How to bring your abandoned call rate down.

Where they actually come from

A predictive dialer places more calls than it has agents, on the assumption that most will not be answered. When that assumption is wrong for a moment, somebody answers and there is nobody to hand them to.

So every cause is really a case of the dialer's model of your floor drifting away from the floor's reality:

• Pacing set too aggressively for the number of agents actually logged in.
• Agents dropping off together. Breaks, a team meeting, or the end of a shift takes several people out at once while the dialer is still dialling for all of them.
• A list that answers better than the model expects. A fresh, warm or recently-worked list gets picked up far more often, and the dialer is still pacing for the old rate.
• Wrap-up time growing quietly. If reps start spending longer on notes, the model's estimate of when they will be free goes stale.
• Answering machine detection guessing wrong. Every voicemail wrongly classified as a human consumes a connection that a real person could have had.
• Long calls. One rep on a twenty minute call is one agent the model counted on and has not got back.

The fixes, in the order worth trying

These are ordered by effect per unit of effort. The first two solve most of it.

1. Cap it in the dialer, not in a report. The single most effective change is an abandonment target the pacing engine actually respects, so it slows itself down as it approaches the limit. If your current system shows you the number afterwards, you do not have a cap, you have a receipt.

2. Pace to who is logged in, not to headcount. The model should follow the floor in real time. A pacing setting that assumes twenty agents when fourteen are on the phone will abandon calls every lunchtime, and it will do it reliably enough that you can set your watch by it.

3. Tune answering machine detection, then check its mistakes. Detection tuned too loosely lets voicemail reach headsets. Tuned too tightly it hangs up on people who answer slowly, which is worse: it looks like an abandoned call to the person on the other end and often is not counted as one by the system. Listen to a sample of the calls it classified as machines.

4. Stagger breaks. If a whole team goes at once, the dialer loses a quarter of its capacity in a minute. Staggering costs nothing and removes an entire class of the problem.

5. Segment lists by expected answer rate. Mixing a fresh warm list with an old cold one gives the model an average that is wrong for both. Run them as separate campaigns with their own pacing.

6. Look at wrap-up time as a pacing input, not just a productivity metric. If it is climbing, the model's picture of when a rep frees up is stale, and it will keep being stale until somebody tells it.

Measure it honestly

Two counting decisions decide whether your number means anything, and vendors differ on both.

• Is the denominator all calls placed, or only calls a human answered? Only the second is meaningful. Dividing by every number dialled makes any operation look excellent and tells you nothing.
• Does a call the system hangs up on within a couple of seconds count as abandoned? In most definitions it does. A system that quietly excludes those is flattering you.

Track it daily, by campaign, and look at the shape rather than the average. A flat rate of two percent all day is a different problem from a flat zero with a spike at one o'clock, and the second one is usually breaks.

One more: an abandoned rate of exactly zero on a predictive dialer is not a win. It means the pacing is so conservative that you are paying for prediction and getting a power dialer. Somewhere just below your cap is the efficient place to be.

What it costs when you ignore it

The regulatory exposure is the one that gets attention, and it is real. Most jurisdictions that regulate outbound calling put a ceiling on abandonment, and the enforcement tends to arrive as a letter about something that has been happening for months.

The quieter cost is your numbers. People who answer to silence do not just hang up, they mark the number as spam, and enough of that gets a number flagged by the carriers. Then the number stops getting answered by everyone, including the people who would have been glad to hear from you. You will read that as a falling connect rate long before you connect it to abandoned calls.