Playbooks

They already bought from you. Talk to them again.

The list already exists, the trust already exists, and nobody has called it in two years.

Define dormancy→Remove who must not be called→One reason to talk→Qualify→Transfer

The four decisions that make or break it

• Dormancy is your own repurchase curve, not somebody's twelve-month rule. Plot when returns fall off a cliff.
• The exclusion pass runs before the first call: do-not-call, disputes, active accounts, collections, anyone who asked you to stop.
• The call identifies, names one specific detail of the relationship, gives one reason to talk now, asks one question. It does not open with an offer.
• Qualified means four things: right person, remembers you, current need, agreed to talk. Anything less goes back in the sequence, never to a human.

Report recovered revenue and cost per recovered customer. Calls made is not a result and it will not get you a second campaign.

Transfer three unqualified calls to a branch manager on day one and nobody picks up again. That is how this play dies, and it is not a technology failure.

"Not now" is not "never"

Cold leads are not one list, and calling a never-contacted lead with the same script as one you lost to a competitor is why most of these campaigns produce nothing.

Segment first, then rank

• Five segments, five different first sentences: never contacted, contacted once, quoted-not-closed, explicit not now, lost to a competitor.
• Rank on recency, the stated reason for not buying, value — and above all, anything you can name that changed since. A call that can name what changed is a different call.
• Three to five attempts over two to three weeks, not two to three days. The urgency is yours, not theirs.

"Are you still interested?" is a bad question. It gets a polite yes from people who are not and a polite no from people who are busy. Ask about a fact, and make the yes cost a date or a number.

The real output is not transfers. It is a database more accurate than it was three weeks ago, with the reason for every no written down in the customer's words.

You paid for that call. Somebody should answer it.

Pull one month from your phone system: unanswered, abandoned in queue, after hours, weekends. Multiply by your close rate and average ticket. It is the cheapest number in this library to produce and usually the most alarming.

Then build it in this order

• Overflow first, not full coverage. The agent takes what a human did not pick up after N rings. Nothing changes on a normal day, which is why it survives contact with your staff.
• The job is not answering. It is establishing who, what they need, whether it is real, and which location owns it — inside forty seconds.
• Missed-call recovery is one integration and usually pays for the whole deployment. A call returned in minutes converts close to one that was answered.
• Escalation defined before launch: complaints, legal, hardship, and anyone who asks for a human. First time, no negotiation.

A transfer to a phone nobody answers is worse than the missed call you were fixing, because now the customer has been told somebody is coming.

Read fifty transcripts in week one. Inbound tells you what your customers actually want, which is often not what your website says you offer.

Your reps should be selling

Measure two numbers for one week before you change anything: connects per hour, and conversation minutes per rep per day. In most teams the second is under ninety minutes out of eight hours, and that surprises the person who runs the team.

Dial→Voicemail→Dial→Note→Dial→One conversation

What changes

• The list gets prepared first: dedupe, phone validation, timezone per record, consent, branch ownership. Campaigns fail on the list, in the first hour, in ways that look like a broken agent.
• The qualification bar is a checklist, not a feeling, and the people receiving transfers agree to it before launch.
• Below the bar never enters the same queue as qualified. The moment those mix, the queue is worthless and reps go back to their own lists.

Week one, review every single transfer with the rep who took it. Reps ignoring AI-sourced opportunities because the first three were bad is how this play quietly dies.

Then measure the same two numbers. That comparison is the whole argument.

One company should not run thirty different ways

Every location will tell you they are special. Some are. Most of the differences are accidents that accumulated because nobody wrote the process down.

Sort every difference into deliberate or accidental

• Accidental gets standardised. Deliberate becomes a variable in the config. There is no third category.
• One campaign with location variables, not one campaign per location. At five either works; at twenty only the first does.
• Write the ownership rule before launch and get it agreed by the people whose compensation depends on it. Then define the tie-breaks.
• Central owns the script, the qualification bar and compliance. The location owns hours, transfer target and capacity. Everything else is frozen by default.

Branch-level reporting from day one. Retrofitting it later means rebuilding the identifier chain, which is the most expensive rework in this whole library.

The most common cause of a failed transfer is not technology. It is that nobody was there at 12:40.

If revenue moved, you should know why

Attribution is not a report. It is an identifier that survives five system boundaries, and if it breaks anywhere everything downstream is a guess.

Source→Contact→Conversation→Opportunity→Sale→Revenue

Where it breaks, in order of frequency

• Between the call system and the CRM, when the match is on phone number and the number is shared.
• At the opportunity, when a rep creates a new record instead of using the one that exists.
• At the sale, when whatever books the money has no field for where it came from.

Three fields, never one

• Source, worked by, closed by. The moment a rep's compensation depends on a field the automation writes, you spend more time arguing about data than selling.

Name the three numbers you will report and the record each lives in, before any dashboard exists. If you cannot name the record, you cannot report the number.

The owner's page is five lines and no scrolling, and the fifth one names the location that is off trend. That line is why it gets opened on Mondays.