We have just re-launched JUnit Max. See http://www.junitmax.com for details.

Jumping from testing as a separate activity to integrating it as a part of the programmer’s editing session is also something that I think would help AOP. I like the idea of AOP (I’ve used AspectJ) where you can focus on the cross cutting concern. Yet it is ‘separate’ from the code itself. This bothers IT managers (code generation, etc), and is less ‘integrated’ to the programmer.

I just don’t like the ‘flatness’ of code, or the ‘greyness’ of code: When I first walk into some product and have to ‘learn’ it, I can’t tell which class is important or how it fits into the overal picture. Basically, that’s because I don’t have a picture to put it in yet. I have to paint that entire picture to find out what things are and where they belong, and their place in the scheme of things. Then I know which classes and parts of them are hills, valleys, trees, sky, and mountains – I have a landscape, I know where to look for things, how they fit together, etc. But in the beginning, everything is flat – the same shade of gray. As I understand things, classes and such take on shape or color if you will.

Aspects help to color things in a cross cutting way – logging, error handling, etc. Other things besides the happy path. Brakes and seat belts – oh, that’s part of the safety system, engine and transmission – part of the locomotive system, etc. Yet the code has them all weaved together and you can’t see one from the other without filtering it in your head – noisy. I’d love to see a way to turn on or delineate these aspects – IN THE CODE. Annotations sort of help to do this. You see them in the code, yet they’re short little snippets instead of blocks of code. Sometimes you can collapse a comment block or method to cut down on the noise – so something like that capability for aspects would be useful I think. It would be cool if you could click on an aspect, and the code would kind of move out towards you, while the rest recessed. “ah, now I can see how this aspect is being done throughout the code.” Its important to have all the code in there, but it would be nice if the editor simply helped me to view the aspect I want to concentrate on.

Your thoughts?

Thank you for the suggestions. As you point out there were a number of marketing techniques I didn’t try with JUnit Max. The lack of an “enterprise” story was a big problem. Selling programs to programmers is like selling buns to a baker.

While I might have increased revenues 5-10x with the measures you suggest, the fundamental problem remains: there just aren’t that many programmers testing minute-by-minute. Without the market, marketing won’t help.

To be fair, the product needed to be better–faster, more complete interface, more scalable. However, I still think it was good enough to let me measure the market (this could just be rationalization). I didn’t see any path in the future that would have generated 200-500x the revenue, which is what I would have needed for a sustainable business.

1. What if you sold the project to companies directly instead of to individual programmers?

2. Was the lack of revenues a failure of JUnitMax or a failure of the subscription model?

3. What if you sell JUnitMax for $79, or $49, a one time price?

4. I noticed when I was selling my iPhone app that whenever I announced future features my sales dropped — people postpone buying if they think a nicer version is coming out soon.

5. JUnitMax’s subscription model might suffer from the same psychological effect: People might prefer something which is declared done (even if it is just a snapshot of an evolving app). If it is in the process of evolving they might postpone buying.

6. Selling JUnitMax as a product instead of as a subscription should be a cheap experiment to try out. Put the bits on a website. Use PayPal for payment processing. And then watch the Apache log with tail -f to see anyone bites.

7. The marketing for this might not necessarily need to focus on tangible engineering-oriented factors like fast response time. It could appeal to the emotions of developers and their managers. Do you want the free JUnit which is okay, or do you want to develop your code with JUnitMax which is like JUnit on steriods and possibly the best test framework that money can buy? Even if developers are not sold on the value of quick feedback they might just like the assurance of using something high-quality to develop their code.

8. You could try different types of messaging by playing with the headlines of your Google ad campaign to see which generates traffic. Or you could try them on friends. Sometimes one headline or appeal just works a lot better than others.

Thank you for the feedback. My claim to understand the market for test tools well enough to make a business decision comes from my experience with JUnit for the past 12 years, my experience with Agitar for 3 years, and my recent experience actually selling into the market. I also talked to the people you mentioned. I might have missed a big opportunity, but I need to weigh that against the opportunity cost of not doing something else that seems to me to have more potential.

It would have been great to let Max bump along quietly for a while and see if it caught fire later, but I had six months or so of work before it would have been solid enough to drive the support costs down. As it was I had high support costs, low revenue, and shrinking capital. That’s not a sustainable situation. My decision was to move on.

Have you talked to people like atlassian, instantiations etc about the market? Usually people are more than willing to discuss, even if you are competitive.

I pride myself in keeping up with technology. I use every method I can get to find new things and I never even came across junit max until it was featured on hackernews.

E.g.
http://javablogs.com/Search.action?query=beck&sortOption=Relevant&dateOption=Last+Month
http://www.dzone.com/links/search.html?query=kent+beck&x=0&y=0
http://blogs.atlassian.com/news/2009/02/from_startup_to.html
http://blogs.atlassian.com/news/2009/04/5_days_x_thousa.html

Your blog isn’t even mentioned on javablogs! By all means do what you wish, but please don’t claim you know the market.

We have some Eclipse-based API in Mylyn that you might be interested in building on for this effort. Check out: http://eclipse.org/mylyn/new/#general (see Integrated Bug Reporting and Report from Help Menu). This gets more interesting when you consider how it can be hooked up to the Eclipse-based usage tracking technology: http://wiki.eclipse.org/index.php/Mylyn/Integrator_Reference#Monitor_API (see the two Resources listed).

Mik Kersten
CEO of Tasktop
Lead of Eclipse Mylyn

I still don’t see how I underestimated the market. I didn’t say I’d sold all I ever could sell. I’m sure there were more sales out there. There just weren’t enough to justify the cost and time of getting them.