I hate when apps send tons of notifications, but I hear they’re good at keeping people engaged. How do you find a good balance?
Short answer -> test, test and test.
Long answer ->
When I built out the push notification team a year ago, we started with a clear mission – we wanted to drive in app engagement (e.g. users doing the core actions we care about), and we weren’t there just to drive fluff DAU (e.g. people accidentally tapping a notification, opening the app then bailing). We set up our experiment dashboard to clearly show how our notification experiments were performing on this metric (core app actions), and have used this to measure success (while keeping an eye on coverage, CTR, uninstalls, notification disables).
We then started by building the best notifications which were contextually relevant (timing, location) and personalized (used the particular user’s tastes and preferences). In an effort to expand coverage, we had to eventually build some not-so-personalized and not-so-contextually relevant ones. However, we had a rigorous frequency cap, cool down and prioritization schema setup so that we are always trying to send the best possible ping to a user in a given moment, and trying to minimize annoying users with irrelevant noise. In addition to building new content modules, we experimented frequently across all our pings with timing, copy and location optimizations so we were continually improving what we send when to who and how often.
The right balance depends on your app and use case. So the only way to know is to make an educated guess, then test alternatives until you find a sweet spot that engages the most users (and pisses off the least).
Just want to add 1-2 things.
If you dont have much idea about how, when and what you should do, just study the posh notifications of your competitors and the big Unicorns of that particular sector.
For example if mine is an e-commerce app, I should check the notifications of Amazon.
Use this tool, to check what notifications a particular app sends to its users and how often they are sending.
And if you still need any kind of consultancy, feel free to contact.
Push notifications are like cowbells. Sending them too much will most likely be annoying. The general rule of thumb is to send push notifications when you have something timely or urgent to say. For example, if a sale is ending or if they forget to register for an account. Don’t send out push for every new feature or new piece of content. That’s what in-app messages and emails are for (the power of multichannel).
To find the right balance and right cadence of push notifications, think about how people would use your app on a daily basis. For example, if you have a gaming app, there’s a good chance that you have power users who use your app on a daily basis. This type of activity would warrant more frequent push notifications for things like leveling up and new points, etc. On the contrary, if you have a travel app, you would send your users push notifications less frequently because most people aren’t traveling every day or even every month. That activity is less frequent, so your push should be as well.
Then there’s segmentation. You can send more frequent push notifications to more active users who you know are engaged and loyal without risking uninstalls.
Here’s a helpful link.
For a start, make sure you plan your engagement strategy assuming push notifications are not there, which will be true for a large set of your iOS user base. From there on, plan to use push notifications only where they add value to the strategy.
I used to manage at lot of the top Indian apps when I was at MoEngage, and our message to all our clients was to all ways use targeting, better yet, use automated triggers, as they inherently have a lot of context and personalisation. We also pushed clients to have a frequency cap on notifications, which helped a lot.
I definitely agree with Marissa that you have to test! Every user base is different, and you have to find what works best for your product.
I read this article from Andrew Chen’s blog a couple weeks back — it has some interesting data about peak times people open push notifications. A really important component of a push, in addition to the content or the frequency, is the timing — does it fit contextually into a users’ day and how they want to interact with their phone or your app?
In the past, we have thought about these 4 things while thinking about Push notifications.
a. Frequency: Unless its a social app, where notification is about something a friend did etc., we believe users don’t like more than 1 notification per day. So definitely caping it and making sure users approached in one bucket – do not duplicate in another.
b. I still see non-actionable notifications or weak triggers. Like Tumblr sends me notifications like “you would dig this” “You don’t wanna miss this” etc. Its super spammy and I have no idea how to engage with it except they wanting to share it.
c. The information push notifications can sometimes work – but it depends on how will you are able to customize it. PH sends me product notifications, Periscope sends me when people I am following are LIVE – they feel its customized but its not. Its not customized to the PH product I liked today – I am more eager to hear about it more. Same thing with Periscope – its taking my entire follow list and then letting me know rather than the ones I engaged with the last and probably push me something similar.
d. Immediacy : To me push notifications work best when it has immediacy related to it. A flash sale, a moment passing by on Periscope, a friend who just posted something, an email I just received etc.
Overall, Push notifications should be used sparingly (unless triggered by user activity in social activities) , be highly customized, actionable and require immediate action.
Because you want to increase your app engagement. Plus, in the absence of push notification, you have no reliable medium to re-engage your inactive app users. Since an average app loses 77% of its DAU within the first 3 days (source), push notification plays a pivotal role in app retention.
When, how often?
To me, both of them look like the same question because if you could answer ‘when’ you can essentially answer all the related questions around ‘how often’, frequency etc.
In fact, amid the ongoing abuse of push notification as a channel when an average user is getting more than 60 push a day (source) resulting in the decline of opt-in rate, it has become critical for every mobile brand to answer this question of ‘when’ to themselves.
Obviously, the answer to ‘when’ is not one-size-fits-all. It would depend on your business case. Like several notifications from Uber and I hardly mind one but one from Flipkart (promotional) and I am all tempted to delete the app.
To answer ‘when to send push’ you basically would have to identify the right triggers to send push. Now, there are several stages in buyers journey like checkout, inactive, new launches etc where sending push is conventional and warranted. But besides those points, you could certainly innovate unique ways to nudge your users. The hack is to find the right trigger. We had written a post some time back where we had discussed on employing the cognitive approach to identify triggers for push.
We discussed the following 5 cognitive methods and deconstructed examples of some the big brands employing it:
As pointed out by Merissa, ‘test test test’… which is exactly why I’ve cofounded DataEagle, to automate that testing for all!
• In order to find the answer to how often, DataEagle automatically creates dozens of A/B tests, each for a different frequency.
The A group are people being messaged and the B group are people not being messaged. We then track changes in user behaviour after they’ve been messaged for a couple of days (so session frequency, purchase…) and compare it to B group. This helps us measure the impact of notifications for each frequency. (side note we also make sure enough people participate in each A/B test so that the results are statistically valid and reliable)
• For when: we’re keeping track of user’s login times and finding when they typically use the app. This is done for each user so it’s customized to them.
• We’re also using machine learning to group people that are similar together and run tests on them, so we find frequencies that work best for each segment. This approach has improved overall message performance by about 40%.
Once A/B test results are in, we use learnings to decide on when to message a user and automatically send it out when it would have the highest impact.
We’ve often seen that waiting a couple of days after the user’s last session to send a message is much better than sending a message immediately or daily.
Using this we’ve often seen +150% increase in push notification re-engagement effectiveness versus scheduling notifications on a specific day or to everyone at the same time.
With DataEagle our clients are sending out fewer notifications (thus less annoying) but are ultimately much more successful in re-engaging users.
The best thing about it (for marketers and product managers) is that integrating it doesn’t require any engineering effort or app update, since we connect to services that you already have in your app (e.g: OneSignal).