In researching the how-tos of podcasting one of the more interesting things is the amount of data used in downloading files.1 I also noticed that audio is seen as the poor cousin to video. Video files are much larger, but very few people pay for their own video hosting. I doubt YouTube would be as popular if it required a monthly fee to keep your videos online.
The usual recommendations for podcast files seem to say to encode the episode file (MP3) at 128kbps. This works out to about 1 MB per minute of audio. These are usually stereo files, as most encoders default to that. To test this theory, I looked at a recent episode of a popular show. By opening the
get info panel in iTunes, we can see what settings Serial uses.
The episode is encoded as a 128kbps stereo (2 channel) MP3 file with a sample rate2 of 44.1kHz. The file is 49.8 MB in size for a duration of 55 minutes (not shown). This is pretty much textbook file preparation. It also sounds good in both headphones and car speakers.
Notice that it’s also a two channel file. So to play at 128kbps (128 thousand bits per second) each channel must be half of that. In simple terms it’s two 64kbps streams played at the same time. One stream is sent to the left speaker and the other to the right.
If this file were to be reduced to a single channel (mono) it would be half the size and that one channel would be played from both speakers. There are some sound advantages to stereo like panning3, but doubling the file size is serious disadvantage. Especially when you might not have a large corporation paying for your bandwidth usage.
Data usage for audio
This is where the rubber hits the road, or in this case the dollars fly out the window. File hosting is relatively cheap. Current rates range from free (Dropbox, etc.) to $0.0300 per GB on Amazon’s S3 service. But data storage (uploading the file) is not equivalent to data usage.
In the case of our Serial file, uploading it to Dropbox the file size would be equal to the data usage. The file is being sent from our computer’s hard drive to Dropbox’s hard drive exactly once. This is the same as when we download something. The file size is the amount of data usage. This is why so many online services are free. The data storage is cheap, and for the most part the data usage is low when used by one person. The free services all have data caps. It’s just that most people never run into them.
The problem for podcasting is that you want lots of people to download and enjoy your show which is the direct opposite of what your file storage company wants. Dropbox has a cap of 20GB (gigabytes!) per day. Others are similar. As compared to a mobile phone data plan that seems like a huge amount, and it is.
That Serial episode is 49.8 MB, which is about the same as two raw image files from a mid-range DSLR camera. The most recent version of Firefox is around 80 MB is size. Podcast files are in that middle range where people won’t blow their mobile data cap, but are still large enough to be a headache to whoever is paying for the download bandwidth.
For the sake of this post (and that Amazon provides a nice data usage calculator) lets assume we’ve uploaded a podcast episode that is 50 MB in size to Amazon’s S3 service. But first, math!
50 MB = 50,000 kB (kilobytes) 0.05 GB (gigabytes)
Now that we have the file size in gigabytes we can figure how much bandwidth our show might use. One of the largest podcast hosts produces a show, and they talk about statistics. One interesting stat is that the median (half above, half below) downloads for all of the Libsyn hosted shows is around 140. So if your show gets more than 150 downloads, it’s likely to be in the top 50% of podcasts. While a show like Serial gets around a million downloads per episode. Also, to make cost-per-thousand (CPM) advertising worthwhile, it’s recommended to have over 5000 downloads. Then the CPM rate is multiplied by five. Any less than that and it’s not worth the trouble to deal with CPM-based advertisers.
Here are the bandwidth calculations for each example.
Median show: 0.05 GB x 150 = 7.5 GB -- Minimum CPM show: 0.05 GB x 5,000 = 250 GB -- Serial-class show: 0.05 GB x 1,000,000 = 50,000 GB (!)
Using Serial as an example the S3 tab of the calculator page gives us a cost of $4731.70 for those million downloads.
That’s just for one episode. The Serial team also releases one episode per week. My guess is that they’re spending around $20,000 in bandwidth per month during their season. Not all of those downloads happen at once, and this doesn’t count downloads of the back catalog. So it could be higher.
Even for the 5000 download show, the bandwidth would cost around $25.
Most of the podcast hosting companies (Libsyn, Blubry, Feed Press) charge around $15/mo. for their service. You can see this is a pretty reasonable charge for what’s provided.
Hosting a moderately popular podcast on Dropbox or a shared web host (Bluehost, Hostgator, etc.) can get you blocked quickly. Using the Dropbox limit of 20 GB/day, 400 downloads in a day would earn a block.
0.05 GB x 400 = 20 GB
A shared host that offers 1 TB (terabyte or 1,000 GB) per month would have an average daily limit of around 33 GB.
1000 GB / 30 days = 33.3 GB/day
Being a shared host, you also have to consider processor load. It takes CPU cycles to power the downloads. If the downloads take up too much CPU power, your account could be blocked. That means your website and everything you have on that server is inaccessible until the “time out” is over.
For the most part the established podcast hosts are a good deal and provide lots of other features like download statistics and RSS feed maintenance. Self-hosting the files can be done, and might even make sense for a first time podcast that has a high probability of staying in the bottom 50% range of downloads.
After looking at the numbers, I’ve found the following to be a useful guide.
- The no-budget option: Files go to Dropbox, and use a free WordPress.com/Blogger/Tumblr site to create the RSS feed. Cost $0. Downloads stop if you get popular, but hey it’s free.
- The WordPress.com Premium4 Option: Everything is done out of a WordPress.com Premium account. $99/yr includes the VideoPress option which will cover the bandwidth of a podcast. No download stats are available, but if you already have this podcasting is a free add-on.
- The all-in-one: Sign up with a podcast hosting service and everything is taken care of for you. Cost is around $150-180/yr. Download statistics are included and might be worth the cost if you’re shopping around for advertisers.
- The privateer: Host your files on Amazon and hope for the best. It might be almost free, it could bankrupt you. But you’re the sort of person that likes the risk.
This site is hosted on a WordPress.com premium plan, so adding a podcast is free for me. The 13 GBs of storage will hold over 250 episodes, so I think I’m okay there. The only problem is that I won’t have statistics. One option is to publish my feed through Feed Press ($4/mo.) and get the numbers that way.
There’s lots of options, and I thought I’d show my work when can to thinking about the costs associated with podcasting. I hope someone finds it useful.
- For this post I’m counting streaming as downloading. Because either way the file is pulled down from the internet to the player. Streaming is just a fancy way of saying the podcast/song starts playing before the file is fully downloaded. ↩︎
- Changing the sample rate doesn’t effect the file size. It’s simply a measure of how many times per second the sound waves are measured. ↩︎
- That’s creating the illusion of movement by raising the volume in one channel while lowering it on the other. ↩︎
- In the media settings WordPress includes everything needed to produce a single RSS feed, including the iTunes tags. ↩︎