Tuesday 21 December 2010

How do the NBN financial assumptions stack up?

The NBN business case is out - if you want to wade through all 160 pages of it (like I did), you can find it here. You can follow a reasonably informed debate on this at Catallaxy, where I originally posted much of this comment.

There are a number of assumptions driving the financial models underlying the business case. As with any model, you can tweak the numbers you put in to get any result you want. Garbage In, Garbage Out.

The amount of revenue the NBN will gather is based on two things - how many customers it signs up, multiplied by the average revenue per customer.

The business case gives us these numbers.

The average revenue per customer is termed ARPU - Average Revenue Per Unit. You are a Unit - get used to it. I am a Unit on a Bike. Or a UNOB. Hmm - don't like the sound of that much.

The plan states that the ARPU will be $34 per month. It's worth looking at how they calculated an ARPU of $34 per month.

There is a table on page 101 that sets out the wholesale prices for each plan

12mb - $24 month
25mb - $27-30 month
50mb - $34
100mb - $38
250mb - $70
500mb - $100
1000mb - $150

There is a graph on page 118 that shows the subscriber split by speed.


I check these things carefully, because I used to do this sort of modelling at work from time to time, and I know how much bullshit went into the models.

From eyeballing the graph, I get the following splits for 2014:

12mb plan - 53% at $24 month
25mb plan - 16% at $28.50 (splitting $27 and $30)
50mb plan - 24% at $34 month
100mb plan - 7% at $38 per month

From that, I calculate an ARPU of $28 per month ex GST.

In 2020, you get:

12mb plan - 45% at $24 month
25mb plan - 5% at $28.50 (splitting $27 and $30)
50mb plan - 15% at $34 month
100mb plan - 25% at $38 per month
250mb plan - 10% at $70 per month

This gives an ARPU of $34, but it is driven by the truly heroic assumption that 10% of the population will be on a 250mb plan with a wholesale price of $70 per month.

It then goes on to assume that by 2028, 40% of the population will be happy paying a wholesale price of between $70 and $150 per month.

I call that "fantasy". I'd love to see the hard data that supports those assumptions about what Australians are willing to pay for their broadband.

3 comments:

Anonymous said...

I think there's 10% of the population that always wants the best and the fastest, even if the average westie won't be interested. So it's probably not unreasonable.

Going from $28 to $34 over 8 years only requires an increase of 2.5% per year. So it really only needs a few people to upgrade each year, which could well be encouraged by the applications that NBN speeds will make possible.

Simon said...

Anonymous is speaking bull. This is as you say a clear case of garbage in to get the result that is wanted. Projecting technology take up rates so far into the future is always going to be nonsense. If it were a private business a much high rate f return would be required to give a faster pay back over a more predictable time frame (as well as being higher to simply justify the financial risk). And for such a large investment a pilot project with measurable take up rates (or more likely multiple rolling stages) would be done before commiting to the whole thing. They have no data to support the take-up rates. The last time we saw business plans like this was when AOL merged with Time Warner and didn't that turn out well (and at least they had a couple of years worth of data to work with). These numbers are politically not commercially driven and it is all drivel.

Peter Dyson said...

How many nation building projects would not be started if they had cost-benefit analysis done against them? Sydney Harbour Bridge? At what point is or should vast nation building projects be less subjected to being forced to turn a profit or not exist? Public parks are wasted real estate oh very high value in some cases, they don't make a profit, why do we allow these spaces to exist then? Where was the cost-benefit analysis of these places?