Streaming Usage Running out of Steam? Or Just a Lull Before the Next iPlayer Storm?

[#1]

It's about time for another update on the monthly usage data and in particular our now regular focus on streaming and iPlayer. A few headlines to start with based on all customers (both business and residential):

• Streaming usage per day is starting to flatten, February saw a near 25% increase in streaming traffic per day over January, but by March that had slowed to 6.61% over February and April was up only 3.19% per day over March.
• Per customer, streaming increased by just 1.54% in April over March, a decrease from 4.66% in March and 23.04% in February.
• Evening (4pm - midnight) streaming downloads now account for 9.47% of the available network capacity. Up from 8.81% in March and 8.27% in February (and 4.86% in August last year).
• Total streaming (upload and download across the whole day) now accounts for 6.39% of all usage, up from 6.05% last month.
• The busiest hour of the day for streaming is 9-10pm where streaming now regularly accounts for over 11% of total downloads.
• The busiest hour of the month was 9-10pm on April 30th with 12.51% of the total downloads being streaming.

So it looks like we're getting towards the plateau in terms of the current trends of streaming. Of course this doesn't mean it will stay like this, our expectations are that this is just a natural levelling off which will ultimately take a further upward curve either as applications like iPlayer and ITV's recently launched catch-up service become even more mass market or as more applications become available (plenty more content on the horizon) or as the quality (and thus bit-rate and actual usage) increases. We might also see a shift in usage habits as different shows are on TV and with the weather. The sunnier days we've had this last week or so has certainly meant that there are a few less people online or people are going online later or for less time as they enjoy the long bank holiday weekend and the warmer evenings. This blog notes the impact that a show like Doctor Who has on our network. With the ability to watch shows at any time over the following week via iPlayer we might see more streaming later in the evening or when people are at work for example. 30th April was an interesting day, not only did it have the two busiest hours ever for gaming, but it also had the busiest hour of the month for streaming (9pm to 10pm, the same as the second highest gaming hour). The main "loser" in that hour was web traffic. Normally around 70% of traffic at that time of day web traffic was down to 64.7%. It's also interesting looking at how usage has changed over the last 6 months. This graph shows streaming usage over the last 450 days. It's quite clear to see now how quickly streaming has grown but you can also see the levelling off that's starting to happen (green lines are downloads, blue lines uploads, the darker colours represent the maximum amounts used, the lighter colours the average). Streaming now accounts for more than 10% of downloads on some of our products, when 6 months ago it was around 5%. A big chunk of that usage shift (in percentage terms and in actual usage per customer) is coming from P2P and Usenet. On average in August 2007 customers that had P2P usage used about 75MB per day while those using Usenet downloaded about 25MB per day (note: that includes both text and binary usage). Streaming was about 7.4MB per day. However, by April those figures have changed to P2P 57MB, Usenet 17MB and streaming 13.5MB per day. (Note: these figures only count those customers who had at least 1 byte of P2P, Usenet or Streaming). Average usage per customer (for customers that used at least 1 byte in the month) grew slightly from August to April from 208MB to 209MB per day with the highest figure per day being 222MB in February and the lowest was 207MB in December. Next month may well see streaming break the 10% of evening downloads, we'll be back in the middle of June to let you know. Dave Tomlinson PlusNet Product Team

[#2]

So in short are you saying that the iPlayer seems to be having only a small effect on total bandwidth used per customer, causing more of a shift in type of usage than an increase in quantity? Also, what category to iPlayer downloads fall into? Streaming or P2P?

[#3]

There's an increase in usage as well as a shift, the actual usage per month or per day has increased while the percentage of traffic is shifting towards streaming and away from P2P. Take a look at the monthly usage patterns on the sheet on this post: http://usergroup.plus.net/forum/index.php/topic,6064.0.html I'll update it next week hopefully with April's data (it takes a few hours to produce). iPlayer is indentified as Streaming (both the Flash and download versions).

[#4]

Dave, Nice info, but please please please can you be clear and unambiguous in what you write. For example: "Streaming now accounts for more than 10% of downloads" - what does that mean? 10% of all downstream traffic, or 10% of some group of traffic which you're calling "downloads". It may be obvvious to you, but it's not to me (and I'm a Plusnet punter). There are lots of folk following these articles in interesting places, it would be good for them to not misunderstand. Thanks!

[#5]

Good point. 10% of downloads means 10% of all downstream traffic. Will try and make sure it's fully clear in the future.

[#6]

The bbc just released their figures for Iplayer usage and it was reported at: http://www.digitalspy.co.uk/broadcasting/a96400/bbc-iplayer-use-grew-22-percent-in-april.html I loved the final comment at the bottom: "The BBC figures suggest the behaviour of users on PlusNet, a relatively small ISP, was not reflective of others in the UK."

[#7]

Perhaps it's a reflection of our relative customer make up? Also, we've not been afraid to talk to our customers about iPlayer for a while now, so perhaps we shifted the take up curve earlier?

[#8]

Perhaps our customers are more likely to be early adopters than other larger ISPs? So maybe this signals that the Beeb's own growth figures may slow next month or the month after?

[#9]

[...] Examples of both of these can be forwarded to us for analysis. Whilst this may not result in immediate improvements, it will help us and Postini to further refine the accuracy of the spam filters. We do expect messages from some mailing lists and bulk mailers to be incorrectly identified as spam due to some of the characteristics they exhibit. In many cases the best way to deal with these will be to add the originating addresses to your approved-sender white-list (See ‘What’s Next?‘). [...]