How to Pay Chatters: What Each Pay Structure Produces

Most conversations about chatter pay begin with a number and never reach the structure. The number is local (your country, your shifts, your language) and anyone quoting it as a market rate is quoting a guess. The structure is not local. Fixed pay, commission, or a mix produce different behaviour in the inbox with the same team and the same training, and that behaviour reaches revenue long before it reaches anyone’s complaint.

This page is the agency’s own cost side. What the agency charges the creator is a separate argument, made in flat fee versus commission. What follows describes how these packages are commonly built; it is not employment or legal advice on yours.

What are you actually paying for with each part of the package?

Two different goods, and most broken pay structures are a confusion between them.

• A fixed rate buys presence. A named person, on a named block, reachable, in the queue. It is the only instrument that reliably covers a shift nobody volunteers for.
• A commission buys outcomes. Sales on threads that person actually worked, on a base you can compute the same way twice.
• A bonus buys a direction, at best. It works as a shared pool on an account’s result and fails as an individual monthly target, because a target is a deadline and a deadline gets paid for out of the conversations.

Notice what is absent: activity. Pay per message sent, per pitch made, or against a nightly quota, and you have bought motion. Motion is the one output a chatter can produce all evening without producing a single sale, and the fastest way to produce it is to stop reading before replying.

What does a commission-only package do to a conversation?

It pulls the pitch forward, which is the most expensive thing that can happen to a thread. Across 42,852 conversations, 1.5M messages and 38,879 sales, a sale attempted before the sixth message converts roughly a third worse than one placed after about ten exchanges.

The mechanism is not greed, it is arithmetic the chatter is doing correctly. Under commission with per-shift credit, a thread left to mature is a thread handed to whoever comes on next. Patience is unpaid, so patience stops. Three behaviours follow, and all three are rational:

• The pitch moves to the opening messages, where our corpus says it performs worst.
• The price floor softens at the end of a shift, because a discounted sale credited to you beats a full-price sale credited to a colleague tomorrow.
• The queue gets picked over. High-value threads get the hours; the quiet ones, the fans nobody has spoken to in weeks, get nothing, because commission never reaches the dormant end of the list.

None of that is fixed by telling people to be patient. It is fixed by paying for the thing you want, or by removing the reason not to do it.

What does a fixed rate do that a commission cannot?

It buys the hours and the tasks where the money is not. That sounds like a weakness and it is the entire point: coverage, handovers, notes, the dormant list, the fan who will never buy but whose thread has to stay alive.

Its failure mode is quieter than the commission one. Fixed pay rewards being present, and presence is easy to fake in writing: a reply per thread, none of them demanding, none of them working. That failure never announces itself; it only shows up when you read the threads.

One decision the fixed part settles on its own: night coverage. Our data has conversion collapsing between 2am and 6am and peaking in the evening, so a rotation through the small hours cannot be paid on results. There are not enough results. Buy those hours with fixed pay, and buy them as continuity rather than as growth.

So what does each structure actually produce?

Different problems, on the same roster. Read the failure-mode row first; it is the one you will be managing.

	Fixed only	Commission only	Base plus commission
Who carries a weak month	The agency	The chatter	Shared
Where the pitch lands	Late, sometimes never	Early, the costly end	Depends on attribution
Unpopular shifts	Covered	Refused or padded	Covered
Dormant threads	Worked if scheduled	Never	Worked if scheduled
Price floor	Holds	Softens near payday	Holds if the floor is a rule, not a target
A new hire’s first weeks	Survivable	Punishing, and they leave	Survivable
What you must measure	Quality of the threads	Attribution, constantly	Both
Failure mode	Presence without effort	Early pitching, cherry-picking	Complexity, disputes

The mixed package is not a compromise between two half-good ideas. It is the only one of the three where both jobs (being there, and selling well) are paid for. Its cost is administrative, and it lands in the management overhead line of what a chatting team costs.

What should the commission be calculated on?

On tracked sales, on a base you can name in one sentence, paid from the agency’s own share. Three bases circulate and they are not variations on a theme:

1. The price the fan paid. The largest number, and it includes a share the agency never receives.
2. What the platform transferred: net revenue, the only base where every unit being divided has arrived.
3. The agency’s share of that, after the split with the creator. The narrowest, and the one that matches what the agency actually pays from.

Two rules survive contact with a payroll. Whichever base you pick, the chatter’s commission comes out of the agency’s share rather than off the top of the sale. Otherwise the creator funds the same team twice, which is not the deal she signed; the full stack sits in creator agency commission. And name the sale types by line: subscriptions, PPV, tips, custom orders. A blanket phrase about “sales” holds until the first month tips are the biggest line, and then it holds nothing.

Who gets paid when three people worked one thread?

Everyone who wrote inside a defined window before the sale, if you want anyone to wait for the tenth exchange. This is the clause that changes behaviour most and gets written last. Five rules are in use; the simplest is the most damaging.

Attribution rule	How it works	What it rewards	What it breaks
Closer takes all	The sender of the message before the sale	Speed	Patience: nobody waits for the tenth exchange
Last shift	Whoever held the block when the sale landed	Simplicity	Fairness on threads built earlier
Window split	Everyone who wrote inside a defined window before the sale	Groundwork	Payroll simplicity
Thread owner	One named chatter owns the fan	Continuity and memory	Coverage, when the owner is off
Account pool	A share of the account’s result, split by hours worked	Team behaviour, dormant work	The link between effort and pay

Window split plus a small account pool is the pairing that pays for the two things one rule cannot pay for at once: the message that closed the sale, and the exchanges that made it possible. Whatever you choose, write the window in hours and publish it. An attribution rule people have to guess at gets tested, and the test is an early pitch.

What happens when a sale reverses, and when does the money arrive?

Whatever you wrote down before either question was asked, which is the only version anyone will accept. Both get asked for the first time in a bad month. Settle four things before the first payout:

• The refund rule. No clawback, netting reversals against the next period, or reversing the individual line. Netting is the common middle ground; a chargeback landing weeks later is the case that makes people argue.
• The payment date, and the cut-off it is computed from. A calendar month and a platform payout cycle are different windows, and a sale near the boundary lands in one or the other.
• What the chatter can see. Someone paid on a number they cannot verify eventually stops believing the number.
• What happens on departure. Sales already closed, sales that reverse afterwards, and the final period: three lines, written once, in the offer.

Written down before the first payout, those four stop being negotiations and start being lookups. What is left is not a pay problem at all: it is the ramp, and that belongs in training a chatter.

What would you change first?

Attribution, before the rate. If your team closes early, look at who gets credited before you look at what anyone is paid. The odds are that waiting for the right moment currently costs the person doing the waiting.

Then hold each end of the package against the job it was meant to buy. If the fixed part is not covering the unglamorous work, it is not doing its job. If the variable part pays on anything other than a tracked sale, it is buying activity. And if the whole team sits on pure commission, patience is priced at zero on the behaviour our data ties most clearly to revenue. There is a version where the variable cost only exists when a sale does (that is how justonedash is priced, at 20% of sales with no subscription), but the rule holds either way: pay for the outcome, never for the motion.