Thursday, January 27, 2011

Arguments for a cost of 5 cents per GB for usage based billing

In my post yesterday, I argued that I think that usage-based billing (UBB) does make sense, but only if the rate is on the order of 5 cents per GB, and dropping over time. ISPs in Canada are charging 10 to 100 times this, which makes downloading video outrageously expensive.

My argument yesterday was that 5 cents per GB would be sustainable from a consumer's perspective since the cost of downloading a high-definition movie would be limited to be around a dollar (above and beyond the rental/purchase price of course).

Today I will add some additional arguments, both in favor of very cheap per-GB rates, and also in favor of usage based billing at those low rates. I will try to be conservative in my estimates so by arguments will be hard to challenge.

What is the marginal cost of providing an extra GB of download to a consumer? We don't know exactly, and it will vary from neighbourhood to neighbourhood, but we can do some 'back of the envelope' calculations to show that it must be extremely small.

In the worst case, the ISP's lines to the network might be near capacity. In this scenario, the ISP would need to provide an extra line along its existing conduits, with the attendant switching hardware and head-end equipment. These days, 1Gbps lines are typical. What return on investment could an ISP get from such a line at 3 cents per GB (which is less than the 5 cents I talked about earlier to allow for wholesaling and profit)? The total number of GB the line could serve in a month would be 328725. If that entire number of GB were sold, then the revenue would be $9861 per month. In reality, the line will only be used to capacity at peak periods, and it may take years before it reaches capacity even then. So lets imagine it is used on average at 25% capacity over a 10 year lifespan, that gives revenue of $2465 per month. If this was used to pay interest at 5%, then the ISP could afford to pay $232,000 to install the new line. I am certain that in the vast majority of cases, the line would cost less than this.

Today there is a report from Netflix, which suggests that the actual marginal costs would be 1 cent or less for each GB. Of course they have a conflict of interest, so they are likely to argue for a smaller cost. However, it should be clear beyond any doubt that the 50 cents to $5 currently being charged by Canadian ISPs is vastly out of line.

Netflix argues that at one cent per GB, usage-based billing becomes nonsensical. However let's take the ISP's perspective for a minute. If you allow unlimited downloads, some people will decide to download all the free video they can find on the Internet. We can play devil's advocate and imagine an obsessive-compulsive character that finds a way to download video 24 hours per day; at 10GB/h they could consume 7300GB per month; I think it fair that such a download-fiend might be expected to pay the $365 he would be charged at 5 cents per GB. Clearly he would be consuming capacity at a level greater than the ISP's marginal of providing the service.

Let's look at two other cases: The first is the person who entirely replaces their 7h of cable-TV-watching per day, by watching 7h of Internet video per day. They would consume 2120GB per month, and at my rate it would cost them $105. This would be a lot more than the equivalent cost of buying cable TV. However I think the extra cost is justifiable: Cable TV channels broadcast to multiple customers at once so makes much more economical use of the infrastructure. Put that $105 in perspective though, at the current ISP rates our Internet-TV watcher would be paying $1050 to $10500. Obviously excessive. Yet Internet TV served by multiple vendors is clearly the way of the future. Cable/Phone/Internet companies are only protecting their existing duopoly by charging per-GB rates that make this impossible.

The final case is the person who rents three two-hour movies per week, or watches 6 hours of ordinary TV per week. That would take 260GB per month at the good-quality 10GB/h rate I have been using, and would cost them $13. At current ISP rates it would be between $130 and $1300.

There is a lot of backlash in Canada against UBB and the recent CRTC decision forcing independent ISPs, which buy capacity wholesale from the the major operators. to be subjected to the same excessive UBB rates, with only 15% discount. I suggest that the independents should be either able to buy a fixed percentage of total capacity on each line, or else be charged a wholesale UBB rate of something like 3.5 cents per GB.

It is  clear to me that the CRTC made a terrible decision by failing to consider the actual cost of capacity.

Tuesday, January 25, 2011

Usage-based billing for Internet service: Only evil if it is out of line with incremental costs and cramps people's style

Most of us pay for electricity, water, and natural gas by usage-based billing (UBB). Yet there is a great outcry against Internet Service Providers applying UBB.

An example of this outcry is this opinion piece by David Beers in today's Globe and Mail. His basic argument is that the increasing tendency for ISPs to use UBB will kill Canada's ability to compete in the global information economy. He feels his Canadians will stop being able to afford to download video.

His argument is valid to the extent that ISPs charge excessively, which to me means well beyond the incremental cost of providing additional Gigabytes. However, that does not invalidate the entire concept of UBB.  I believe that a good argument can be made for metering at a rate of about 5 cents per GB to allow ISPs to recover the incremental costs of extreme volumes of use.

When I was a child in England, phone calls were all paid for using UBB. Even local calls cost money, and long distance was considered an expensive luxury. This definitely had an impact on everybody's life; people avoided using the telephone, and kept calls very short, for fear of raking up huge charges. Over the decades, the incremental cost of providing minutes of phone service, including long-distance, have diminished to almost nil. The main cost to phone companies is the fixed infrastructure, which is why it is common now to pay a flat rate for unlimited calls.

However voice is cheap to transmit digitally. High definition video, on the other hand, requires vastly more network capacity. So the incremental costs have not yet come down to the level where 'unlimited' Internet usage makes sense. This is especially true for wireless services, where spectrum and towers are a limited and expensive resource, but is true also for cable and DSL landline service, where there are limited-capacity lines into neighborhoods that are shared. ISPs used to offer unlimited plans, but they have largely stopped doing this due to the fact that today's heavy Internet usage is largely due to watching video.

In the following I am going to use a round figure of 10GB per hour for good quality video (in practice it can range from 2GB/h to 40GB/h depending on the quality you want). If a two-hour movie costs $5 to rent, then paying $1 to download it would not be unreasonable. But download prices in excess of this would indeed likely start to cramp people's Internet usage style, as Beers suggests, just like telephone usage style was cramped when I was a child. Rough calculations suggest that five cents per GB might be a tolerable metered rate for wired Internet, with even that amount diminishing over time as networking technology gets better and cheaper.

If you look at Rogers' Internet service rates, this is exactly ten times the 50 cents per GB they currently charge. But you only get this rate if you are paying for their 'Ultimate' package, at $99/mo, and exceed the 175GB cap. Unfortunately, people with 'Ultralight' service pay 100 times my suggested 5 cents per GB base rate (for $28/mo and when exceeding the 2GB limit). This to me is clearly excessive, since the incremental cost to Rogers is the same. Rogers defines their service levels based on three factors: a) speed in Mb/s), b) the usage cap, and c) and cost per GB for excess usage. I think that the cost per GB for excess usage should be fixed for all service levels at my suggested rate, so as to not overly cramp people's style. Note that competing Internet services in Canada, such as those from Bell, are roughly comparable in terms of their charge for excess GB..

I think that the ISPs are charging these rates currently for two reasons: Firstly they don't want competition from other video providers (as they provide their own video services), and secondly, they can get away with it since in recent years, most users have only consumed a small amount of video. However if we want proper competition for video, and stimulation of innovation, the cost of obtaining it through cable TV, satellite TV or through Cable, DSL an fibre-based Internet need to be roughly similar.

It should be noted that cable operators are already sending vast volumes of GB/s into each neighborhood by way of digital HD video channels, including on-demand video. They don't charge anywhere near 50 cents per GB for their own on-demand video service, so the argument that this is their incremental cost is facetious.

Consumer awareness and pressure may help to push the rates into  'reasonable' territory. Regulatory action may also be needed. Unfortunately competition in the market is not going to work on its own since Internet service operates as a cable/DSL duopoly.

Unfortunately arguing that UBB is evil and unlimited service should always be provided is disingenuous. It damages the ability to make a sound argument for a reasonable 5 cents per GB metered rate.

Wednesday, January 19, 2011

In trepidation of the zero-button iOS device

There are rumors floating around that the iPad 2 and future iPhones will lack a 'home' button. Instead you would have to use gestures to do the things that currently are supported by the home button.

I think this is clearly a move in the wrong direction. Apple has come up with many smart usability innovations, but its obsession with 'the fewer buttons, the better' is not one of them. Since the very first Mac in 1984, Apple's computers sported one-button mice. Unix devices of the time, and later Windows devices all had mice with more than one button, and for the most part users found this quite straightforward to deal with.

An interface with too many buttons will clearly be more complex. But reducing to zero also adds complexity. Without a home button, users will have to memorize gestures. How to get back to the home scren will not be 'discoverable'. Also it will be increase difficulty when trying to use a device without looking at it, or with gloved hands.

As I said in my earlier post, I think that a three-button device would be better.

Let's home the rumour of zero buttons turns out to be false.

Tuesday, January 18, 2011

Thoughts on Tethering: How to keep and improve its flexibllity and usefulness

Tethering (using an iPhone or other smartphone to connect a laptop to the cellular data network) is a feature I use frequently. I use it when I am waiting for one of my kids to finish a sporting event, when waiting in a garage for car service to be complete, or when on a train trip – it is currently far faster and more reliable than Via Rail's free WiFi for example. I also us it when a passenger on a car trip. For example, when working on a pressing deadline I maintained an almost continuous Internet session on a road trip from Ottawa to London down the 401 highway. I also sometimes find myself using tethering when parked at a parking lot to do a quick task that requires my laptop.

Luckily my wireless data provider (Rogers) doesn't currently charge extra for tethering, or restrict it entirely, as is the case with providers in some countries. It would annoy me tremendously if they decided to change this policy and treat tethering usage as a separate billable item. All I am doing, after all, is using up some of my allocation of data usage. I think all data usage should be treated the same and billed in the same way. I do find it acceptable to throttle connections that try to maximise bandwidth use on a 3G network for extended periods (e.g. video download for extended periods) if total bandwidth at a cell tower (on the cable to the tower, or from the tower to connected devices) is near capacity. However, that should not affect how the data usage is billed.

My ISP sells separate USB devices for Internet access. It seems a waste to have to get one of these when you have a perfectly good smartphone that can do the job. I would not want to carry around another device that happens to be small enough to get lost easily.

Tethering can be done on iOS using a USB cable or a Bluetooth connection. I normally use a cable since it is distinctly faster. iOS is purportdly going to offer instant WiFi hotspot capability in the next release; essentially this amounts to a capability to tether up to three devices at once. This would certainly be useful in many contexts, such as enabling two people on a train to work together. Many iOS apps for sharing data require both devices to be on the same WiFi network. Hopfully the WiFi hotspot capability would also work for these apps. I would like it if you could create a WiFi hotspot when not even connected to a data network. For example, it could be used in a foreign country or the wilderness  just to share data between several devices. We will see what level of flexibility Apple comes up with. I hope they won't be too much influenced by the demands of cellular carriers.

There is one thing, that currently bothers me about tethering on the iPhone: Whenever I connect my computer to the phone to charge it or back it up, a tethering session is established. I can turn tethering off, but then I have to painstakingly turn it on and back off again repeatedly. On the flip side, whenever I connect my computer to the phone to initiate tethering, iTunes initiates a synchronization/backup session. Finer control is needed. I therefore suggest the following behaviour should be implemented in iOS:
  • There should be an option when you turn on tethering in iOS to declare: 'Do not activate tethering when connected to a WiFi network'. If my phone is connected to a WiFi network, I can most of the time directly connect my computer. This would prevent unnecessary tethering when at home or work.
  • There should be an option in iTunes that says, 'Sync iPhone automatically at most once per n hours.' I think 2 would be a good default for n, but the user could easily adjust this, and can always sync manually.

Thursday, January 13, 2011

Umple update: Top November post on modeling-languages.com

Here's the status of the Umple project:

My post about Umple on modeling-languages.com was rated the top post in November. This shows that there is a lot of interest in the technology.

Some important improvements to UmpleOnline have just been installed. These make it much easier to lay out UML class diagrams. As you drag classes around, associations now don't cross their class boxes. If you have recently used UmpleOnline, you may need to clear your browser's cache for the new version to be loaded.

A couple of teams in the ICSE SCORE student project competition have been using Umple, and I have been very impressed by their work.

Our open-sourcing work is progressing and we hope to complete the transition to Google Code in about a month. Meanwhile Umple can still be accessed from the Umple home page.

My collected posts about Umple can be found here.

Thursday, January 6, 2011

Mac App Store - A good idea with some rough edges

The Mac App Store went live today, and I have had a look around. In this post I  give some observations about pricing and usability.

In almost every detail, Apple has copied the iOS App Store that has proved so successful. This is both good and bad. The good is that it is a familiar interface; the bad is that it brings along some iOS app store weaknesses such as being too busy, and lacking in discoverability.

I think having an app store is fantastic. The ease of installation, purchase and update will make consumers install (and pay for) much more software. The benefits are clear for consumers and for new and small developers, or developers with inexpensive apps: the benefits for these developers will trump complaints such as the  fact that Apple reviews all submissions and gets a 30% cut.

Apple's cut on purchases as a disincentive for developers: Thankfully, it looks as though Apple does not intend to restrict alternative ways of installing apps, so developers can continue to use their own sales and installation channels. This is critical, since I predict that many existing app developers, and developers of expensive apps, will be reluctant to sell on the app store. The 30% cut apple charges is unreasonable when the price exceeds about $15. I think Apple should have a sliding scale that reduces its cut for more expensive apps. How about 30% of the first $15, 20% of the next $15 and 10% above this level. That would encourage developers to use this channel for their more expensive apps.


Flat pricing as a disincentive for consumers: There is the lack of a sensible way to 'try out' expensive apps. The price points of Mac apps are in general much higher than iOS apps. If all I have to spend is 99 cents each, I am willing to be bold and download a bunch of apps that may prove to be not very useful. But there is no way I will explore a bunch of $20-or-more apps. And even if I do want to use an app, I am highly reluctant to pay $40 or more for an app I might use only once or occasionally. Apple needs to do two things, both of which should be win-win propositions:
  • Firstly Apple needs to build into all its sales (above, say, $5) an automatic 'demo' mode, where you can use the app for a couple of days at no charge, and only be charged after that period if you don't 'return' the app. With the Digital Rights Management (DRM) Apple has built into its store, this should be quite easy to do.
  • Secondly, Apple needs to provide a capability to charge for the app on a metered basis, up to a certain maximum. For example, I might want to use a particular image-editing capability for just one small operation, and perhaps imagine I will use it 3 times ever. If an app for this costs $50, I would be reluctant to pay, especially if I am not even sure if the app can do precisely what I want. I would like to be able to buy the app on the understanding that I would only be charged $5 for every day that I make any use of it, up to a maximum of $60. This would cut my risk as a consumer, and would benefit developers since it would attract many more sales of their expensive apps. Apple has a micro-payments infrastructure in place, so this shouldn't cost Apple much. I think a 20% premium for this would be reasonable, hence my suggestion above that the cap is $60, whereas an outright initial purchase would only be $50. I would buy many, many apps on this basis, that I otherwise would never try.
Deletion and accidental installation: There is no obvious way to delete an app, even a free one. In your list of purchases the only choice available is to go to the app's page. I manually deleted an app from my Applications folder, but the app still appeared in the purchases list, now with an 'Install' button ready for you to reinstall it. Uninstalling has always been non-intuitive on iOS; now it seems impossible to do so without trace. Apple needs to provide a 'uninstall' and 'delete' menu items wherever it displays the word 'installed' in its interface (it should also have 'return' as an option for demo mode, as discussed above).

I also found it too easy to accidentally install apps. Like with the iOS app store, once you have signed in you can do several operations over a short period of time. Simply clicking on the word 'free' installs an app with no prompts. There should always be a confirmation before installing any app.

Discoverability and search: Even with only 1000 apps on opening day, navigation of the apps to find good ones was tricky. They are categorized into broad categories, and you can search by keyword. However, the interface is busy (like that of iTunes and the iOS app store). I would like the ability to list or search for 'tags' that would identify features within categories.

You can examine an app in the store by displaying the entire page describing the app, or you can show lists with just the short name and icon of each app. It would be much better if you could also see lists of apps with an intermediate level of description, perhaps 3-4 lines describing the key features.

Also when you search by keyword, the 'hits' that result don't show the context of the keyword, in other words, I want to know where in the app's listing the keyword appears. This is irritating at times. I would like search results to show me the sentence containing the keyword, as Google search does.


A final usability issue: The App Store was very fussy about allowing me to paste text into it when writing an app review. Beware of writing a review in an external text editor, you may not be able to paste it into the app store.

Incidentally, most of the above issues apply to the iOS store too