Hacker News Reader

Comment by bborud | Hacker News Reader

It would be more interesting to address why they chose a pricing model that guaranteed their products would never be a more mainstream consumer product and pretty much irrelevant for almost all enterprise contexts where a computational or analytical capability is needed.

Because I don't think they are so stupid this wasn't a conscious decision on their part.

View on HN

I think to make it mainstream would require making it different. The number of people who actually have a use for Mathematica is pretty small. Very few people understand both math and programming. Most people who can program are already employed as programmers. In a larger organization, rather than buying 10 licenses for 200 bucks each, it's better to buy 1 license for 2000 bucks and have everybody bring their math work to the "math person." Organizations see this as a natural division of labor. And it's easy to justify that cost because you can show that the math person is using Mathematica full time.

They would have to add more non-programming and non-math features. And at the end of the day, they'd still be competing with free.

In my own case, I've used Mathematica, and I love it. But the standard arguments in favor of using free tools have all materialized over the span of my career, and I'm happy enough with Python.

Well, let's for arguments sake consider the most important demographic in my mind: me :-). (That's what most people do anyway).

Joking aside. I'd love to make Mathematica part of my toolbox, but the steep pricing makes me think about all the other software I pay license fees on that I depend on, and I need another software expense like I need a third buttock.

If it wasn't a recurring license, or a recurring license at a significantly lowered price, that would make it different.

Indeed, in my case, I've got my basic toolchain installed on more than a half dozen computers. At work, I do hardware related R&D, so I need my tools at multiple workstations in the lab. At home, a work-at-home computer and a personal computer.

If I share something, my victims need to install their own copy of the tools.

Then there's the whole world of code running on microcontrollers, which is how I prototype hardware gadgets, also in multiple locations.

There is one piece of domain specific software that I use, and that is tied to a single computer, but I only use it sporadically when I need to do that particular kind of design work.

I appreciate Wolfram's vision for a universal computing platform, but aside from the problem of getting everybody to agree on what that should look like, it's not a realistic dream unless it addresses things like economics and logistics.

You point out something extremely important here: the ability to share work with other people. Tool-use tends to promote or limit networked effects. If I'm the only one who can afford Mathematica in a given context where I collaborate with other people, then we won't be using Mathematica. Simple as that.

On the flip side: If I introduce a tool in a networked context, my enthusiasm may well end up converting dozens of other people.

I think Wolfram could have built a much broader and more stable revenue base if they'd dropped the professional licenses by roughly 1/10 and gotten a bigger user base.

Mathematica is available as a non-recurring license, though you'll have to pay for upgrades if you want to upgrade in that case.

I tried to do this for a few years. I really tried, I renewed, I did various analysis in it... But it's really hard to make it performant on data processing tasks, learning how to do things efficiently in the language.

The curated data is... honestly a little basic and overrated imo. If you are doing general work, maybe their population info is ok, but you don't really know the source of the data entirely, when it was acquired, etc. If you are doing real research, it's not gonna cut it.

I wanted to do analysis of ethereum, but their blockchain library was incredibly weak. It did not support reading data at given block ids (ie: historical analysis), and when I requested the feature the answer was basically "maybe in the next major revision."

I'm not a fan of python, but I went there. I used typed python, which ultimately I think the best solution moving forward is probably typescript, since typed python has a lot of type holes in it :-/

I don't really know but lets play their advocate.

Probably they are stuck in academia market or really research market because that is who they know and this is their target audience.

Moving to more enterprise context would cost them effort, they would have to hire enterprise sales people and spend money on making new connections.

Then to gain real foothold they would have to implement security/auditing/SSO and other enterprise requirements. Which would be quite heavy investment and quite high risk one, because they will never know if any of it will be useful. Like they get first customer, roll out all of that and then growth stagnates and maybe they are stuck on 2-3 customers who keep making demands, but there is no growth to cover for the "enterprise features" investments.

Keep in mind also that not everyone wants to be a unicorn, if they are happy with what they have and roll with it, it is also fine.

Not sure about that. If Matlab is their main competition, then they are much cheaper and you get tons more (that you may never use but its there...) for the price of just Matlab's base software. The downside is that the engineering community by and large uses Matlab because it's what the legacy code is and it is taught in school. I've used Mathematica and I enjoy using it (and Python) more than Matlab but at the end I have to use Matlab because that's what my colleagues use.

Are they that close now to be often interchangeable? Last I looked, they were as different as Lisp is to C: Mathematica is very strong for symbolic manipulation and discrete simulations and DSLs and can somewhat do numerics, whereas Matlab can get you serious performance in numerics but at the cost of awfully inflexible language.

Wolfram’s suite of software seems like it’d require very specialized support staff. Maybe the higher price point helps keep the customer support side of the business more maintainable?

Reply on news.ycombinator.com