Sales follow-up cadence: how often to call a lead

A good follow-up cadence is two attempts a day, one in the morning and one in the afternoon, at a slightly different minute each time. After a week without an answer, move both attempts to hours you have not tried yet. If the lead asks for a time, use it. Stop the moment they say they are not interested or ask not to be called, and stop when your team marks the lead won or lost. That is the cadence we built into our own dialler. This post explains each choice, including why the chase has to stop the instant the CRM records a no, wherever that no was written.

Why "call five times, then drop it" is the wrong rule

Most teams run a fixed attempt count. It ends the chase for the caller's reasons, not the buyer's: the team got busy, the list got long, somebody decided five was enough.

It also tests the same hours over and over. Ringing a number at 3 pm every day for a month tells you one thing: this person is busy at 3 pm. It does not tell you whether they are interested.

So the useful question is not "when do we give up?" but "when do we try next?" A cadence is a series of small experiments about when one particular person can talk. The first call matters most, which is the case for speed-to-lead and for calling within five minutes. Follow-up is what you do when that first call goes unanswered.

How many attempts a day

Our default is two: one in the morning, one in the afternoon. A workspace can drop to one, which is a quieter setting, or raise it to four.

Two covers two different parts of a person's day without turning your number into one they learn to ignore. Beyond that, each extra attempt mostly tests the same hours again.

Count what happened on every attempt, and keep the outcomes apart:

• Answered. The counter resets, because the rotation is a statement about which hours are failing, and an hour that produced a conversation is not failing. What happens next depends on what the buyer said.
• No answer or busy. The attempt is logged and the next one is booked.
• Failed. The call could not be placed at all. A number that rings out belongs to a person who was busy; a number that fails to route is usually a wrong number. Only one of those is worth a human looking at, so do not merge them.

What times, and why they have to move

Our defaults are 10:00 and 15:00 in the workspace's own timezone. After fourteen unanswered attempts at that pair of hours, which is a week at two a day, both slots move one hour later. The next week probes two hours nobody has tried. The table assumes a calling window of 10:00–21:00, the hours TRAI's default time bands leave open for promotional calls.

The rotation wraps inside the calling window set for the workspace. A slot that would move past the end of the window starts again at the beginning, so each slot tries every hour of the window in as many weeks as the window has hours (eleven, for 10:00–21:00), and no computed time ever falls outside it. There is no path by which the rotation walks a lead into a call at an hour you are not allowed to use.

Two details decide whether this works in practice.

• Count stages in attempts, not days. Days pass whether or not you managed to ring anyone. A weekend, an outage or a closed calling window would otherwise use up a week that never tested those hours. Counting attempts means a stage always means what it says: fourteen real attempts at these times, none answered.
• Reset on an answer. A buyer who picked up at 11:40 once has shown that 11 is not a bad hour for them.

The minute matters too

Ringing at exactly 10:00 every day is the signature of a machine. It is also worse at the job. Someone whose 10 o'clock is always a meeting is unreachable at 10:00 forever and perfectly reachable at 10:23.

So each attempt lands at a different minute within its hour. The minute is worked out from the lead and the attempt number rather than drawn at random, for two reasons. The schedule reads the same every time someone opens it, and a queue that reshuffles itself on every refresh is one nobody trusts. And two hundred leads imported together do not all ring at 10:17.

One trap comes with it. A varied minute can make a slot look unused: a lead rung at 10:17 goes back on the list for 10:47 the same morning, because 10:47 is a different time. Two attempts a day quietly become four or five, and a fortnight of rotation compresses into twelve days. The fix is to treat a slot as spent once its hour has arrived. The wider lesson: test a cadence against a clock, not against a description of it.

A time they asked for beats the rotation

When a buyer says "call me on Tuesday at four", Tuesday at four is when the next call goes out. The whole point of asking "when suits you?" is to use the answer, and a rotation that overrode it would make the question dishonest.

• Clamp it into the window. A buyer may happily ask for ten at night. That is not yours to agree to, so the call moves to the first permitted hour of the day they asked for.
• Use it once. A requested time is consumed by being scheduled. Otherwise every future attempt would land on the hour they once asked for, which is not what "call me on Tuesday" means.
• Treat it as a yes. Asking for a time is the opposite of refusing, so it revives a chase that an earlier refusal had ended.

On outbound calls, a buyer who is driving or in a meeting should be asked for a specific day and time before the call ends. See inbound vs outbound AI calling.

When to stop

A chase should end for a small number of reasons, each written down.

A clear no

"Not interested", said in a conversation and recorded on the lead, stops the chase. It can be reopened, because "not interested" is a sales judgement and someone who said no in March may be worth a call in September. But reopening it takes a person deciding to, never an automatic rule.

What is not a no: "not now, I'm driving", "abhi nahi" (not now), "budget nahi hai" (there is no budget), "call me later", "my wife will decide". Those are objections and reschedules. Picking up the phone is not a request to stop being called either.

An opt-out

"Dobara call mat karna" (don't call again). "ફોન ના કરતા" (phone na karta: don't phone). "Remove my number." A request not to be called goes onto the suppression list the moment it is said, and the number is never dialled again. Unlike "not interested", it cannot be reopened from the follow-up schedule, because it is a compliance record, not a sales one. It is also separate from the national DND register, which your team scrubs against through its DLT access.

Won, or lost

Both are your team's decision, and both stop the chase. Be clear about one thing: our scheduler has no attempt count at which it gives up on someone who never answers. It keeps moving the hours. If your policy is to stop after a set number of weeks of silence, write that number down and mark those leads lost. Choose the rule on purpose rather than letting a dialler, or a tired team, choose it for you.

When the basis for calling runs out

A cadence also sits inside the rules on commercial calls, as of September 2026. Under TRAI's third amendment to the TCCCPR (PDF), opens in a new tab, commercial communications may be sent on the basis of a customer's enquiry "only for a period of seven days from the date of such inquiry", and the enquiry must be in writing or by digital means and kept in a verifiable form. After that you need another basis, and a dialler cannot tell which one a lead has. Our TCCCPR guide and the TCCCPR glossary entry cover the bases; our post on calling hours covers when calls may be made. This is not legal advice.

Why the chase must stop the moment the CRM says no

A refusal can be recorded in many places: by the agent in the middle of a call, by a salesperson editing a lead, by a spreadsheet re-import that changes a status, by a sync from another CRM. If only one of those paths knows about the follow-up schedule, the others let the calls carry on. The buyer said no, and the dialler rings them tomorrow anyway. That is the worst call you can make: it turns a lost lead into a complaint.

So in our system the chase is reconciled with the lead's status every time the lead is written, by any path, and checked again whenever the next attempt is booked. "Not interested" stops the calls no matter where it was typed.

Three more rules came from failures:

• Pausing clears the schedule. Once, pausing a workspace set a flag and left every booked time in place, and the dialler went on ringing people while the screen said it had stopped. Now a pause removes the booked times, and resuming picks each lead up at the stage it had reached.
• The reason is kept. "They told us no on the fourteenth" is a fact about a conversation. It should survive someone reopening the lead later.
• No lead falls out silently. Every active lead with a phone number always has a next attempt booked, and a sweep repairs any that lost one. The opposite failure, a lead nobody is chasing and nobody knows about, is the other way a cadence breaks.

Why "not interested" and "do not call" have to stay separate records is covered in CRM hygiene for teams using AI calling.

What to measure

• Answer rate by hour of day, so you can see which slots earn their place.
• Chases stopped, grouped by reason: not interested, opt-out, won, lost.
• Qualified leads that came from a follow-up attempt rather than the first call.
• Site visits booked from follow-ups.
• The cost of all of it against what it produced. See how to measure cost per qualified lead honestly.

The operations series

• How to measure cost per qualified lead honestly: defining qualified, counting every cost, and a worked example.
• CRM hygiene for teams using AI calling: one number format, deduplication, statuses that mean one thing, and why "not interested" is not "do not call".

For how follow-up fits a property sales team, see AI calling for real estate.

What to do next

Write your stop rules in one paragraph: what counts as a no, what counts as an opt-out, and when silence becomes lost. Then check that every place a lead's status can change enforces that paragraph.