Julia has been around for about 7 years, but the price of base Matlab has been quite stable. Actually if you take into account the bundling toolboxes, it is getting more expensive. Since Julia can do most of the stuffs Matlab offers (at least in my case), I consider Julia as a substitute of Matlab to some extents. It puzzles me that Matlab is still maintaining a high price.
They have corporate support and people who are used to it and can get productive in it don’t want to learn new things, as they might have hobbies outside of work etc. These tools tend to be sticky. I imagine a lot of professor still use SAS, SPSS, Minitab, S-Plus even!
The customer service at Matlab is actually not as good as most people think. In my case, when I send them my issue, then it is slower than posting the question in stack overflow or directly ask my colleagues. When I call them, it is difficult to describe my issue.
Julia Computing can provide corporate support as well. I would say that this discourse is quite good in terms of community support!
I imagine there are a lot of legacy use cases with simulink that depend on having matlab available
There’s no way mathworks can compete on price; Julia is free after all. It would also be extraordinarily difficult for them to compete on core language features without creating something which is not matlab. But regardless of that, matlab has a huge install base and mindshare and there’s plenty of existing systems which rely on software written in matlab.
So mathworks has something which is mature and valuable and they know how to sell it well and who to sell it to. They’ve got existing customers who are invested in the product and expect to pay good money for it. That’s all great for the company in the medium term.
What would lowering the price then achieve? They immediately get less revenue from their large existing customer base. Perhaps they can reduce the rate at which customers leave the platform by reducing the subscription. On the other hand, potential new customers who are price sensitive may choose one of the free option regardless. The actual effect of changing the price depends on how these and other effects trade off against each other, but maybe you can see how companies with mature products to get stuck in a local optimum?
I should say I’ve no actual knowledge of how mathworks/matlab is doing in the market right now, but simply reducing the price might not have the effect you imagine.
Right. At what price would you want to use Matlab? For me the answer is some very large negative number - that is someone would have to pay me a great deal of money to use it over Julia.
I think the “Innovators Dilemma” applies here. The Mathworks can’t suddenly change their business model to be open-source, and it may not even appear to them that there’s any need to do so. They have many entrenched customers who will keep paying their service contracts or buying new licenses; for these customers there’s no need to reduce prices.
I don’t forsee the Mathworks going out of business soon; I see many years of revenue, though with low or negative growth.
Julians ignore the benefits of matlab at their own peril.
It has an excellent IDE and debugger that works out of the box from a point and click install and no configuration or difficulties upgrading, the language is beginner friendly and intuitive, and the speed is generally pretty good in practice (eg MKL and multithreading built in without any tweaking required on installation and even for loops aren’t that bad any more), the plotting library to choose is obvious, never stops working and is fully documented, etc.
As for the price, if you are a scientist, professor, or engineer, the price is irrelevant to you personally since you either have a site license or can justify the cost if it is essential to your day to day work. Why would they need to decrease their price when it is an essential part of the workflow for their target market?
… Of course, I am not listing out the many reasons that Julia is preferable in many cases and for many reasons (and how it dominates matlab for professional developers) . But I only post this to highlight the advantages of Matlab for the many users who don’t like programming and are likely to stay beginners their entire careers.
I understand the rest of your points, and they are very good points, there is definitely a great deal of value in a cohesive, mature product, but I’m not really clear on what the “peril” to Julia is.
Fair enough. The peril I have in mind is if that one of the long-term requirements for “massive” success of julia being attracting a large set of current matlab and perhaps some python users.
If so, then it is easy as an insider who lives day to day Julia to assume it will eventually win and siphon off a large number of users simply based on its massive technological advantage. My point is that this can’t happen without (1) understanding what it is up against and (2) making decisions in the design of the language, the triage of features, trade-offs in convenience vs. purity in implementation, etc. that attract beginners. I won’t point out the specific issues on this, but there are many.
Of course, if matlab users are not a major target, then there is no need to worry about any of that and the sights can be set on other “competitors”.
I think you made a great point. I really want Julia to have a simple way of enabling MKL using all cores.
The price of Matlab is is not a big deal for academia users, but for startup companies the cost is pretty high, especially in developing countries.
Yes, rather than competing on price (basically impossible because alternatives are free), matlab competes on features which make the system as a whole easy for domain experts who aren’t programmers.
As an example, someone was showing me the neural network toolbox recently. My impression was that it makes setting up, training and assessing simple models strikingly easy. It has a nice GUI and some good defaults which make it at least somewhat likely that a non-specialist will get their model to work and assess its accuracy in a valid way.
Building such tools is not something that the free software world excels at! We tend toward building tools for ourselves so we try to build powerful composable libraries usable by programmers. This ignores an entire huge audience of people who aren’t particularly keen on programming as such but just want to solve their domain problem quickly and easily. There’s nothing wrong with that. Actually I think it’s a good thing because it provides opportunities for companies to succeed in the ecosystem by filling in the usability gaps with great products.
also license management really is a headache. It seems to me that code in academia has a good chance of life in the public domain, and it would be a shame if that code were to be released and can only run under matlab (there’s always octave…)
If a business is going to have difficulty competing in the current market, the optimal strategy can simply be to extract revenue from the existing base while slowly winding down. I’m not saying this is happening at MathWorks (I have no idea what their new customer data looks like), but it is a possibility, in which case their current pricing model and trends are rational.
I have put it like this but this seems fairly accurate. I always boggles my mind when someone complains “I have been working for 8 years and I still haven’t made XYZ position”, and I was said “well, there are clear gaps in your technical skills, e.g. programming”, “well,look at person ABC, they can’t program that well either and they are one level above XYZ”. People do progress without very strong technical skills in a technical field, and they will those that come next that “soft skills” are more important. Maybe… but if only every tries to improve their programming…
a paid for Julia-based NN GUI with good defaults?
At the risk of pissing a lot of people off, Julia’s numbers are still quite far from Matlab’s. I think it is safe to assume that people who know Matlab from all stretches of life are still far more than those who know Julia. Also, from the Mathworks website, Matlab has:
- Software installations at over 90,000 business, government, and university sites
- Customers in over 180 countries
- More than:
- 3 million users of MATLAB worldwide
- 4 million files downloaded from File Exchange on MATLAB Central each year
- 225,000 contributors worldwide to MATLAB Central apps
- 500 third-party solutions that build on MATLAB and Simulink
- 1900 MATLAB based books in 27 languages
So while Julia’s features are far more striking in my opinion than Matlab’s, history is still on Matlab’s side. But if the past 5 years is any indication of the future, things may be shifting very fast (I hope).
Very good arguments, which I can fully understand. I do not come from the Matlab corner, but am an intensive R-user. At the moment I look at Julia intensively because I believe in her potential.
But to learn Julia, the learning curve is quite steep. Not because Julia is complicated, but because the documentation shows still clear improvement need (I work on it!). I can imagine that if this point has improved, Julia will gain a larger market share. And “Julia” needs to listen more to the potential users when it comes to usability. Juliapro is certainly a good step in this direction.
The reason to the OP is quite simple. Matlab is to a large extend payed by public money (at least in germany where universities are non-private). So in the end, the price of Matlab is even payed by those people that care about Julia.