November was a busy month – busy with things that are not ripe for public talk yet. The snow has covered the ground here, and in the coming weeks I will have time to unpack my Nodal Ninja 5. Stay tuned.
The whole purpose of this blog is to generate a response chain – feedback of any kind into the hugin project, primarily volunteers to spend time on different aspects of the project: the code, the bugs, the graphics, the documentation, the building and packaging. But any kind of response counts.
Ten months ago I started publishing on this site snapshot Windows installers for download. My installers have been popular – installed over 35.000 times. What response rate have we got from Windows users?
Type Of Response Kind
One minute donated to hugin is one minute of very valuable time, no matter if it is for a critical bug fix or a simple user support post to a mailing list. The time spent to advance the hugin project is all equally valid and good response, as are donations, in-kind or financial. For the purpose of this analysis, all response is equal.
Response rate is the number of responses in proportion to the total number of exposures. The total number of exposures is 35.000 – that’s how many times my Windows installer has been used, until I took it down.
I may have counted multiple downloads by the same person. This makes the response rate seem smaller. It is difficult to separate Windows users from the rest. This make the response seem larger. I have used conservative estimates. some of the feedback may go to other Open Source projects and can’t be measured, as pointed out by Leo Sutic.
Return rates can be measured for many types of exposures. A response rate of 0.0005% is excellent for spam but very bad for targeted ads. Similarly one can not expect the same response rate for shareware and for free open source software.
Two weeks ago I started a usage poll. As of today, 65% of those who have seen the poll have voted. This is a good response rate.
At 40% the merchandising poll has attracted less response – it is obviously less relevant to the readers of this blog. A further off-topic question will attract even less response.
The usage poll indicate how high interest for the installers is. At the cut-off time to write this article 45% of respondents used hugin weekly or daily; 33% used it monthly; 17% used it seldom and 5% never used it. Adjusting for the 35% who did not bother to vote and assuming the worse case (i.e. that they never use hugin), we end up with:
- weekly or daily: 29%
- monthly: 21%
- seldom or never: 50%
The above break-down applies across all readers. It is assumed that Windows readers do not differ from other readers. This means that of the 35.000 Windows users who have used my installer
- 10.150 use hugin at least weekly
- 7.350 use hugin at least monthly
- 17.500 never really bothered and are probably just curious or are hunter-gatherers.
What is the added value of my installer for them? Commercial software with a comparable feature set fetch prices of 99 EUR and above. One could argue that I’ve been giving away a value of 3.500.000 EUR in the name of the project; and that 1.750.000 EUR was a waste (because it is not being used). That’s not exactly true: the price influences the quantity. But it is undeniable that 10.150 users find enough value added in hugin.
Imagine for a moment if they would put in 10 cents for every week that they use hugin? that’s 5 EUR per year and user – much less than the 25 EUR per year that PTgui users pay to stay current with updates. 50.750 EUR per year would be enough to pay a junior engineer for an 80% job – not full time, but a lot more resources than the project has currently available. Dream on…
Expected Response Rate
So what kind of response rate can I expect? The right answer is none, because Open Source is a give economy, not a take economy. I can only influence what I give, not what I get. I can come to the conclusion that what I give cost me more than what it is actually worth, and stop giving it.
Actual Response Rate
If these users were exposed to spam the response rate would likely be 0.0005%, or roughly 2 users. And if they were exposed to target ads, it would likely be between 1% and 5%, so between 350 and 1.750 users.
Why comparing with ads? because the installer download is, in a sense, advertisement. The software is given away to advertise a community and attract tester, developer, translators, graphic designer, tutorial writers, and all other sorts of talents and supporters.
The actual response rate to my Windows installer, all type of responses considered, is closer to spam than to targeted ad.
By now it is safe to assume that most internet users know about the online ad model. I experimented to see if my installer was at least worth a click on an ad. For almost a month, I presented about 13% of those who intended to download it with an ad page. That’s roughly 900 impressions. This was done using a redirect that was already in place against deep-links.
The ad page has a twist: it closely watches user behavior and starts the download only for well behaved user.
I’ve experimented with different kind of observations. The results are discouraging: some users even try to reload the same page a dozen times, expecting that it is an error and that the download they want will come soon. None of them even bothered clicking on the ad.
The feedback I get from Windows user indicates that my installer is not worth the effort, and I may as well spend my time doing other things. Otherwise stated, it seems that every platform attracts a specific type of user. The Mac attracts honest people that are used to pay for what is valuable to them; Linux attract gregarious, social and righteous people that are used to share and help each other. And Windows?
Hugin uses the Cmake cross-plattform build system. It is an amazing tool that helps build the same codebase on different platforms. It is powerful and in concept relatively simple, but the details of the building process can be quite tedious and the printed documentation is steep.
The online help that comes with Cmake is helpful, but require command line typing which is not that practical when concurrently editing the CMakeLists.txt files in a text editor.
Introduce CMake Help, a simple and neat graphic user interface that will parse the Cmake online help and display it nicely in a window. Written initially by Aleix Pol and released under the GPL, it has been visually improved by Aurélien Gâteau. Both versions work nicely in Linux but not in Windows, because the line ending character in Windows is CRLF (actually two characters), while it is LF in Linux (and CR for the Mac).
I quickly fixed Aurélien’s versions to work in Windows:
- Download and install Python for Windows. Version 2.5.2 works for me.
- Download and install PyQt for Windows. Version 4.4.3-1 works for me.
- Download and uncompress the cmakehelp archive. You may need 7-Zip in the process.
- Doubleclick on main.pyw to start.
I hope you find this helpful. Happy Cmake-building!
Zoran and Bruno represented the hugin team at this year’s Google Summer of Code Mentor Summit. Bruno has already stitched his panoramas of the summit and of the day the two were touring San Francisco.
One of the thing they noticed is that many other project teams have their own T-shirt. Bruno has an idea of a T-shirt design with a panorama mapped to the entire surface, but normally T-shirt printers only give a 10 inch square for the design. Any idea of a flexible T-shirt printer?
And if there was hugin merchandise up for sale, woud you buy it?