paulmeinshausen

And ML can help recognise users who feel that way and "get off their lawn" faster. It's absolutely a tooling problem.

It can be so much better than this. It doesn't have to be an average of most other users. It's perfectly feasible to run user-level optimisation that runs by modeling user preferences and propensities to respond and then letting those models manage the notifications. That's what we've built at www.aampe.com

I used to think this. Then I learned/realised how absolutely crappy the tooling available to those marketing departments is. The solution is for push to not be a manually run operation, it has to be run entirely by a kind of hybrid reinforcement learning system that can automatically manage the explore/exploit tradeoff in order to learn a person's preferences and then update its assessment as those preferences change over time.

Agree this happens. But it's a tooling problem. We wrote an ebook on how timing/frequency optimisation can solve for this problem: https://www.aampe.com/ or https://global-uploads.webflow.com/639b3023f879838c502e7d9f/...

The thing is, technology (including apps and websites), is supposed to help its users. If you want to make room for something in your life (running or any form of exercise, or films, or news, or a particular genre of shopping/retail, etc, etc), then you're also theoretically open to the technology helping you attend to the topic/problem at the frequency/rate and depth that is helpful for you.

The problem is that messaging/notifications don't do that 99% of the time. But not because they can't in theory. It's because the tooling for that technology (i.e. CRM platforms like Braze, OneSignal, CleverTap, MoEngage, Salesforce Marketing Cloud, Iterable, etc) don't make it possible for the app/website (the operators who manage the channel).

Lifecycle notifications and CRM are primarily designed as rule-based systems and they're manually managed. The result is Crap. Crap copy, crap timing, crap delivery.

That's why my cofounders and I started Aampe (www.aampe.com)- to make messaging actually useful for people.

I wrote a blog post about this in the context of food delivery notifications here: https://www.aampe.com/blog/aampe-for-food-delivery-app-notif...

And here are 2 videos about the infrastructure necessary to make notifications work at a user-level, at scale.

"Segmentation comes after messaging - not before" https://youtu.be/SBE0FpWbgyA

"A User Story" https://youtu.be/bFsnB6Thneo