they do drift.