Writing a message people do not delete
A push notification gets a glance before a thumb decides. Here is what to put in the first few words, and what to leave out of them entirely.
A lock-screen notification is read in the gap between picking a phone up and unlocking it. That is one line, maybe two, and a decision that takes less time than reading this sentence.
Most shop notifications lose in the first four words.
Key takeaways
- Lead with the shop’s name and one true fact. The customer’s only reason to care is that they like your shop, so spend that advantage first.
- Nobody publishes a character limit. Apple and Google both decline to give one, and the numbers quoted around the web are invented.
- Write for the first clause rather than to a word count, because the tail of your sentence may not render on a phone with large text.
- Relevance matters more than send time. The research that exists points at whether the content is worth reading, not at the hour it arrives.
- Anything you would send twice is the message that gets your pass deleted.
How much room do you actually have
Less than you think, and nobody can tell you exactly how much.
You will find confident numbers online: 40 characters for an iOS title, 65 for Android, 120 for the body. None of them come from Apple or Google. Apple documents a payload ceiling of 4KB and gives no character limit for the title or body at all. Android’s documentation says only that notification text is truncated to one line by default unless you use an expandable style.
The numbers cannot exist, which is why they are not published. What gets shown depends on the device width, the notification style, whether alerts are grouped, and the text size the customer chose in their own settings. It is a pixel budget, not a character count, and a reader who has set large text sees materially less than you do.
So the practical rule is not a length. Put the decision in the first clause. If the customer reads only the opening few words, they should still know who you are and what is being offered. Everything after that is a bonus you may not get.
The “four words” in this article is a working habit rather than a measurement. Nobody has published a credible study on push copy length, and any article quoting an optimal word count is guessing.
The shape that works
Three parts, in this order, and the order is the point.
1. Say who it is
Not “Hi there!”: the shop’s name. The customer has one reason to care about this message and it is that they like your shop; leading with anything else spends the only advantage you have.
2. Say the true thing
One fact, not a mood. The strongest facts are ones the customer already has a stake in:
- how close they are to a reward
- something that is genuinely only true today
- a change that affects their next visit
3. Stop
No sign-off, no emoji run, no “see you soon!”. The message is a notification, not a letter.
Templates worth stealing
The nearly-finished card, which in our experience is the message that earns its send most reliably:
Ninth stamp is on your card.
One more and the tenth is on us.
Something that is only true today:
Sourdough is out of the oven at 3.
About twelve loaves today.
A change that affects the next visit:
We're open till 8 on Thursdays now.
Same kitchen, later.
A quiet shift, without a discount:
Table free tonight if you fancy it.
Kitchen's on till 9.
A win-back that does not beg:
Been a while. Your card's still got
four stamps on it.
Read each of those out loud. If it sounds like something a person behind the counter would say, it will work. If it sounds like a brand, it will not.
What to leave out
Exclamation marks. One is a shop that is pleased to see you. Three is a shop that needs something.
The word “exclusive”. It is only ever true in marketing copy, and every customer knows it.
A discount you did not need to give. If the card is nearly full, the pull is already there. Adding 10% on top does not increase the response; it just costs you 10%.
Anything you would send twice. The fastest way to lose a notification permission is to become predictable, and the second identical message is where that starts.
That last one has support behind it. In a seven-week experiment with 17,500 users of a retail app, sending generic notifications more often both increased uninstalls and reduced the open rate of the notifications themselves. Sending more did not get more read. It got less read, by fewer people.
What the platform will actually let you say
Worth knowing, because it constrains the copy in ways most advice ignores.
On Google Wallet, you author the message. It is added to the pass and triggers a notification, with Google controlling how it renders on the lock screen. You get a maximum of three notification-triggering messages per pass per 24 hours, and going past that returns an error rather than a silent failure.
On Apple Wallet, the push carries no text at all. Apple’s update push is an empty payload that tells the phone to fetch the pass again. The words the customer sees come from a change message attached to whichever field changed. So the text is a property of the card, not of the send.
The practical consequence is the same on both: you are writing one short line attached to a card, not composing an email. Neither platform hands you an unmetered broadcast channel, and that constraint is mostly doing you a favor. The full comparison is in Apple Wallet vs Google Wallet.
Relevance beats the perfect send time
Send-time optimization gets more attention than it deserves for a shop this size.
The in-situ notification research that exists points that way. Pielot and colleagues found most notifications get looked at within minutes of arriving, which limits how much a clever hour can add. Reporting earlier work by Fischer and colleagues, they note receptiveness was driven by whether content was interesting, relevant and actionable, rather than by the time it landed.
Hold that loosely. The Pielot study followed 15 people on Android in 2013, and the receptivity finding is second-hand within it. It is a reason to spend your effort on the sentence rather than the send hour, not proof that timing never matters.
That does not mean timing is irrelevant. A message about this afternoon has to arrive this afternoon. It means the gain from moving a send by two hours is small next to the gain from having something worth reading, and most shops are better off spending the effort on the sentence.
Frequently asked questions
How long should a push notification be? Short enough that the first clause carries the point. There is no published optimal length and any specific number you see is invented. Write it, read the first six words alone, and check they still mean something.
Should I use emoji? We could not find a controlled test on emoji in push notifications at all. The published tests on email subject lines contradict each other, and none report a sample size or significance. Given that, the deciding factor is whether it sounds like your shop. For most independent businesses a single emoji reads fine and a row of them reads like spam.
What time should I send? Whenever the message is true. For anything tied to a shift, send it in time to act on. For anything else, the evidence that a specific hour matters much is weaker than most advice suggests.
Is it worth personalizing beyond the first name? The useful personalization is the fact, not the greeting. “Ninth stamp is on your card” is personal in the way that matters. “Hi Sarah!” followed by a generic offer is not.
How do I know if a message worked? Look at visits in the days after, not at the open rate. The point of the message is somebody walking in, and a notification that gets opened and changes nothing has not earned its place.
Where to go next
The words are the last step. Which channel carries them is push or email, who should receive them is loyalty segmentation, and the three messages worth automating are in automated loyalty campaigns. For what to send when the shop is empty, see the quiet Tuesday problem.
Drafting these, and sending them as a push, an email or a text from the same loyalty program, is what Passumo does.
Sources: Apple, Generating a remote notification and PassFieldContent. Android Developers, Create a notification. Google Wallet, Trigger push notifications. Atilla Wohllebe et al., “Mobile apps in retail: Effect of push notification frequency on app user behavior”, Innovative Marketing 17(2), 2021; 17,500 users over seven weeks, non-personalized notifications only. Martin Pielot, Karen Church and Rodrigo de Oliveira, “An In-Situ Study of Mobile Phone Notifications”, MobileHCI ’14; n=15, Android, fielded 2013, reporting earlier findings by Fischer et al. on receptivity.