If you own an app, you can send users push notifications. That’s a standard feature. At Aampe, we believe that if you treat this feature as only a feature, you’re leaving an immense amount of value on the table. App notifications are a core strategic asset, not only for engaging users, but also for learning about your user experience and your market - but only if the notifications are created, delivered, and monitored in a way that brings out their value. Those details are hard to get right. That’s why we built Aampe: it was so difficult (and expensive) to make notifications live up to their potential, that for the most part only very large tech companies got consistent value from them. Aampe makes that capability more accessible and affordable. We believe everyone should be able leverage their notifications, and we believe users want them to.

At Aampe, we’re staking the future of our company on push notifications. This post explains why we think this is a really good bet.

Apps rent users’ attention

Apps need users. People use phrases like user acquisition, retention, marketing funnels, loyalty, experience optimization, and dozens of other terms to talk about this challenge of getting and keeping users. Particularly for apps, one of the central challenges is that getting and even keeping a user doesn’t deliver uniform value. You may get a new user to sign up for your app, and they may use your app every day. Or they may never use it again after their initial sign up. You may see a user return several times a week, or you may see a user return less than once a month. 

The term “acquisition” implies possession. If you acquire a computer, or new shoes, or a potato, you have those things. They are yours. Apps don’t acquire users in the same sense: the mental model of “purchase” isn’t really accurate. The challenge of user acquisition is the challenge of convincing users to let you rent their attention when everyone else (not just other apps - literally everything else in the world) is trying to rent that attention at the same time. When you acquire a user, or when you retain a user, or when you engage your user in any form, you don’t buy their attention. You make one more rent payment. 

These rent payments are made in the currency of happiness delivered. You could use a term like satisfaction or value instead of happiness if that works better for you - the principle remains the same. When your app allows a user to say “I bought a thing I wanted” or “I learned a new thing” or “I was entertained for the past hour”, it has paid the rent. If something else - perhaps another app, perhaps not - elicits those responses, that means the user rented their attention to someone or something other than you. That’s as it should be: it’s their attention. All users know they command a scarce commodity. 

The general terms of the rental agreement are no secret. For example, Cisco Appdynamics does an “app attention” index that measures all kinds of ways people interact with their devices and apps. Some of the questions are kind of funny (“Have you ever sworn or cursed out loud due to problems or frustrations with digital services?” 44% said yes. “Thrown or damaged the device?” 13% said yes.) Other questions essentially spell out the terms on which users are willing to rent their attention: 

  • 70% of users want an app experience that’s actually more personalized than face-to-face communication. 
  • 83% of users who have a bad experience don’t give the app a chance to put things right. 
  • 63% actively discourage others from using an app after they’ve had a bad experience. 

This isn’t new information. Other studies show the same sort of thing: users want apps to cater to them as people, and they’re ready to walk if they don’t get what they want.

It’s the details of the rental agreement that pose a difficulty. What makes any individual user happy at any given time depends on a million different things - most of which we can't measure, and almost all of which change frequently. That’s what makes it hard to pay the rent. It’s not just a matter of finding out what users want. It’s a matter of finding out what users want, and when and under what conditions they want it. 

Because of all of the above, apps tend to constantly re-acquire many of their users: they get them, then lose them, then they have to go back and get them again as if they had never seen them before. There’s an extensive AdTech market that handles all this. It’s expensive. And it’s not even the most important part. Getting users in the door is important, but getting them to actively use your service is what ultimately matters

Push notifications are rental requests

For an app, push notifications are the best way to renew a user’s attention rental. Email, SMS, and other communication methods treat the situation as a re-acquisition: the user has already established a way to keep in contact with you - they downloaded your app. Channels other than push notifications are the channels cold-callers use. If you want to be treated like a cold-caller, use those channels.

To be fair, you will need to actually re-acquire some users, even if they still technically have your app installed on their device. That’s because, if a user has refused to rent you their attention for long enough, there’s really no difference between that situation and a situation where you’ve never seen the user before. Push notifications, used well, can reduce the number of users who distance themselves that much from your app. 

Push notifications are the only channel that any app user will recognize as a channel they themselves established. No one gets push notifications and says “hey, how did they figure out how to contact me?” Any user who installs an app expects to receive some communication from that app. The key is to keep the user willing to have that channel open. And that means no annoying or frustrating them with mindless or pointless notifications. In other words, your notifications should deliver happiness.

Keep users the same way you keep your friends

There are a lot of loaded marketing terms like "user experience" or "customer service" or "relationship management", but those big ideas tend to gloss over the unavoidable fact that, ultimately, we're talking about renting the attention of individual people - one by one. If we’re going to deliver happiness, we have to deliver it individually - one by one. So the details are the only thing that really matters here.

You rent people's attention by offering them the things they want more consistently, more easily, less expensively than other people do, even though what they want changes frequently and without warning. In other words, you win by being in the right place, the right time, with the right stuff, packaged in the right way, at the right price (and so on), for as many individual human beings as possible. You could summarize this as "being there for your users."

Make this personal: what does it mean to "be there" for an individual friend? It's not complicated: check in regularly; listen; if they need or want something and you can help with that, go ahead and help. Everyone knows this.

Doing that at scale? Not so easy.

Current tools make you a lame friend

The suite of tools that collectively falls under the umbrella of buzzwords like “marketing automation” and “experience optimization” have trained us to forget that users are people, who want to be treated like people. It’s not hard to find both blog posts and SaaS tools attached to these buzzwords, and the things they talk about make a lot of sense. For example:

  • User segmentation. Find users that act and think alike and tailor your messaging to those preferences.
  • Lead scoring. Rate users by how likely they are to do certain things you want them to do, then tailor your messaging to where they are in terms of their journey.
  • Lead nurturing. Determine where each user is in their journey toward doing the things you want them to do, and tailor your message to encourage them to take the next step forward from wherever they are located at the moment.

All of those things sound at least reasonable, and possibly really exciting, because they conveniently ignore the fact that users are human. Again, let’s make this personal: what would these methods look like in an individual friendship?

  • User segmentation. Get other people to tell you what your friend likes to talk about, then next time you visit that friend, only talk about that.
  • Lead scoring. Look at things your friend has talked about in the past. Score each topic in terms of its relevance to a new set of topics you want to talk about. Next time you visit your friend, only talk about the highest-scoring topic. 
  • Lead nurturing. Decide ahead of time what conversation topics lead to the conversation you really want to have. Next time you visit your friend, only talk about the earliest topic in the chain that they haven’t already talked about. 

If a friend treated me like that, I would start avoiding them. Which is what many do when it comes to push notifications. Marketing automation turns your app into a really crappy friend.

And, no, it doesn’t really help to sprinkle some AI on the whole thing. Analyze millions of data points about other people to find attributes that predict a propensity score for each of a number of predetermined conversation topics. Pick the topic with the highest propensity score. Next time you talk to your friend, only talk about that topic. See how well that works.


Scalable "being there" is hard

A push notification is like corresponding with a friend by text message. Imagine what would happen if you only communicated with all of your friends all at once, in a blast that sent all of the messages on one topic on one day and one time. And imagine what would happen if, no matter how they responded, they still got the same boilerplate message at the same standardized time next week. Wait, that’s obviously inauthentic. So let’s get fancier and keep track of who responded, and text one standardized message to those people, and one standardized message to those who didn't respond. Imagine being on the receiving end of that kind of communication. Would you feel that the sender had really put in the extra effort to be a good friend?

You don't have to imagine what that would be like: I've just described your push notification strategy. No matter what segments you're using, and no matter how complicated your marketing automation is, and no matter what predictive analytics you've trained based on customer data you've mined, you’re still only checking in on your schedule, not theirs; you're still talking more than you're listening; and you're still only minimally responding to what they signal about what they want and need. You're treating your users like metrics, not like friends. It's easy to understand why most every app does that:

  1. Copy creation. Each friend that replies to your message needs to get a response that is appropriate given what they said, and that doesn't repeat something you've already said. That's easy to do for one person at a time. It's really hard to do that for 50,000 people at a time. The copy creation all by itself is onerous.
  1. Inventory recommendation. When you and a friend have exhausted a subject, and you want to keep the conversation going, you have to change the subject to something that's different from what you were talking about, but isn't something entirely out of the blue. Again, that's easy to do for one person, hard to do for thousands at a time. Content/inventory recommendation is hard.
  1. Preference measurement. People don't think about or want just one thing at a time. At any given stage in the conversation, your friend might give you multiple, possibly-conflicting signals about things they like and don't like. When it's one person, you can usually make a pretty good guess about how to balance all of those things, and you can course-correct if you find your guess was wrong. Again, easy for one conversation, hard for thousands at one time. Friendship at scale is a complex measurement/metrics problem.
  1. Complex coordination. Practically everyone knows how to carry on a text conversation, even with autocorrect messing things up on a regular basis. You can keep track of who you're talking to, what text needs to get sent to which person, and how long is appropriate to wait before responding. Once again, this is easy to do for one conversation but hard to do for thousands. Adaptive communication at scale involves a whole lot of plumbing that doesn't come pre-installed on your cell phone.

Aampe let’s you be a decent friend at scale

Aampe solves the content creation problem with our Composer tool (screenshot below), which allows you to break message copy down into building blocks that can be recombined. You write just a handful of notifications, but you get thousands of ways to talk to your users that are both appropriate and fresh.

The graph below shows relationship patterns between inventory items for one of our customers. The relationships are based on actual engagement order - the more users engage with one inventory item after engaging with another, the stronger the relationship. That allows us to generate personalized content recommendations to keep the conversation going even when users have exhausted other content options or have given you very little information to work with. If you want, you can send new content to every user every day, tailored to their personal usage and message history.



We use model-driven metrics for balancing the extent to which users show interest in different features and framings of your app, as well as for summarizing all the different ways a user can "engage" with your app. In other words, “last day seen” isn't the only meaningful metric. In fact, it’s not even a very good one. The following graph shows retention rate for the user of one of our customers based on the last day we saw them (the grey line), compared to our activity score which combines recency, volume and other measures of engagement into a single representation of each user’s willingness to rent their attention to you.

We also allow you to generate the data you need to understand user behavior, rather than trying to fit the square pegs of your current data into the round holes of your decision making about how to talk to your users. Want to know which users will react to an incentive? We can generate a metric for that? Want to know which users find that a particular value proposition resonates? We can generate a metric for that. By calculating metrics that match what you really want from your users, we all allow you to make nuanced decisions about listening to, learning from, and responding to user's individual behavior.

Aampe packages all of the above and more into a scalable system that automates expertise from experimental design, reinforcement learning, artificial intelligence, and other professions. For example, here’s an architecture diagram of one part of our system:


We make sense of things like that so you don’t have to. You don't need to hire a whole new workforce in order to build your notifications into a strategic asset.

Treat your users the way they want to be treated

Think about the costs associated with “being there” for your customers without a toolset. You’ll need an agency (or the internal equivalent of one), just to manage the copy. You’ll need a data science team to manage the setup and analysis of all your tests. You’ll need constant management involvement to make decisions about what to do about results (and to debate their interpretation). And, of course, you’ll need to allocate scarce engineering resources to building and maintaining the infrastructure to make all of the above run reasonably smoothly. That’s an immense investment, which is why, historically, only really big tech companies have invested in this level of continuous experimentation and learning. Each of those companies built their tools to fit their specific use cases. At Aampe, we’ve built the tools to adapt to the specific needs and goals of many different apps. You can make your push notifications a core strategic capability by plugging Aampe into your app. It saves time and money, and it could save your relationship with your users. Companies that learn to treat users as people are the companies who get - and keep - users’ attention.