How to Train a Chatter: Documents, Order, and Ramp-Up

Hiring ends the day someone accepts. What decides whether that hire was worth anything happens over the following weeks, and most agencies run it as an improvisation: a call, a document or two, a few days sitting next to someone, then a live shift. The sourcing and testing side is covered in hiring chatters. This page is about what comes after: the documents that carry the ramp, the order the learning has to happen in, and what a departure takes back.

What does a new chatter need in front of them before the first shift?

Six documents. Anything a trainee has to ask about is an hour taken from your best writer, and you pay for that hour again with every hire.

#	Document	What it settles	What breaks without it
1	Voice guide	How this creator writes: vocabulary, length, punctuation, warmth	Every message sounds like the chatter
2	Price list and floor	What things cost, and the number below which nobody goes	Discounts invented under pressure
3	Annotated conversations	What good looks like, including what lost	Training by opinion
4	Handover format	What the next person is told	Threads restart from nothing
5	Fan record conventions	What gets written down about a fan, and where	Purchases re-offered months later
6	Escalation and refusal rules	What a chatter may never promise, and who to ask	A promise you have to honour or withdraw

Two of those are worth building as templates rather than prose, because they get filled in weekly: the voice guide and the handover note. The other four can start as a page each. A short document that exists beats a thorough one that is still being written when the new person starts.

Why does the order of learning matter more than the length?

Because each stage is only correctable when the one before it is stable. Hand a trainee the price list before the voice guide and they will sell in their own words. Now you are correcting pricing and voice in the same review, and neither correction lands.

Stage	What they do	Off-limits	Exit test
1. Reading	Read closed threads, predict the outcome	Writing to any live fan	They can say which of two versions sold, and why
2. Voice	Draft replies, reviewed before sending	Unreviewed sends, any pricing	Five consecutive messages you cannot distinguish from the reference
3. Money	Quote prices, refuse, hold the floor	Discounts, custom promises	They turn down a discount and the thread stays alive
4. Timing	A capped live shift, plus thread pickup	An uncapped queue	Pitches placed late, and a first message after handover that repeats nothing already bought

Run the stages in that order and each failure is legible. Run them in parallel (the usual approach, because a shift needs covering) and you get a trainee who is vaguely wrong at everything and specifically wrong at nothing.

Which document actually shortens the ramp most?

The annotated conversation set, and the annotations matter more than the conversations. A folder of transcripts teaches nothing; a folder where someone has marked this is where the thread turned, this reply cost the sale, this refusal kept the fan teaches the decision itself. Shadowing only ever shows what someone did, at the speed they did it.

Build it with four properties:

• Include the losses. A set of wins teaches a trainee that the thing that happened is the thing that works. Losses are where the mechanism is visible.
• Anonymise properly. Names, handles and amounts out, before it reaches anyone new.
• Annotate the decision, not the outcome. “He bought” is not a lesson. “She waited two messages here rather than answering the price question” is.
• Keep it current. Add one thread a week from live work; retire the ones that stop resembling how the account sells now.

What are you actually training out?

Instincts, mostly. Nearly everything a capable new chatter does naturally at the point of sale is something our corpus of 42,852 conversations prices as a loss.

• Pitching early. A sale attempted before the sixth message converts roughly a third worse than one placed after about ten exchanges (where the pitch belongs in a thread). New chatters pitch early because it feels decisive and because the shift is nearly over.
• Ending on an ellipsis. The worst-performing sales-message ending we have measured, and the one people reach for when they want to seem playful rather than pushy.
• Closing with a closed question. A yes-or-no at the moment of closing costs several points of conversion. It feels polite; it hands the fan a one-word exit.
• The personal callback at the wrong moment. Referring back to something personal the fan said, placed at the exact moment of the pitch, costs points even though it looks like the most human thing in the message. This is the personal callback mistake, the most counter-intuitive result in the corpus, and the one trainees resist hardest, because everything else in their training tells them to be personal.

Teach these as rules with the measurement attached, not as style preferences. A trainee who knows a rule came from an outcome follows it on a busy evening; one who thinks it is the trainer’s taste stops as soon as nobody is watching.

What does a departure reset, and what does it not?

Everything that lived in one person’s head, and nothing that lived in a document. That is the whole return on the six documents, and it is only visible when someone leaves.

Resets: their fan-level knowledge, their sense of which threads were fragile, their unwritten workarounds, and the four exit tests for whoever replaces them. Their annotations are the one recoverable part, so collect them in the notice period rather than after.

Does not reset: the voice guide, the price floor, the handover format, the fan record, the annotated set, provided the record was kept on the thread rather than in one person’s habits.

That gives you a real test of a training system: ask what a departure would cost you this week. If the answer includes anything a fan would notice, the training is living in people rather than in documents, and every departure is going to charge you for it again.

How do you know the training worked?

By watching where sales attempts sit in the conversation, not by how confident the new person sounds. Confidence arrives before competence and reliably outruns it.

Three checks, in the first month and then monthly:

• Pitch placement. Where in the thread the sales attempt lands, for that chatter, against the rest of the team. Drifting early is the first symptom of overload, not of attitude.
• A blind voice read. Six consecutive messages from one thread, handed to someone off the account: how many people do they think wrote them?
• Pickup quality. Take one thread they inherited and check whether the first message repeats anything the fan already bought or already refused.

Score those consistently rather than by impression. The criteria and the trap of ranking on raw revenue are in the chatter scorecard. And keep the review separate from the pay conversation. A trainee who thinks a voice correction is going to cost them money starts optimising for the number instead of the thread, which is why what the pay package rewards decides more about a new chatter’s habits than the training does.