We are aware that rental is the easiest
We are aware that rental is the easiest billing model to set for devs but it has many issues for users - there is big sudden flat charge, you have to manually rent Actor (hard to automate - we could build API for this but it is another step), users are charged with confusing compute etc. MCP/AI is just a small part of this problem but a growing one.
When we designed PPE, we wanted to create a flexible billing model that will cover all use-cases (technically, you can charge monthly there as well). But we are aware it is still cumbersome to set up and we have many ideas how to improve it.
As the future of rental goes, we want to speak with devs, get feedback and decide on that. We definitely want to discourage using it for new Actors but we won't remove it in near future or suddenly.
5 Replies
I have many thoughts...
First of all, we as developers really appreciate how is Apify interacting with its developers community to shape the experience...
To start off, it's true that the rental model is the easiest to set up, but that’s not the only reason why I, and probably many other developers, use it. Another key reason is that, in addition to the mental load of setting up PPE, especially when you have dozens of published actors, we don't want to risk ending up with negative margins. And from what I’ve seen on this Discord server, that seems to happen more often than expected, since websites frequently update their structure, APIs, or anti-bot measures.
We accept the 20% cut that Apify takes under the rentals model, and I think that's a very reasonable trade-off. But with PPE, it looks like Apify still takes its cut and leaves us to absorb the risk of fluctuating costs. If the cut were only taken from positive net profit, that would be a different conversation, but in its current form, it’s just not appealing, at least not for me.
Also, I doubt I'm the only one who has a full-time job while maintaining dozens of actors. Even just keeping them operational is extremely time-consuming, let alone migrating them to PPE. This is far from the "passive income" that Apify tends to promote.
Finally, the fact that we’re even having this discussion & looking at your recent social media activities, they give the impression that Apify is pushing us toward adopting AI or MCP, but maybe some of us/users simply don’t want to go in that direction? Apify reached the stage where it's now without -AI/MCP- adoption and you should focus on what you're good at instead of trying to catch up with the AI/MCP train, yes AI/MCP is a good addition but don't try to revamp your platform around it
I guess a problem could occur if Apify increases the price of CU for some reason (like inflation, for example). Then, someone who has a lot of PPE or PPR actors would need to update each one to avoid ending up with negative profit.
1. Thanks for the great feedback. There is definitely a lot of work Apify needs to do to make PPE almost as easy to set and maintain as rental. Just recently, we did a change that negative profit from one Actor doesn't affect profit of others. More improvements to tooling, platform and analytics will come.
2. Yes, PPE is definitely riskier (we are gonna improve that) but it also has much higher upside. One high use customer can potentially generate thousands of $. Most of top earning devs are there because of PPR-PPE.
3. Your 20% cut point misses one important thing and that in rental model, Apify gets 100% of the compute revenue. Basically, with rental Apify got all the credits from high-use customers and almost all money went to Apify. That all changed with PPR/PPE and now the payouts skyrocketed and large part of Apify's revenue goes directly to devs. This makes our margin much worse and we are becoming more of a marketplace but that's what we decided to focus on.
4. As I explained in other posts, MCP is just one of many issues with the rental model. We are definitely not revamping the platform, it is gonna keep improving for the use-cases it has. MCP is just one new way to run Actors and no one even knows if it is gonna be revelant in a year. Or have you noticed we made something worse because of AI or MCP?
plain-purple•2mo ago
Regarding your point 3:
I think the root issue is that there are basically two "types" of actors: ones with a lot of traffic and ones with a few users.
For the high-traffic ones, PPE makes total sense and I agree with your comment.
But that is only a few of the top-grossing actors on the platform. As with any app-store, there is a long tail of Actors that have just a few users per month. I think both myself and @azzouzana are saying that currently PPE makes little since for these low-volume actors.
Btw, one fairly easy thing that could make this more attractive would be an option to have the platform automatically kill the actor if/when profit goes negative (or some other minimal threshold). It should be pretty east to implement since you are already calculating profit in real-time, and it would provide a good guarantee to creators that even if they screw up the calculations at very least they will not be paying for it out of pocket.
Yeah, there definitely needs to be better control over the margin of runs. We are now working on making the 1 result runs have better profit.