Most outbound calls go nowhere. The phone rings out, a voicemail picks up, the number is switched off, or someone answers who is not the customer. If an agent dials by hand, a large part of the shift disappears into ringing tones. Diallers exist to recover that wasted time, and the different dialler modes are really different bets on one trade-off: agent idle time against the risk of a customer answering and finding nobody there.
The cost of waiting
Start with a rough model. Assume three in ten dials connect to a person, a failed attempt (ring out, busy, unreachable) takes about 25 seconds to resolve, a successful one rings for about 10 seconds before being answered, and a conversation plus wrap-up takes 180 seconds.
On average it takes about 3.3 attempts to get one connect, so before each conversation the agent sits through roughly 2.3 failed attempts and one successful ring: about 68 seconds of waiting. Out of each cycle of about 248 seconds, only 180 are spent talking or wrapping up. That is roughly 72% productive time, and this is with automatic dialling one line at a time. With manual dialling and hand-typed numbers it is worse.
Every mode below is a way of attacking that 68 seconds, with different side effects.
The four common modes
Preview dialling. The system shows the agent the customer’s record first: name, account history, previous call notes. The agent reads it and then clicks to dial, or the system dials after a short countdown. Productivity is the lowest of all modes, but the agent walks into every call prepared. It suits high-value or complex conversations: B2B sales, high-ticket collections, premium customer retention, or any case where a wrong opening line could cost a lot.
Progressive dialling. As soon as an agent becomes free, the system dials the next number automatically, one line per agent. Nobody types numbers, nobody chooses whom to skip, and the customer is never left waiting for an agent, because an agent is always reserved for each call. Agents still hear the ringing and the failures, so the waiting time from the model above remains.
Power dialling. The dialler places a fixed number of calls per available agent, say two or three, on the assumption that most will not connect. It is a blunt version of the next mode. When more people answer than expected, the extras have nobody to talk to.
Predictive dialling. The dialler watches live statistics (current connect rate, average handle time, how many agents are about to finish their calls) and continuously adjusts how many lines it dials so that a connected customer arrives just as an agent becomes free. Only answered calls reach the agent; ringing, busy tones and, with answering-machine detection, many voicemails are filtered out. Talk time per hour rises sharply. The cost is that the dialler is betting on probabilities, and sometimes it wins too much: a customer answers, no agent is free, and the call is dropped or held in silence.
Abandoned and silent calls
An outbound abandoned call is one where the customer picks up and the operation has nobody to connect. From the customer’s side it is a phone that rings, is answered, and goes silent or cuts off. Repeated, it feels like harassment, and in some markets it is treated almost that way.
Several regulators set explicit limits. In the US, the FTC’s Telemarketing Sales Rule caps abandoned calls at 3% of answered calls per campaign, measured over 30 days, and requires a recorded message within two seconds when no agent is available. In the UK, Ofcom’s persistent misuse policy also uses a 3% ceiling, measured per campaign over 24 hours, along with rules on caller ID and repeat calling. Indian rules focus more on consent, customer preference registration and calling-number series than on a dialler abandon percentage, so an Indian BPO serving US or UK clients has to configure its dialler to the client’s market, not its own.
The pacing setting in a predictive dialler is essentially the target abandon rate. Lower it and the dialler plays it safe, agents wait longer, and productivity drifts back toward progressive levels. Raise it and agents talk more but more customers hear silence. Small teams are especially tricky: with only eight agents on a campaign, the dialler has little statistical room, and a sudden burst of answers causes a burst of abandons. Many centres switch small campaigns to progressive mode for exactly this reason.
Answering machine detection, briefly
Answering machine detection (AMD) listens to the first second or two after pickup and guesses whether it is a person or a recording. It saves agent time, but it has two failure modes. It can mislabel a real person as a machine and hang up on them, and the listening delay itself makes the customer say “hello?” into silence. Some operations turn AMD off for sensitive campaigns because the customer experience cost outweighs the time saved.
Choosing a mode
| Situation | Mode that usually fits | Reason |
|---|---|---|
| Large consumer list, short calls, strict productivity targets | Predictive | Enough agents and volume for the statistics to work |
| Mid-sized team, regulated market, low appetite for silent calls | Progressive | No abandons by design, still no manual dialling |
| Complex or high-value calls needing preparation | Preview | Quality of the first 20 seconds matters more than volume |
| Callbacks the customer asked for | Preview or scheduled progressive | A promised call should never be dropped |
Mode is not the only setting that matters. List strategy (which records to dial first), retry rules (how many attempts and how far apart), time-of-day windows, and consent and do-not-call screening often move results more than the dialling mode itself. A predictive dialler on a poor list simply reaches the wrong people faster.
The useful way to think about a dialler is not as a speed button but as a pacing engine with a risk dial. Agree with the client what abandon rate and contact strategy are acceptable, then let productivity follow from that, rather than setting a talk-time target first and letting the dialler find the abandons needed to hit it.
Written by the RP Strativa Team. The timing model is illustrative; regulatory thresholds should be checked against current rules for the market being called.
Leave a Reply