“Auto dialler” covers three quite different pieces of software. Vendors use the words loosely, and the wrong choice is the most common expensive mistake in outbound calling.
The three kinds
Preview dialling. The agent sees the next customer’s record, decides when to dial, and the system places the call. The agent is in control.
Good for: anything where context matters. Collections, insurance renewals, B2B sales, anything where arriving unprepared loses the call.
Progressive dialling. The system dials the next number as soon as the agent finishes, one at a time. No waiting, no deciding.
Good for: straightforward lists where every call is similar. Order confirmations, appointment reminders, simple follow-ups.
Predictive dialling. The system dials several numbers at once, predicting how many will be answered, and connects the ones that are to whichever agent comes free.
Good for: large floors making the same call repeatedly, all day. Twenty-plus agents, high volume, simple script.
Why predictive is usually the wrong answer
Predictive dialling is the one on every feature list, and the one most Kenyan businesses should not buy.
The maths only works at scale. The system is making a bet about how many calls will be answered, and when it bets wrong the customer answers to silence — because no agent was free. That silence is the “abandoned call” you have experienced yourself, and it is why people stop answering unknown numbers.
With five agents the prediction has almost nothing to predict from. You get the silence without the efficiency. With twenty-five agents on identical calls, it works.
If you have fewer than about twenty agents dialling constantly, progressive is what you want. It never produces silence, because it only dials when somebody is genuinely free.
We do not sell predictive dialling, which is worth saying plainly before you read further. We think it is wrong for the desks we serve, and if you need it we are not the right supplier.
What matters more than the dialler
The list. A dialler working a bad list calls wrong numbers faster. Time spent cleaning the list beats any feature.
The do-not-call list. This should be checked again as each call goes out, not just when the list is imported. People ask to be removed between the import and the call.
What the agent sees. A dialler that connects a call without the customer’s history on screen has made the agent’s job harder, not easier.
What happens afterwards. Every call needs an outcome — promised payment, wrong number, call back Tuesday — and those outcomes need to drive what happens next. A dialler without good outcome handling is a faster way to make the same call twice.
The Kenyan specifics
Mobile is the default. Most of your list is mobile numbers across Safaricom and Airtel. Dialling behaviour that assumes landlines misreads busy signals and voicemail.
Caller ID matters more here. An unrecognised number gets answered less. Show a Kenyan number people can call back, and the same one every time — not a rotating pool, which looks like exactly what it looks like.
Airtime is a real cost. Outbound minutes are the bulk of an outbound desk’s running cost, far more than seat fees. Ours are KSh 3 a minute billed by the second; whatever you pay, model it on your real volume before you commit.
Calling hours. There is no single rule covering every kind of call, but calling people at nine in the evening damages your brand regardless of what is permitted. Set hours in the system rather than relying on agents.
Working out what you need
Answer three questions.
- How many agents will be dialling at once? Under twenty: progressive. Over twenty on identical calls: predictive is worth discussing.
- Does the agent need to prepare? If yes, preview. Preparation beats throughput on complicated calls.
- What is the call worth? High-value calls justify preview dialling and a slower pace. Low-value, high-volume calls justify automation.
Most Kenyan teams land on progressive dialling with click-to-call for the exceptions, and that is what we build.
The inbound and outbound article covers why these two jobs need different staffing and measures, which is the decision underneath this one.