I have invented a new methodology that can benefit the buy side (asset management firms, hedge funds, pension plans, insurance companies, etc.) on making portfolio management decisions. Shall I publish the methodology or turn it into a software and sell it?

Both

I am going to read an awful lot into your 2-line post and the fact that you chose to post it in the Careers forum. Since I don't believe there are many meaningful similarities between how portfolio management decisions are made in insurance companies, pension plans and asset management firms (I'm not even going to touch hedge funds), you have either come upon something incredible fundamental or you are mildly delusional. Insurance companies care about book yields and credit ratings. Pension funds listen to actuarial consultants who tell them that they can discount their liabilities at a higher rate if they invest in stocks, rather than what they should be investing in. Asset management firms are (for the most) trapped in a mean-variance world defined by alpha targets and tracking error risk budgets against some benchmark. Common to all of them is that they have fiduciary duties to the ultimate asset owners, so they tend to take a dim view of black boxes. So if you think you can produce software that is actually useful, Gamal is probably right -- you will have to make it crystal clear what the software is doing and why it works, i.e. publish about it. And then worry about the possibility that your invention will get ripped off by competing providers, but that may be a luxury problem. If it were me I would lean toward publishing and aggressively promoting your idea at conferences and use the resulting fame/notoriety to improve your career prospects.

Maybe I over-generalized my statements. Let me put it this way. This is a fundamental technique in probability/statistics, and I have no doubt it can be applied in the mean-variance world. Actually, its applications can be very wide, but let's not go there. Let's say my ultimate goal is to monetize it directly or indirectly. If I publish it, any software providers can implement it immediately, while I make zero cash. I might make my name known if the technique gets popular. However, how does it REALLY improve my career prospects given that I'm not an academic?BTW, I don't think I need to disclose the algorithm for clients to use it, or the clients care to look at the algorithm. The clients desire some properties of an estimated metric, and my method delivers exactly that. They can simply use it as a black box. Think of it as a Matlab function. The user only needs to know the input and (the properties of the) output of the function. He doesn't need to look at the code behind the function.

If you are so sure about your trick why don't you use it in your own trading? You may lever your position as much as you want.

QuoteBTW, I don't think I need to disclose the algorithm for clients to use it, or the clients care to look at the algorithm. The clients desire some properties of an estimated metric, and my method delivers exactly that. They can simply use it as a black box. My 2 cents.I think that this is a huge oversimplification and is very optimistic. What are you offering a) a software product, b) services, c) both?Point a) will be very difficult for many reasons, such as you are presumably a 1-man show, clients want to see source code (+ escrow) and the closer you get to the actuaries the most scary it gets because they want to know everything and will avoid high-falutin' mathematics. Using terms like homotopy and algebraic geometry is taboo.Some software issues would be: is the product per workstation, will it be embedded in another system, support, upgrades, bug fixes, interfacing with external sources etc. Unless you let a software partner do it for you.QuoteThink of it as a Matlab function. The user only needs to know the input and (the properties of the) output of the function. He doesn't need to look at the code behind the function.AFAIK Matlab is used as a 'second opinion'. Then a black box solution is OK I suppose. They might accept you black box solution.Have you considered a patent if it is so unique? Maybe some white papers? Or maybe sell your idea to one or more of the larger software companies in this space?Or sell it to each client on a NDA basis; the client cannot sell it on or tell others.Anyways, go for it and best of luck. Quote If you are so sure about your trick why don't you use it in your own trading? You may lever your position as much as you want. Indeed. Have you tried it on a live test case?

QuoteOriginally posted by: CuchulainnAFAIK Matlab is used as a 'second opinion'. Then a black box solution is OK I suppose. They might accept you black box solution.Matlab is used as a language in production in some pension plans AFAIK. Anyway, that's not the point. I can implement the methodology in other languages.QuoteHave you considered a patent if it is so unique? Maybe some white papers? Or maybe sell your idea to one or more of the larger software companies in this space?Or sell it to each client on a NDA basis; the client cannot sell it on or tell others.Is it possible to patent a math/stat methodology? I thought it's not doable. What is the difference between writing a white paper and publishing my paper in a refereed journal? If I sell my idea to a large software company, do I still own the intellectual property so that I might publish the methodology at some point in the future? How do you determine the price of a methodology?QuoteQuote If you are so sure about your trick why don't you use it in your own trading? You may lever your position as much as you want. Indeed. Have you tried it on a live test case?This is not a trading strategy, although it's a tool commonly used in devising portfolio management strategies. My tool can make the process more efficient and effective. I have tried it on a real application and it works.

There's no reason to hide anything except trading strategies. Publish.

It depends on what your method and its performance are (you didn't provide us with such information -- it doesn't require going into the method's details). If it's something simple (e.g. close to mean-variance optimisation), there's a risk that it's commonly known (nobody bothers publishing it). If it's something less obvious, it's understandable that your potential clients will either ask you about the details of the method or, if you want to sell a black box, want to see a convincing proof of its performance. Mind that you can still keep the real secrets to yourself while talking about / publishing the details of your method if they are in the numerical solutions (which, I suspect, is quite probable considering the subject). If the method performance is impressive, someone might ask why you don't start your own fund ("I have more interesting and ambitious things to do with my life" is a type of an answer their brains won't process). This is essentially pretty obvious..."I have tried it on a real application and it works."I have a feeling that I wouldn't agree with you about criteria of testing a model, and with this feeling I find it hard to believe that you have something valuable.

QuoteOriginally posted by: katastrofa Mind that you can still keep the real secrets to yourself while talking about / publishing the details of your method if they are in the numerical solutions. Why? Once the details of the method are published, there are no secrets anymore.

QuoteOriginally posted by: londonerQuoteOriginally posted by: katastrofa Mind that you can still keep the real secrets to yourself while talking about / publishing the details of your method if they are in the numerical solutions. Why? Once the details of the method are published, there are no secrets anymore.Maybe an example will make it clearer for you... I once published a method, which required using a semi-analytical integration to be numerically feasible. The algorithm was a small but the trickiest part of the fun. It's quite possible that there were others before me who came up with the method, but couldn't properly implement it without finding the algorithm. I published the method together with the algorithm + the numerical code, even if the method (supported by tests, application examples and a comparison with other approaches) was sufficient for the article (it was during my times at a public scientific institution).

If you even have to ask then you should publish. There's little easier than writing a paper, there's little harder than trying to sell. Statistically speaking, you haven't got the right personality for running a business.P

To expand Paul's point, think of it as a net present value calculation.The effort you have already put in as investment is sunk. You are where you are and the optimisation is to get the most from it going forward.Then we add in risk.There is some non-zero chance that this work is not as valuable as you hope. It may be that others are working on the same thing (this happens a lot) and so all you have is a slight time advantage.There is also a difference between what people find useful, what they say they want and what they will pay for. These sets overlap, but not as much as you'd like.To sell the software you have to explain it, not all the details, but enough that someone with the right skills can duplicate it, which will include major firms with big tech staffs and even bigger marketing budgets.The more valuable the s/w, the more it may be reverse engineered.On the other hand if it is a highly useful insight then it is in effect a good CV if published in the right places.As a quant, with a better reputation, your earnings over time may be jumped up a level.So you need to compare the cashflows from a s/w product against improved earnings.A middle case is to write a simple version of the software and present it at conferences, seminars etc.That's a lot less work than writing a whole commercial grade program.You haven't shared what your program might do, but there are some things it probably needs anyway.Data imports.Lots of time on parsing different types of files, files that are nearly right, etcA disgnostic of the form "this file is wrong, try again" won't fly.Not rocket science, but soaks up time.It has to work on various versions of Windows, Unix etc. Tedious and fiddly.Are the graphs pretty enough for good demos ?You also need to writre a script and demo for your sales people.Yes, you will need sales and they will want a cut.Documentation. Lots, worked examples.Support, even if your s/w is flawless, people will still need help and if you think it's foolproof remember they upgrade fools regularly.Then there's producing new versions.That's good for revenue and to kill off bugs but it's work and don't imagine that you won't have people harassing you about the version from threee years ago.I'm not saying don't do it, but do the maths.

What's the half-life of this product-to-be?