When it comes to ESD, size matters!
An important metric we track is download completion rate. It's fairly "simple" to calculate -- of all those who start a download, what percentage complete it? Of course, the actual mechanics of how this is calculated may vary widely from site to site and company to company. (For example, if a "product" is made up of 5 segmented files and a customer only downloads 4 of them, is that a "completed" download? And how do you aggregate transactions against the multiple files, not to mention multiple attempts by the same user?)
Bottom line, though, is we've studied this a long time and the larger the file, the lower the completion rate. That sounds pretty intuitive, but just the same, we had to measure it carefully and then prove it, which we've done repeatedly.
This really became a hot issue when we first released the Solaris Operating System for download about 4 or 5 years ago, as the files were very large and the completion rates very low. We formed a Sun Sigma team to investigate the issues and generate an improvement plan. I'll assume I shouldn't post exact results here, but suffice it to say our efforts have greatly increased success rates over time, especially as they started under 30%.
Today the Solaris OS files are cumulatively over 2 GB in size, making for a challenging download. And please don't try it if you're on a dial-up modem -- it'll take you about 70-90 hours! (Please order the media kit instead and save us all a lot of trouble.) We're pleased to say download success rates have held up well, partly due to use of the Sun Download Manager. (I will write a separate post on that later).
What I really wanted to get at is some very interesting research we're doing into advanced file size compression technology. Zip works fine and is pretty much a de facto standard, but it's "off the shelf" functionality. If we customize the compression technology specifically for the Solaris OS content and for specific platforms, we can greatly increase the compression over that offered by zip. The result is much smaller files, yielding solid benefits:
- Faster downloads for our customers
- Smaller files = higher completion rates = better adoption of our technology
- Bandwidth and cost savings for Sun
- Bandwidth and cost savings for customers with ISPs that charge for and/or limit bandwidth usage
- More satisfied customers and fewer support issues
This all sounds really great, but there are concerns. First is the increased cost, especially when you consider we provide the downloads and software at no charge. This proprietary technology (sorry, I can't name names right now) must be licensed and paid for. We're taking a close look at how bandwidth savings can offset the cost and may even be able to show a positive ROI in some cases. Second, it does take a lot longer to restore the file on the client machine than zip. But the extra "restore" time is more than offset by reduced download time. And the slower the connection, the more time is saved.
How do you see the trade-off? If we could save you an hour downloading a huge file but it took 50 minutes to restore on your desktop or workstation, is that trade-off worth it to you?
