No results found

Encoding trends from our first 20M files

Since we launched just two years ago, Zencoder has successfully encoded over 20 million audio and video outputs. (This only includes successful video and audio encodes.) In a growing startup, it's hard to know what to think about a milestone like this. On the one hand, it's exciting - 20M is a lot of video and audio files. And there is something satisfying about rolling over a bunch of 9s on an odometer, even when we all know that there is nothing special about round base-10 numbers. Why didn't we celebrate our base-2 1000000000000000000000000th video back in January? On the other hand, in the grand scheme of things, 20M files is tiny. We're already on pace to more than double that in 2012 alone, even if we don't grow. (We certainly intend to grow.) So we're not too concerned about the 20M number. We're more excited by 200M. Or 0xB000000, or 0b10000000000000000000000000000. And we plan on getting there soon. But we're still going to celebrate, and in true startup fashion, we're going to celebrate by mining some data. What interesting trends can we pull from our last 20M encodes? Here are just a few; we'd like to publish more soon, and if there is anything in particular you'd like to see, let us know.

Average number of outputs

Many of our customers still just encode a single output for each input file they submit. But more and more customers are doing multiple outputs: targeting multiple bitrates, resolutions, and devices. For those customers doing more than 1 output, the number of outputs trends upward:

Growth of HD video

Error rate

Zencoder can't support every file. There are a handful of codecs we don't support (go2meeting, I'm looking at you); we occasionally do something wrong; and some files are just corrupt. But our actual error rates are pretty low, and are trending in the right direction. This chart shows our overall error rates.

Overall growth

What would you like to see?

We'll be mining more data over the next few weeks, and we take requests. What sort of information would you like to see from the last 22-or-so months of encoding that Zencoder has done?