While you wrote an extensive post, your mostly describe what actually happens already and I find few points that I can directly answer, but I'll try:
I didn't mean to ask anything - largely because I think I get the transient nature of data and the implications it has on the supply side of an MNO business - and instead was simply trying to in part by analogy set out that the technical considerations simply don't cut it because of the nature of end users. I went with the train analogy rather than the hotel rooms conundrum.
1. You're right that it is the Network Operators and ISPs that did the conversion of bits/second to bytes in constructing data products. You're wrong that it's just the MNOs, it's an idustry standard. This conversion is actually alien to the underlying technology and underlies most of the misunderstanding that arise by selling a 'bucket of data' when you actually sold access to a network for a specific period of time with a limit on usage.
Well I indicated Telkom and the MNOs, and I don't think its a broad industry standard but far more of a South African practice - there is probably a dozen other markets where it has happened but not enough to be broad - the idea of deeming "overuse" according to total data transferred over a time period is quite different to the idea of selling data by bundle sizes. If you look at "wireless providers" in the US -
AT&T and Verizon - which pretty much matches VC and MTN they have moved towards caps not from them and the approach towards overusage is not an OOB shark but rather to charge a big bundle fee with a big bundle. The bundle sizes in play matter here and I am afraid the rates at under R100 a gig are significant.
Now I get that our market is a very different animal, and I strongly believe that the US approach to cellphones might work for the US (and would be of benefit to some high usage South Africans) but it won't work in SA, but the market is certainly lucrative as the financial statements show.
I suspect though that a CEO of a competitor to Vodacom actually made the call towards entrenching buckets of data.
The correct way to sell such a network would to make available a certain bit-rate for a specific time, say 1Mb/s for 30 days. Of course this would put the service financially beyond the reach of most people. Inversely, if I have to ensure the price is low enough (say R100 per month), the bit-rate would be so low, it would be unusable, something like a few bits/second. Unfortunately neither of these work for most consumers (enterprises buy this way, BTW); subscribers want sporadic access at the highest possible speed. The only way to do this is to convert to bytes. If you have a better way to sell such a network, I'm all ears.
A detailed proposal would only be sent with an invoice after being solicited
However part of your costs problem is found in the fact that you are setting time frames of 30 days - which is a long time and frankly is still orientated towards VCs billing system.
But as an idea into the hat:
I imagine you'd find quite a few people on this forum who wouldn't object to a well priced prepaid product for downloading distros (you know all the linux users) that is available in time capacity blocks of 512k/convenient factor (down/up) and 15mins on a repeat basis but can only be purchased online. So from a sales portal I select my time blocks, capacity and repeat cycle for up to 30 days and check it out - I can also download download managers and so on - pricing for time blocks is set according to the time of day which is built around demand but you (the MNO) have more flexibility so it doesn't occur that the network has every user who wants night owl rates kicking in at the same time. My thinking for implementation would be a special APN (which I am sure would allow authentication on a sim basis and which can run side by side with a "normal" implementation) and the user having their PC side of things "dialing in" through either a smartphone or usb modem the software on the PC side can get the schedule to know when to work (a nicely designed console would ideally allow the PC to have another internet connection and would have a download manager running, getting a newsgroup reader to "always use dialup" isn't a problem and the same applies for mail etc ...). The product would also be suitable for an entire array of uses where a batch of data needs to be transmitted on a daily basis but you don't require "always on" connectivity or you have always on connectivity through 3G but the big transfers are periodic - a mail server at a remote site, offsite backups, providing for an employee to work from home or on the road and accessing a corporate VPN. Apart from banking machines I think you could pretty much absorb the residual dial up market if it is well priced, hell a product of 15 minutes at 512k daily at R25 or less a month is pretty convenient and affordable for the "rural areas" - and Telkom would probably like to mothball the copper . With some network overhead and so on I reckon a base session amounts to transferring 50 megs and so you should get pretty close to a gig and a half a month. I suspect a fairly easy software upgrade on the Webbox could also see that product using the service. You also have possible applications for content sales - purchasing music etc ...
Of course using the 1Mb/s for 30 days benchmark, can you provide such a product with consumer grade contention ratios and consistency for R480 a month because then you could compete with
TI ADSL, even if we use the business rates you looking at less R800 a month. And I am the first to acknowledge that the mobility factor comes into play but if your goal is 3G as a general broadband product that is the competition to look at.
Morevover if an ECNS licence holder were to approach VC with a business proposal of a "wireless local loop unbundling" nature what would Vodacom or MTNs response be - I am hoping that the Afrihost MTN arrangement moves in this direction, but it shouldn't be difficult to see where I am going. As an additional thought the notion of selling blocks should probably be run as a fairly autonomous business unit - office space in Jbay is cheap
BTW, I'm one of those "
guys that work out the retail products[\i]" and I do believe I understand the landscape - but if you have any suggestions on how to do the conversion while losing the time-component, I'm ready to listen. 
See above. Also my point isn't that the time component needs to be lost it is that you can't rely on the underlying system as an argument for a pricing model "defect" (perceived or real). And honestly you seriously need to have words with the marketing people and internal corporate communications - because the message needs to be: so your usage is about 200 megs a month you should get 500 megs a month, with data usage more is always better; with all of VCs tv adverts I can't recall a single advert delivering the message that you should look to secure data at 150% of your estimated usage.
As I see it the longer a "bucket" of data has validity the more expensive it should be, so I do not see why Vodacom cannot sell non-expiring data at R200 a gig other than that it hurts sales of small expiring bundles with some users then purchasing a gig to last for a year or so for a not-so-smart phone; which is a valid business reason not to do it - but then it needs to be said that the decision is profit driven and accept the criticism that there is pillaging happening.
2. You make a few suggestions on what should happen at bundle end:
a. Pay OOB - we have this in place including the ability to pay OOB=IB (though you don't mention it as an option)
b. Throttle the sub - this is actually quite a bad idea as the vast majority of users don't understand this and end up experiencing what they believe is a bad network. Even if you tell them.
c. Allow the sub to buy another bundle - this is the option I also prefer and thus you get a notification (soon with click-through links) at 70%, 90% and 100% to do exactly this. Busy making this even more sophisticated.
You forgot an important 4 option: Stop the access - We do this today with TU and CLL.
My suggestion isn't that one of the options be followed indiscriminately but rather that the user gets a choice which can either be made in advance or once the cap is hit (in the same way of those bank authenticating thingies work). I also added that there are obvious and not obvious problems with the approach.
In any case, back on topic - so we've determined that Vodacom and MTN offers roll-over the others operators don't. And yet they get attacked for it while the guys that don't are not even in the discussion.
Well we have also established that the other operators are significantly cheaper - at least outside of the large enterprise clients - and I've presented the position which I think is broadly accepted that VC and MTN have better coverage and so forth which does impact on their costs to deliver.
As I've often posted on this topic, it seems the only way out of this is to remove roll-over so as to avoid any confusion on the benefits.
I don't think that would help your situation because there is now a reasonable expectation of roll-over - and it is this problem of reasonable expectation and the creation of a consumer expectation that comes into play.
