Study: 42,852 Fan Conversations, Five Measured Findings
Five findings measured across 42,852 conversations, 1.5M messages and 38,879 sales: when to pitch, how to close, what time fans buy, and one counter-intuitive mistake.
Published 29 July 2026
Most advice on selling in conversation rests on one good salesperson’s instinct. We wanted to see what the numbers say. This page gathers the five findings we measured across 42,852 conversations, 1.5 million messages and 38,879 sales, and just as important, what those numbers do not allow us to claim.
What exactly is in the corpus?
Real conversations between content creators and their fans, timestamped message by message, each tied to an outcome: sale or no sale. That tie is what makes the analysis possible: a specific move can be connected to what it produces.
| Quantity | Volume |
|---|---|
| Conversations | 42,852 |
| Messages | ≈ 1.5 million |
| Sales tied | 38,879 |
Three definitions the rest depends on:
- An exchange counts when the fan replies, never when the seller sends. Three messages in a row with no reply move the counter by nothing.
- The rank measured is the first offer: the message carrying a price or paid content, not the payment.
- Rank, not duration. Two conversations at the same rank count the same, whether they ran ten minutes or three weeks.
Finding 1: pitching early costs a third of conversion
An offer sent before the sixth message converts about a third worse than the same offer sent later. The optimum sits after roughly ten exchanges. Full development, with the method caveats, is in when to pitch a sale.
The effect is counter-intuitive for a simple reason: the seller who closes fast looks efficient. They are actually burning conversations that would have produced more.
Finding 2: the last line of the message carries weight
Two moves on the end of a sales message stand out:
- An ellipsis at the end is the worst closer measured.
- A closed question at the moment of closing costs several points of conversion.
These are the two easiest faults to fix, because they need no training and no rewrite of the substance, only the last line. Detail in how to end a sales message.
Finding 3: send time erases wording
The 2am-6am window collapses conversion; evenings maximise it. The gap between those two moments is wide enough to erase any difference in phrasing. A perfectly written offer sent at four in the morning goes out with a handicap no sentence recovers.
The practical consequence is one thing: the fan’s clock, not yours. Details in what time do fans buy.
Finding 4: the personal callback at the wrong moment
This is the most counter-intuitive result in the corpus. A personal callback (“you who love so much…”) placed at the exact moment of the pitch costs points, even though every selling instinct says to put it there.
The explanation is in what the callback does: it builds the relationship upstream, and turns against you when it is used to justify a price. The mechanism is in personalising a sales message.
What do these numbers NOT say?
This is the section that makes the others credible, and the one most studies leave out.
| What the measurement establishes | What it does not |
|---|---|
| Pitching early is associated with worse conversion | That delaying the offer causes a better sale |
| The optimum sits after roughly ten exchanges | That there is an exact rank beyond which all is equal |
| The effect holds across the whole corpus | That it holds creator by creator, price by price |
| Rank is the observed factor | That duration has no effect: it was not isolated |
Two frank limits. First, selection: a conversation that goes far is already a healthy one. Second, the corpus comes from accounts using our tool, with their prices and catalogues. The orders of magnitude hold, the fine tuning does not. Full reasoning in what our data does not show.
How to reproduce the analysis yourself
Without our corpus, two columns recover the two main effects: the rank of the exchange where the first offer went out, and the fan’s local time. Crossed with the sale outcome over a few hundred conversations, they surface your own window within weeks.
That is exactly what justonedash keeps for you: rank and time no longer depend on end-of-shift memory, and the log builds itself, conversation by conversation.
Frequently asked questions
What is this study based on?
On 42,852 conversations between content creators and their fans: roughly 1.5 million messages and 38,879 sales. Each conversation is timestamped and tied to a sale or non-sale outcome, which lets us connect a single move (message rank, send time, how the message ends) to its conversion.
Can these figures be cited?
Yes, attributed to justonedash and kept in the order-of-magnitude form we publish them in. We do not release exact coefficients: they would imply a precision the data does not support. The effects are robust as trends, not to the percentage point.
Do these findings hold for every creator?
The orders of magnitude, yes; the fine tuning, no. The six-message floor and the overnight collapse show up everywhere in the corpus. The gap between the eighth and twelfth exchange, however, depends on the creator's voice, the price point and the fan type. That is what you test yourself, one variable at a time.
Is this correlation or causation?
Correlation, and we say so plainly. A conversation that reaches the twelfth exchange is already a healthy conversation: part of the gap comes from that selection, not the move alone. Our claim is narrower and still usable: pitching early was never associated with better conversion, in any slice of the corpus.
How can I check these findings on my own conversations?
Add two columns to your sales log: the rank of the exchange where the first offer went out, and the fan's local time. Across a few hundred conversations, cross them with the outcome. Your own window appears within weeks, without our corpus.
See what it looks like in practice
The justonedash chatbot holds the conversations, keeps each creator’s voice and works around the clock.