It is pretty as all heck. I'd be A/B testing it against a simpler one-click-and-done pricing scheme, though. I think geeks get a high off of complex pricing schemes like very few other people do, in particular, in a way that very few people who have the authority to sign off on a $3,500 a month ongoing cost do.
I mean, take the hourly cron option, for example. I'd be asking myself the question "Is having this standard feature on the pricing page, where the presence will probably not influence purchasing decisions but will make it one click longer to get through to entering payment details, going to make us more money at $3 per customer per month than it will cost us in lost conversions?"
You don't even have to make the pricing different -- just abuse the power of defaults and hide all the details on another page that 98% of your customers will never click through to. One enterprise plan: we assume you want the top DB, our recommended number of dynos, unlimited backups, and unlimited cronage. One SMB plan. One startup plan. One hobbyist plan. That takes it down from 6 or 7 layers of decision points, some of which are quite complicated and may be beyond the ability of the person actually making the decision, to a quick "What best describes the type of business you work for" check.
(Note that this may result in some large business customers overpaying for services out of a desire to not have to submit reimbursement requests for "Hobbyist" services. Aww shucks, right?)
These are all good points. There is something else though that's kind of amazing about the page. It got a whole bunch of people who make the decision on what hosting platform to use to talk about it and link to it. They somehow turned a pricing page into viral marketing.
Plus it furthers their brand amongst the HN crowd that they have their act together and are worthy of trusting your application to.
Needs a better description of "dynos" than the help section gives. Also, what is a backup "bundle"? Finally, no mention of bandwidth, only storage. Are there any guarantees here?
"A dyno is roughly equivalent to a Mongrel, except that
dynos are spread across multiple servers, so
performance in most cases is greater.
4 dynos are equivalent to the compute power of one
CPU-core on other systems."
A dyno is different from the common RAM+Bandwidth VPS slice, so it's difficult to make a comparison. (I'm going to do it anyway...) I would say that a 4 dyno machine, with everything else dialed down to the lowest value (so, free), you're roughly equivalent to Slicehost's 1024MB slice. (That's roughly $108/mo at Heroku, $70 at Slicehost.) You've got more storage space (so, the db can grow) on Slicehost, but you have to setup everything yourself, and adding power will take longer than a couple seconds.
It should be stressed that Heroku makes everything really simple. You can save a lot of time and energy if you've got the money to get started with Heroku up front.
I like setting up and tweaking a VPS, though. I haven't seen any benchmarks to show that Heroku with 4 dynos performs as well or better than 4 mongrels, or Apache + mod_rails on the 1024 slice, etc.
I think it's a bit expensive, but, man, it's really cool, too.
I'd argue that a Dyno isn't nearly as resource intensive. If 4 Dynos == 1024MB slice, then each Dyno would be using 256MB. Ruby app servers don't need that much RAM each and with a 1024MB slice you could fit a good deal more than 4 Dynos in it.
In fact, the Phusion Passenger documentation says that they recommend 30 app servers within 2GB of RAM. They do say that if you're running lots of other things like MySQL on a 256MB VPS, you probably want to limit yourself to two app servers. Still, it scales up much nicer. For $70/mo you could easily fit 10 app servers with your database on the same box - something that would cost several hundred dollars with Heroku.
That said, Heroku's selling point won't be cost. Heroku makes deployment exceedingly simple. You don't have to worry about installing anything, configuring anything, making sure things stay up, etc. And that has value - especially if you aren't a *nix gearhead. It's more than a slight premium, but it can still be worth it given that it gives you a worry-free environment.
Not sure if I'm just missing this part of their pricing pages, but if your site starts needing more resources than what you planned on, does it automatically and gracefully bump up the appropriate numbers on the pricing plan and bill you accordingly?
