Why a conversation beats a coupon
One telehealth pilot ran two openers at the same paywall. The question got 13% to reply and 14.5% to pay. The discount got 0.8% and 6.4%.
The first text is the whole decision. Someone has just closed your paywall. You get one message before they forget your app exists, and the instinct is to spend it on money off. We ran that instinct against the alternative in a GLP-1 program, and it lost badly.
Two openers, one paywall
The app let us split its paywall-exit traffic across two first texts. One opened with a question about what the person had been looking at. One opened with a discount. Nothing else differed: the same app, the same moment, the same number, the same follow-up after the first message.
| Opener | Replied | Paid |
|---|---|---|
| Asks a question | 13% | 14.5% |
| Leads with a price | 0.8% | 6.4% |
How that was measured: pilot analytics from one GLP-1 companion app, split A/B by opener over the pilot window. Reply rate is people who sent any message back. Paid is people who bought, whether or not they replied. It is one app, run by us, and we are not presenting it as a trial anyone else reviewed.
The gap in replies is sixteen to one. The gap in payment is a little over two to one. Both matter, and the second one is the one that pays your bills.
A discount answers a question nobody asked
A price cut only helps a person who decided against you on price. Most people who leave a paywall have not decided anything. They wanted to know whether the thing does what they need, and the paywall did not say, so they left to think about it and never came back.
Open with money off and you have told them two things they did not want to hear: that you think they are cheap, and that the price you showed them a minute ago was not real. You also close the conversation. There is nothing to reply to in a coupon.
A question is a door.
If the first message cannot be answered, it is an advertisement, not an opener.
Replying is what predicts paying
In the same pilot, people who replied at all went on to pay at 35% to 43%. People who never said anything paid at about 10%. That range is the week-to-week spread across the pilot, not a confidence interval, and it is drawn from the same pilot analytics as the table above.
So the reply is not a vanity number on the way to the real one. It is the strongest signal in the pilot about who is going to buy. An opener that gets 0.8% of people talking has thrown that signal away before the conversation started.
What Hyzl does with the first message
- It reads what the person was actually doing in your app, and asks about that, not about your pricing page.
- It asks one thing. Not three, and never a paragraph.
- It holds discounts back. Hyzl offers a price only after the person raises price, and only from the tiers you set. It never invents one.
- It checks the payment provider first. Someone who already subscribed is never contacted.
Hyzl still sends discounts. It sends them second, to the person who told you money was the problem, which is the only person a discount was ever going to help.
What we are not claiming
This is one A/B, in one industry, over one pilot window. It does not say a question beats a coupon in your app, and it does not say discounts never work. It says that in a GLP-1 program, spending the first message on price cost eight of every nine replies and more than half the payments, and that we build the opener accordingly until an app shows us otherwise.
If your own numbers disagree, the dashboard will say so. Every move Hyzl makes is logged with the reason it was made, so an opener that is not working is visible in a week rather than a quarter. See how it decides.