Push is unforgiving in a way email is not. When someone gets tired of your emails, they stop opening them, and you can win them back with a better subject line. When someone gets tired of your notifications, they click Block — and that is permanent. You cannot prompt them again. They would have to dig into browser settings and reverse it themselves, and nobody does that.
Every mistake below costs subscribers you cannot get back.
1. Firing the native prompt on page load
The single most expensive mistake, and the most common.
A visitor who has been on your site for one second has no reason to want notifications. Most click Block. That Block is permanent for that browser on that device.
Fire your own soft-ask box first. Only the people who say yes to yours ever see the browser dialog. People who say no to yours cost you nothing — you can ask again next week.
2. Asking before the visitor has got anything
Even a soft ask fails if the timing is wrong. Asking at the top of an article is asking someone to commit before they know if your writing is any good.
Wait until they have read 50–60% of the page, or spent 30 seconds on it. You are asking a warm reader instead of a stranger, and it shows in the numbers.
3. Sending too often
Push rewards restraint more than any other channel.
The failure mode is slow. Your click rate drops a little each week, so each notification earns less, so you send more to compensate, so it drops further. By the time it is obvious, you have trained your whole list to ignore you — and the ones who do not ignore you, block you.
Start at two or three a week. Increase only if your click rate holds.
4. Regenerating your VAPID keys
This deletes your entire subscriber list, instantly and irreversibly.
Push subscriptions are cryptographically bound to the application server key that was used when they were created. Change the key and every endpoint becomes undeliverable. There is no recovery and no warning.
People do this by accident: reinstalling a plugin with a “reset” option ticked, switching push providers, restoring a partial database backup that has the subscriber table but not the key option.
If you are about to do anything that touches push configuration, back up the option that stores your keys first.
5. Writing the notification like a headline
Your article headline was written for search results — for someone scanning ten blue links with intent. A notification is read by someone glancing at the corner of a screen while doing something else.
“10 Best WordPress Caching Plugins Compared (2026)” is a fine headline and a weak notification. “Your caching plugin might be making the site slower” is a weak headline and a good notification.
Write them separately. It takes ten extra seconds.
6. Trusting a click rate measured against “sent”
Many push tools report CTR as clicks divided by notifications sent. “Sent” means the push service accepted the message — not that it reached a device, and certainly not that a human saw it. Devices that are off, browsers that never opened, and subscriptions that quietly died all count.
The honest denominator is displayed: notifications that actually rendered on a screen. It is always a smaller number, so the CTR is always lower, and it is the only one that tells you whether your copy works.
If your tool only shows CTR against sent, discount it mentally. If it shows both, use the displayed one.
7. Never cleaning the list
Subscriptions die constantly — cleared browser data, reinstalls, abandoned devices. A push service returns 404 or 410 for a dead endpoint, which is your cue to mark it inactive.
Skip this and two things go wrong. Your stats get worse every month as the proportion of dead endpoints grows. And your sends get slower, because you are spending HTTP requests on addresses that will never receive anything.
Equally important in the other direction: only 404 and 410 mean gone. A 500 from the push service, a timeout, a rate limit — those are temporary. A tool that unsubscribes people on any error will quietly shrink your list during someone else’s outage.
8. Not sending a welcome notification
Someone subscribes and then hears nothing for four days. By the time your first notification arrives they have forgotten agreeing to anything, and an unexpected notification from a half-remembered site gets blocked.
A single message a minute or two after opt-in fixes this. It confirms what they signed up for while the decision is fresh, and it measurably improves how engaged that subscriber is a month later.
Keep it short and do not sell anything in it.
9. Sending everyone the same thing
Once your list is past a few thousand, sending one message to everyone leaves a lot on the table — and actively costs you subscribers who did not want that message.
The cheapest segmentation with the best return:
- By country or timezone, so nobody gets a notification at 3am.
- By device, because a hero image and two action buttons render on desktop Chrome and Android and are ignored elsewhere.
- By engagement, so people who have not clicked anything in three months get fewer messages rather than more. This is counter-intuitive and correct: sending more to a disengaged subscriber does not re-engage them, it gets you blocked.
- By what they read, if your site covers distinct topics.
The one that is not on this list
Not having push at all. It is a real option.
If you publish rarely, if your audience is mostly on iOS, or if you know you will not maintain the discipline to send two a week rather than ten — the channel may not be worth the prompt you are putting in front of every visitor. An interruption that leads nowhere is worse than no interruption.
Push works when you treat the subscriber list as something you can spend down permanently, because that is exactly what it is.
EasyPush defaults to the right side of most of these: soft-ask prompts with scroll and exit triggers, welcome notifications, frequency caps (Pro), automatic cleanup on 404/410 only, and CTR reported against displayed rather than sent. Free, unlimited subscribers, list stays in your database.
Frequently asked questions
What is the most expensive push notification mistake?
Showing the browser’s own permission prompt on page load. People who block it are gone permanently.
Why shouldn't I regenerate my VAPID keys?
Subscriptions are tied to the key that created them. Change it and every existing subscriber stops working.
How often is too often for push notifications?
If your click rate falls week after week, you’re past it. Two or three a week is a sensible start for a blog.
Should I send a welcome notification?
Yes. A single message soon after someone subscribes reminds them what they signed up for, and it makes later messages less likely to be blocked.
Why is my push notification click rate so high?
Check what it’s measured against. Against sent, it’s flattering. Against displayed, it’s honest.