Recruiting Software Engineers

by Philip Greenspun in November 2010

Site Home : Business : One Article


This article is intended primarily for new organizations that need to build software development teams quickly.

Start by slowing down

Your competitors aren't as smart as you think. Tens of millions of camcorders are sold worldwide every year, with Canon alone selling 13.4 million in 2009. Customers are enthusiastic about dedicated camcorders at first, excited by the excellent image quality, wide zoom range, and low light performance. After a few weeks, however, they discover that editing video is nearly impossible while watching unedited video is intolerable. The camcorder is parked on a shelf. If the consumer does return to video production it is with a mobile phone or Flip camcorder. Videos of 30-90 seconds are captured and shared without editing.

The market opportunity is to sell a dedicated video editing tablet to everyone in the world who is currently buying a camcorder costing $500 or more (Sony's dedicated consumer camcorders range in price from about $400 to $1200).

Why consumers fail at video editing

Why can't Johnny edit video? First, let's ask why we would expect a consumer to be able to edit video. Being a video editor with current software and hardware is a well-paid job. It isn't ordinarily possible for an average consumer to do something that pays a high salary and requires months of training and years of experience.

Consumers start off with the problem that their home computer may not be capable of editing high-definition video. Their operating system may not have the right codecs installed to display what is coming out of the camera. They might have configured their computer with insufficient memory or a slow hard drive. They might have an older computer with a slow processor. They might have anti-virus or other software installed that slows performance and makes response time unpredictable.

Suppose that they have a brand-new very powerful computer on which no additional software has ever been installed. Notably with Windows 7, the videos from the camcorder show up in the file system explorer as "00017.m2ts", with no thumbnail or any other cue as to what might be in the video. The consumer can install Adobe Premiere, Sony Vegas, or Apple's Final Cut (Macintosh-only) and at least get a look at the videos, but these applications launch with five sub-windows and present a daunting user interface, perhaps the most complex of any desktop application sold to consumers.

Let's consider the consumer's opportunities for failure with Premiere, for example. In creating a new project, the consumer is asked to create a "sequence". Before loading in any video from the camera, the consumer is asked which of approximately 200 "presets" is appropriate. The consequences of picking the wrong preset are dire, with the video potentially being distorted with horizontal or vertical squeezing. In the case of a typical Sony flash-based camcorder, for example, the consumer is supposed to know to click "AVCHD". That's not too hard because "AVCHD" is printed on the side of the camcorder. Once AVCHD is selected, Premiere offers a choice of 1080i, 1080p, and 720p. The camcorder's labels offer no clue as to which of these might be appropriate. Suppose that a phone call to Sony reveals that 1080i is the right format. The consumer then gets subchoices such as "AVCHD 1080i25 (50i) Anamorphic" or "AVCHD 1080i30(60i)". Good luck to Johnny once he gets past these menus into the timeline.

What about ostensibly simpler applications such as Apple's iMovie? The author and his cousin, a professional animator who has worked for decades at Disney and Pixar, tried to remove a sound track of random noises that somehow the program had added to a project. After several hours, we gave up and started over, throwing out the days of work that had been invested in editing the original video.

The FlipShare program distributed on the Flip camera is probably the most successful interface for consumers. People with no training are able to trim original footage and combine the trimmed segments into a final product, though the experience is a little cumbersome and the software cannot be used with files from higher quality camcorders.

Start by throwing out the computer

How to improve the user experience of video editing? Let's start by throwing out the computer. The last thing any American needs is to spend more time sitting in front of his or her computer. If a family video has been captured, it should be editable while sitting on a sofa with the family members appearing in the video. How is that possible? We'll use a touchscreen tablet.

The consumer plugs the camcorder into the tablet, typically with a USB cable, and the tablet automatically starts playing the clips as they are transferred. A consumer has the following touch-screen controls:

At the end of viewing all of the clips, the consumer is asked how long a final video product is desired. The tablet then assembles a proposed video based on the highest-rated scenes that together will run approximately the desired length. The consumer, still comfortably seated on a sofa, now has the opportunity to use the touch screen to reorder the scenes, remove one, or drag in some that did not quite make the cut. Using touch, the consumer can also make small adjustments to the in/out points for each scene.

How you build it

Digital photo titled sagrada-familia-rose-window You build the tablet using dedicated hardware that runs no other applications. That way you can be sure of a consistent user experience. The logical way to start is with Google's free and open source Android operating system and a standard Android-based tablet from a Taiwanese company such as ASUS. To this you would have the hardware vendor add a powerful GPU chip from a standard graphics card. You'd build the software on top of existing free and open source video editing applications (list).

To ensure fast and consistent response time, the tablet would have no hard drive, just flash memory. The tablet need not have more flash memory than the typical 32 GB of a camcorder. For maximum performance, it would be best to exclude a battery and have the tablet run only when connected to AC or DC power. Video editing is the most computationally challenging task that a consumer would want to perform at home and therefore it does not make sense to hobble the machine with a slow clock rate in order to make it run on batteries. Also, it might be possible to charge up some camcorders from the tablet via USB.

The tablet has a WiFi transceiver that is used for four purposes: (1) automatically downloading software updates, e.g., codecs to support new camcorder designs, (2) uploading raw footage to a home server or desktop computer with shared directory for archiving, (3) uploading edited videos to a home server and/or online services such as YouTube, and (4) supporting a basic Web browser on the tablet that can be used for captioning, renaming files, and fact-checking.

[A natural question to ask is "Why isn't this an iPad application". Current (2010) iPads offer inadequate CPU (1 GHz single core) and memory (256 MB) and also lack a USB port. The iPad is optimized for low power consumption to achieve long battery life and it runs very simple applications. A tablet built for video editing that can rely on a power plug would have exactly opposite design constraints.]

Why isn't this just software?

Why not build everything as described above and sell it as a desktop application? The first problem is that there is no way to guarantee a high quality consumer experience. Nothing will stop a consumer from installing the software on a machine with insufficient RAM and a slow mechanical hard disk and/or installing it on a machine that is running 50 other programs in the background. Inevitably such a customer will be frustrated. A second problem is that consumers have already tried desktop applications. If they didn't like a desktop app from a market leading company such as Adobe, Apple, or Sony, why would they want to try a desktop app from a startup company?

If sold as a desktop application, the new program will inevitably appear in a feature-by-feature comparison with Adobe Premiere and Sony Vegas (e.g., Wikipedia; different versions of Vegas). How well is the software going to sell when a consumer sees that for $99 from Adobe or Sony he gets 200 features whereas this new product has only five features?

Finally, as noted above, why would a consumer want to spend more time at his or her desktop? Home video creation is a leisure activity. Home video editing should also be a couch-based leisure activity.

[Perhaps after a few years of delighting customers with the tablet experience, the product could be repurposed for sale as an add-on to Windows 8.]

Financing

This would be a difficult project to get financed in the United States. I have talked informally to a bunch of venture capitalists and the conversations all had the same structure: How much financing would be required for this product? The entire user experience could be prototyped with standard PC hardware, a high-performance video card, a touch-screen, and a long video cable. From the consumer test's point of view, he or she is sitting on a couch holding a screen with an attached cable, but it is a video cable rather than a power cable. Thus no hardware development cost need be incurred in order to prove the concept.

All of the required software components are free and open source. A four or five-person development team should be able to put together a first version of the software within six months, mostly building a touch-screen interface to existing video editing code. Figure a maximum of $250,000 in initial software development costs, even if done in Europe or the U.S.

Hardware development up-front costs depend heavily on how much engineering work the tablet manufacturer is willing to do in exchange for volume sales after the product is launched. Figure that a team of four engineers will need to be paid by someone for about three months. Most of the job will be specifying the processor and GPU and ensuring that the packaging achieves adequate cooling. Budget $500,000 for hardware engineering costs and plastic mold development (assuming that the design cannot run within a standard Android tablet case).

Marketing costs for an effective consumer electronics product launch are high, typically starting at $10 million. This is one reason U.S. venture capital firms traditionally have shied away from this area, leaving the market to well-financed Asian companies, however bad their user experience ideas might be. As a video editing tablet is probably not exciting enough to achieve effective viral marketing, it is probably best to assume that the all-in cost of engineering and launching this product is about $20 million.

Competition

Sumo Competition.  Ryogoku District.  Tokyo Reduced competition is one of the advantages of pursuing an idea that everyone else thinks is bad. Nonetheless, there are some logical competitors. Sony, for example, produces video editing software and tablets. They have an existing customer relationship with a plurality of camcorder owners. Why wouldn't Sony build this? For one thing, they are selling $680 copies of their Sony Vegas Pro package. Why cannibalize that highly profitable business? Similarly Adobe sells Premiere Pro for $799 and Apple sells Final Cut for $999 plus a desktop computer and monitor for $3500+.

The closest existing system to what is described here is the iPhone version of iMovie, which can be used effectively (example), but is not capable of handling the much higher quality streams from a standard HD camcorder.

Conclusion

There is a worldwide market of tens of millions of existing owners of high-quality digital camcorders with no current practical way to edit video from those camcorders. At $500 to $1000 per tablet, there is an opportunity for the producer of a high quality experience to collect billions of dollars in revenue in return for a relatively small investment in software development and hardware configuration.
Text and photos copyright 2001-2010 Philip Greenspun.
philg@mit.edu