Believe it or not, push notifications still give the best conversion rates of any mobile marketing channel. But you can’t push out a few messages and hope for the best. Most push campaigns fail before they get off the ground, simply because there was no agreement on what success looks like. This end-to-end strategy guide covers every component of a successful push campaign with no rambling.
Know Which Push Model You’re Actually Running
Under the broad category of “push,” there are two distinct processes for marketing a business, and conflating them is a recipe for disaster for either or both.
The first process is in-app push. These are notifications you send to people who have already downloaded your app and raised their hand as interested by opting into notifications. This is your list. This is a re-engagement play. This (always) costs way more money than you think it should because you’ve already paid to acquire this user once.
The second is out-of-app push ads. This is straight-up mobile display media where you (indirectly) buy access to someone else’s user base and push your message into customers’ or potential customers’ notifications. Because you may be reaching people who have never heard of your product, this is top-of-funnel cold traffic.
A campaign that a business can scale will include both. You should be AB testing in-app creative constantly, and iterating it based on performance; you should similarly be AB testing your out-of-app pushes. But when you pitch the campaigns into the budget system to be funded or starved to death, you are going to present those as different line items because they are scored with very different business objectives: ideally, in-app sticks existing customers onto the repeat engagement wheel. Out-of-app turns a profit.
Why Android Is The Default Starting Point
When developing a push strategy from the beginning and you have to choose a primary platform, Android is the better choice for structural reasons.
Opt-in rates are not the same for both operating systems, and they’re not different by a couple of percentage points. According to Localytics, the average Android opt-in rate is about 61% while the average opt-in rate for iOS is roughly 43%. That difference between being able to market to 6 out of 10 users compared to 4 of 10 users precludes rounding. The math is pretty simple. You can market to more users with a better opt-in rate simply by defaulting to Android and keeping your addressable user base as large as possible on day one before you see your first performance data or read your first configuration guide.
Also, because of the wider opt-in, which is philosophically a gate you want to make your campaign data cleaner and testing faster, you get a faster, bigger learning lab to test creative, look at data, and begin sending notifications. FCM just works better as a testing engine right out of the box with a larger population – and for typically no other reason.
Build The Funnel Before You Touch Creative
Each push campaign, such as in-app or out-of-app, goes through four basic stages: impression, click, landing page, conversion. Too many campaigns fail to generate clicks that achieve anything because you neglect trying to map out this funnel and go straight to “let’s write a catchy headline”.
Before you put pen to paper, pick one KPI. Not three. One. If the goal is awareness, it’s CTR. If the goal is a sale or signup, it’s CPA. If you’re doing paid acquisition through an ad network, it’s usually ROAS. Whichever you choose, every creative decision, every segment, every test, gets measured against that single number. A campaign optimized for CTR and a campaign optimized for CPA can appear the same and result in wildly different budgets and design choices.
Scale Beyond Your Own List With Ad Network Inventory
Once you’ve got your in-app push program dialed in, the logical next step is to go beyond your current subscriber base. That’s where out-of-app push ad inventory comes into play, and it’s a whole other ballgame.
Ad networks aggregate push placements across tons of publisher apps and mobile sites, and you bid on the chance to be in front of users who have opted to allow notifications from those publishers but don’t yet know your brand. This is how you get big fast without spending years organically building up your own subscriber list. Geo, device, carrier, and OS targeting are typically available, and specifically buying push ads on android lets you tap into that broader opt-in pool, with your ad showing up directly in a user’s notification bar just like an owned push would, but in this case buying their attention at scale.
The details of your bid matter more here than they would trying to grab a publisher’s owned space. You’re not trying to win a “glad you’re back” message slot in an app a user already has. You’re competing for their lock screen in hopes of getting them to download you for the first time. Small differences in your bid can make the difference between being buried in the noise of a user’s notifications or being the only message they see.
Win The First 10-15 Characters Or Lose The Message
Notifications on mobile phones can be severely shortened. Notifications on Android devices typically display around 40 to 65 characters on the lock screen and in the notifications shade. Previews will often shorten it even more based on the user’s device settings.
This means the first 10 to 15 characters have to work the hardest in the entire message. If your hook doesn’t evoke curiosity or urgency within that range, your notification will simply get cut off – and the rest of the content is lost on the user. Lead with the benefit or the most important number, not your brand name. “50% off ends tonight” will perform better than “\[Brand name\] sale” because the value is your first 10-15 characters.
Rich push formats are beneficial since the image, emoji, or action button becomes a second visual cue that is not text, and consistently performs better than plain text, with or without a visual element. But the non-text elements can only amplify a strong hook.
Segment Before You Scale
Sending a single message to your whole contact list may cause the number of opt-ins to decrease and the number of recipients unsubscribing to increase. This problem can be solved by segmentation. The benefit of segmentation is usually underestimated by marketers.
At the very least, divide your list based on location, device used, language, and behavior. Despite reducing the number of recipients by only 5%, a well-segmented campaign (i.e, a campaign that excludes people in the wrong language, wrong time zone, or using a device that doesn’t display images well) typically generates 30% to 50% more conversions than the unsegmented version of the same campaign. This is not a small gain; this is the difference between a profitable campaign and an unprofitable one.
Then there is personalization. Including the recipient’s name, city, or recent browsing history in the message is not just a trick; it’s a sign of relevance which in turn reduces the number of unsubscribes.
Test One Variable At A Time
A/B testing for push creative seems easy to do, but in reality, it often fails because people try to test five different things simultaneously, making it impossible to determine which one had an impact.
You need to test variables in isolation. Start by testing one headline copy against another while keeping the emoji, CTA, and send time the same. Next, test the emoji. Then the CTA. After that, the send time. For each of these, you need a sizeable enough sample and adequate time to ensure the results are reliable. Don’t just go with the first promising lead after a couple of hours and scale up the delivery. You simply need more data to work with.
When you have a clear winner, scale it up, and start testing the next element on top of the new base. The process of optimization takes place incrementally.
Send At The Right Time, Not The Convenient Time
Scheduling every campaign to be sent at a specific time on a specific day simply because that’s when the team is available to ‘flick the switch’ is a clearly suboptimal approach. Send-time optimization involves analyzing your previous campaigns to determine when your specific target audience is likely to be most receptive to receiving a message.
Use open and click timestamps from past sends to find your real windows, and send in each user’s local time rather than one fixed hour for everyone. Revisit the data every few months, since behavior drifts and yesterday’s best time won’t stay the best time forever.
Cap Frequency Before Opt-Outs Cap It For You
Sending too many messages is often the main cause of expensive push campaigns and lower inventory quality. When you send a message to a tired user, you raise the probability that the user will disable notifications. Once the user is gone, they are gone for good.
Impose a strict frequency cap. Most apps should opt to send between 1 and 3 messages daily, depending on the industry you are in. Monitor opt-out rate weekly, not monthly. A rising opt-out rate is an alarm. It’s your steady indication that you’re already pushing too much. You will run into trouble on the delivery front with future sending through FCM or monetarily with your ad network partner.
Don’t Let The Landing Page Undo The Click
You can nail every part of a push campaign and still blow it on the last step. A high CTR doesn’t mean much if the landing page can’t close the deal. Someone saw your notification, liked what it said, tapped it – and then bounced because the page didn’t deliver on the promise. That’s a wasted tap, and worse, it’s a wasted opportunity that you paid for.
The fix is simpler than people make it out to be: match the message to the destination. If your notification promises 50% off, the page you send people to should have that sale front and center, not buried three clicks deep in a menu somewhere. Nobody wants to go hunting for what they were already told they’d get. The second someone has to search for the thing that got them there in the first place, you’ve lost them, and no amount of clever copy in the notification is going to make up for it.
Respect Consent Or Lose The Channel Entirely
You need specific, recorded permission before you send anyone a notification, full stop. And when someone unsubscribes, that’s it – you stop immediately, not “within a few business days” like some fine print buried in a settings menu. Permission isn’t a box to tick on your way to launching a campaign. It’s the thing that keeps your messages landing in the first place.
Play fast and loose with this and you won’t just lose access to that one user. FCM and the ad networks you rely on will start throttling or cutting off your reach altogether, and once that happens, the whole program is in jeopardy, not just the send that caused it. The same goes for playing by the rules at the transport level, whatever FCM or your network partner’s guidelines say. It’s not just red tape. Your sender reputation depends on it, and so does everything downstream of it: conversion rates, deliverability, all of it improves once this piece is actually handled right.
Benchmark Against Your Own Segments, Not The Internet Average
Using a general industry benchmark to assess your performance is not very helpful, because the effectiveness of pushes can be very different depending on the industry, geographic location, and the quality of the target group. Therefore, it’s essential to compare your current Android e-commerce segment with the previous month’s Android e-commerce segment and not with the average number from the entire industry. Only this type of comparison will show you whether you are getting better or not.
If a campaign is built following the above-mentioned rules – focusing on the funnel, working with segments, gradually testing, measuring, and stopping the push as soon as it starts fatiguing, which can be stressful because you have to do this month by month – this channel will stop being a game of hit-or-miss and will begin to function as a typical marketing channel. And those who use it in this way will still use it profitably a year later.
