Push, email or SMS: which one to send
They are not interchangeable. One is for things true right now, one worth keeping, and one that cannot be missed. Sending the wrong one costs the channel.
Push, email and SMS are not interchangeable, and having all three in one place invites a bad habit: sending the same message down all three. It doubles the annoyance without doubling the response, and it burns the only channel that is hard to get back.
The useful split is not “which is cheaper”. It is how long the message stays true.
Key takeaways
- Pick the channel by how long the message stays true. Push for today, email for this month, SMS for the thing that cannot be missed.
- Ignore open rates when comparing channels. Apple’s Mail Privacy Protection loads tracking pixels whether or not anyone reads the email, and nobody can measure an SMS open at all.
- On clicks, the one metric that is a real action, SMS campaigns ran at 12.39% against 0.74% for email campaigns (Omnisend, 2025 data, mostly ecommerce brands). SMS also costs money per message and patience per send.
- Never send a push that would have been fine as an email. Sending too often measurably increases uninstalls and lowers the response to everything after it.
Why you cannot compare these on open rates
Almost every article on this subject ranks the three channels by open rate. Those rankings are not measuring what they claim to measure.
Email open rates stopped being reliable in 2021. Apple’s Mail Privacy Protection downloads remote content, including the tracking pixel that records an open, in the background and regardless of whether the recipient engages with the message. An “open” in your email tool may be Apple’s server, not a person.
SMS open rates were never measurable. You will see 98% quoted everywhere. It does not come from a study, and it cannot: consumer SMS gives marketing senders no read receipt. Carriers report delivery, not reading. Any SMS open rate is an inference dressed as a measurement.
That leaves clicks, which are at least a real action rather than an inference. Across Omnisend’s 2025 dataset, SMS campaigns ran at 12.39% click-through and email campaigns at 0.74%, with food and beverage email slightly ahead at 0.96%.
Two limits on that comparison. Omnisend’s customers are mostly ecommerce brands rather than cafés and salons, so these are not your numbers. And the email and SMS figures come from separate reports with different populations, so this is two datasets placed side by side, not a head-to-head test.
Read that gap carefully rather than as a verdict. It says SMS gets acted on more often per message. It does not say you should send more SMS, because the cost of each one and the cost of annoying somebody are both far higher.
Push is for right now
A lock-screen notification is read within minutes or not at all. That makes it right for things with a clock on them:
- the reward is one visit away
- there are six left and it is four o’clock
- the shop is open late tonight
There is measured support for the “within minutes” part, though it is thinner than you might expect. In one in-situ logging study, median time from arrival to viewing ran at 3 to 7 minutes for messenger-type notifications, against roughly 28 minutes for weekend email. Nearly 70% of notifications arrived while the screen was off, so the alert itself prompted the pickup. That study followed 15 people on Android in 2013. Treat it as directional, not as a benchmark.
Push is wrong for anything a person would want to come back to, because there is nothing to come back to. Once it is swiped away it is gone.
Email is for things worth keeping
Email is slower, survives, and can carry more than a sentence. It suits the seasonal menu, the new opening hours, the thing with a picture.
It also tolerates being ignored. An unopened email costs you almost nothing; an unwanted notification costs you the notification permission, and with it every message you would have sent for the next two years.
One practical note for a UK shop, with a caveat attached. Litmus, analyzing billions of opens, found the UK is the outlier on timing: opens peak in the late afternoon, around four to five, and mobile opens peak around six. Most send-time advice online is written from US data, where the peak is mid-morning.
The caveat is the one from the section above. Open timing is built on the same tracking pixel that Mail Privacy Protection interferes with, so treat the shape of the curve as a hint rather than a measurement. It is a reason to prefer an afternoon send over a 9am one in the UK, not a reason to tune to the hour.
SMS is the expensive one
Per message, and per unit of goodwill. It gets acted on more than the other two and has the lowest tolerance for being wasted. That makes it a poor default and a good last resort: the appointment reminder, the “your order is ready”, the win-back to somebody genuinely worth the cost.
Worth knowing before you build a list by hand: in the same Omnisend data, automated SMS flows had a lower deliverability than campaigns, 92.8% against 96.6%. That is ecommerce data and says nothing directly about numbers typed at a till, but the general point holds. Phone numbers decay, and with SMS you pay for the failures.
The three side by side
| Channel | Job | Use it for | Why it matters |
|---|---|---|---|
| Push | Timely messages | Today’s offer, one visit from a reward, open late tonight | Reaches every member with no address or number, and lands in minutes |
| Staying connected | New menu, monthly offer, events, anything with a picture | Survives being ignored, and carries more than one line | |
| SMS | Messages that matter | Appointment reminders, order ready, high-value win-backs | Gets acted on most, costs the most, and forgives the least |
The reason to have all three is not that three is better than one. It is that a customer who leaves your shop is reachable in different ways depending on what you know about them, and the message you need to send has a different shelf life each time.
What sending too much actually costs
This is the part that does not show up until it is irreversible.
In a seven-week field experiment with 17,500 users of a retail app, increasing the frequency of generic push notifications increased uninstalls and lowered the open rate of the notifications themselves. The authors are careful to note they tested non-personalized messages only, so the finding is about generic broadcasting rather than about relevance.
The mechanism is what matters for a shop. Every extra send does two things at once: it slightly lowers the response to this message, and it slightly raises the chance somebody removes the card. The first is recoverable. The second is not, because a deleted pass cannot be restored and that customer will not scan the code again.
So the discipline is not a frequency cap. It is a question asked per send: would this customer be glad this arrived? Writing a message people do not delete covers what passes.
Worked examples
Same shop, four messages, four different answers.
| The message | Channel | Why |
|---|---|---|
| “Sourdough out of the oven at 3, about twelve loaves” | Push | True for four hours. Useless tomorrow. |
| “Autumn menu is live, here is what changed” | Worth keeping, worth a picture, no deadline. | |
| “Your appointment is tomorrow at 2” | SMS | Costly to miss for both of you. |
| “Ninth stamp is on your card” | Push | Timely, personal, and it costs you nothing to send. |
The pattern: the shorter the shelf life, the more it belongs on the lock screen. Anything you would be happy for somebody to read next week belongs in email.
Frequently asked questions
Can I send the same campaign to all three channels? You can, and it is usually a mistake. The same words that work as a one-line notification make a thin email, and paying for an SMS that duplicates a push you already sent is spending money to annoy somebody twice.
How often is too often? For most independent shops, more than about one message a fortnight per customer starts to cost more than it earns. There is no published number that fits every trade, and the honest answer is to watch your own opt-outs rather than copy a benchmark.
Do I need an email address to reach someone? Not for push. A wallet pass is reachable with no address and no phone number, which is why a name-only join form still gives you a contactable member. Email and SMS need details you have to ask for. What to put next to the till covers that trade-off.
Is push really free? There is no per-message cost, which is why it makes a good default. The cost is the permission, and you spend a little of it every time you send.
What about benchmarks for my trade? Treat them skeptically. In Airship’s 2026 food and drink figures, the direct open rate spans roughly 50 to 57 percentage points between the 10th and 90th percentile of senders, depending on platform. When the gap between good and bad execution is that wide, the average is not a target worth aiming at.
Where to go next
Which channel is only half the decision. Who receives it is loyalty segmentation, the three messages worth setting up once and leaving alone are in automated loyalty campaigns, and the words themselves are in writing a message people do not delete. If the goal is filling your slow shifts, start with the quiet Tuesday problem.
Running all three from one loyalty program, so the channel is a choice rather than three separate tools, is what Passumo does. What each one costs, what it needs from the customer and how much is included every month is set out on the channels page.
Sources: Apple, Mail Privacy Protection. Omnisend, email marketing benchmarks (published May 2026, 20+ billion campaign emails across 27,000+ brands, 2025 data) and SMS marketing benchmarks (published June 2026, 246M+ SMS sends); both samples skew toward ecommerce brands rather than local businesses. Martin Pielot, Karen Church and Rodrigo de Oliveira, “An In-Situ Study of Mobile Phone Notifications”, MobileHCI ’14; n=15, Android, fielded 2013. 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, seven weeks, non-personalized notifications. Litmus, best time to send email, analyzing roughly 8 billion opens January to August 2021; open-timing data rests on the same tracking pixel Mail Privacy Protection affects. Airship, 2026 push notification benchmarks.