Syllabus › Module 05, Chatting
Reading an unlock rate: what a figure must disclose before you trust it
In brief
An unlock rate is the share of paid media sent that were bought. Said like that, it is simple. But depending on the scope of fans, the window and the unit counted, the same tool can show very different rates, each accurate for its own definition. This lesson gives the grid for reading a rate, yours as well as a vendor's, and the questions to ask before believing a percentage.
What you will be able to do
- Define the unlock rate and know what it measures and does not measure.
- Recognise the five details a rate must carry to be comparable: definition, period, volume, scope, price.
- Spot the three calculation choices that change a rate, and question a vendor or a chatting team with a checklist.
The previous lesson noted that only one of the seven tools compared published a performance indicator with its definition. This lesson explains why that remark was not a detail. The unlock rate is the figure everyone quotes, AI vendors, agencies hiring chatters, training courses, and almost nobody says how it is counted. Yet it is a figure that bends very easily, with no need to lie.
What an unlock rate measures
The basic definition: of one hundred paid media sent to fans, how many were bought. A media sent and opened counts as one unlock; a media sent and ignored counts as a send with no unlock. The rate says nothing about price, nothing about revenue, nothing about the number of fans. It says one thing: when you offer, does it go.
That is already a lot. A low rate on a large volume of sends is a machine that spams. A high rate on a small volume is a team that does not dare to offer. The rate is always read next to the volume, never alone.
The five details without which it means nothing
| Detail | The question to ask | What changes if it is missing |
|---|---|---|
| Definition | What counts as “sent” and as “unlocked”? Do resends of the same media count as several sends? | A tool that counts once per media and a tool that counts every resend do not share a denominator. |
| Period | Over which window? A week, a month, since the beginning? | A week is chosen. A cumulative figure since launch is not. |
| Volume | How many sends behind the percentage? | 40% on 50 sends and 40% on 50,000 are not the same figure. |
| Scope | How many creators, which markets, all fans or only active fans? | Restricting to fans who reply raises the rate mechanically. |
| Price | At what average or median price? | A 60% rate at $3 and a 30% rate at $15 do not tell the same economics. |
A figure that carries these five details can be compared with another that carries them too. A figure that carries none asks to be taken on trust, or not.
Three calculation choices that change the figure
These are not mistakes: they are legitimate calculation choices that produce different indicators. The problem only starts when the choice is not stated.
The scope of fans. A rate can be computed on every fan who was offered a media, or only on those who replied after the offer. In the second case both sides of the division must be filtered: purchases by the retained fans, divided by sends to those same fans. The result can be higher or lower than the raw rate depending on how the audience behaves; it is a different indicator and should carry its own name.
The window. A launch, a month end or a promotion can give rates that differ from the average. A cumulative figure since launch smooths these variations; a single week shows them. Both are useful as long as the period is written next to the figure.
The unit counted. One can count media sent, offers made in a conversation, or conversations where at least one offer was made. Each unit gives a different denominator, so a different rate. When a published rate changes from one period to the next, the first question is whether the unit stayed the same.
Hybrid and AI alone: two figures that do not compare
In a hybrid mode, the AI opens and a human chatter closes the sale once the fan spends. The displayed rate then measures the work of two actors, and it is impossible to know which one converted. In a 100% AI mode, the rate measures the AI alone, from the first sentence to the payment. A hybrid rate of 30% and an AI-only rate of 30% describe neither the same cost nor the same organisation: the first includes chatter salaries, the second does not. Before comparing two percentages, check that both tools count the same work.
Two published figures, and what it takes to compare them
A vendor publishes “43%” and adds, on its About page, the definition: of one hundred paid media sent by the AI, forty-three are opened; cumulative statistic, platform-wide, since launch. One can debate the level, but the figure can be read: definition, period and scope are there, and the cumulative basis forbids picking the good week. That is the format to demand.
Another vendor publishes a dashboard screenshot: “Unlock Rate 29.8”, 10,233 PPV sent, 3,054 paid, over a dated one-month period. Here period and volume are visible; what is missing to compare it is the scope, one account or all clients, and the definition of a send. When a rate is published with none of these details, it compares with nothing, and the right reaction is simply to ask for the method in writing.
The checklist, for a vendor as for a chatting team
- How do you count a send, and how do you count an unlock?
- Over which period, and chosen by whom?
- How many sends behind the percentage, and across how many creators?
- All fans, or only those who replied?
- Does the figure include sales closed by a human?
Five questions, five written answers. A chatting team that announces “we do 45%” has to answer them like a vendor. And in your own reports, display your definition next to your rate: it is the only way to compare it from one month to the next.
What this lesson does not say
It does not say which rate is “good”. A good rate depends on price, market and platform, and a high rate obtained by giving media away is a bad result. It does not rank vendors: the grid of the previous lesson does that, on their public facts. It gives you the reading, not the verdict.
Key takeaways
- An unlock rate without a period, a volume and a definition of the unlock is not a figure, it is a promise.
- The same account can show different rates depending on the scope of fans, the window and the unit counted. Each is accurate for its own definition.
- The rate of a hybrid mode, where a human closes the sale, and that of an AI alone cannot be compared: they do not measure the same work.
- A vendor that publishes its definition and scope can be compared; a vendor that publishes a bare percentage asks to be taken on trust. Ask for the definition in writing.
Do this now
Compute your own unlock rate three different ways, on a single account. Thirty minutes with your dashboard.
- Over the last thirty days, count every paid media sent and every one that was bought. Divide. That is your raw rate.
- Do it again keeping only fans who replied at least once in the period: their purchases divided by sends to those same fans. That is your rate on active fans; note the gap with the raw rate.
- Do it again on the best week of the month. Note the gap. You have just produced three honest, incomparable figures.
- Write in one sentence the definition you will keep from now on, with period and scope, and display it next to the figure in your reports.
- Reread the page of a vendor you are considering. Look for the sentence that defines its rate. If you cannot find it, write not defined in your grid from the previous lesson.
Sources
- OnlyChat, About page, definition of the 43% unlock ratio (paid media opened per 100 sent by the AI, cumulative, platform-wide, since launch), checked on 2026-11-11.
- OnlyChat, 'OnlyChat vs Substy' comparison, Substy dashboard screenshot quoted with its period (10 January to 9 February 2026, 10,233 PPV sent, 3,054 paid), checked on 2026-11-11.
- Telegram, Bot Payments API, Telegram Stars section (a media price is a whole number of Stars), checked on 2026-11-11.