Author Archives: Michael Griffiths

The Newspaper Industry is Changing

By

Martin Langeveld has written a post proclaiming the impossibility of charging for online news. While I think his fundamental conclusion is correct, I think that his analysis is both incomplete and misleading.

His initial premise is controversial. Namely, that charging for access to online news has price sensitivity; or, that demand is elastic according to price. Mr. Langeveld presents some numbers which he says “come from the top of my head” and concludes that newspapers can never match current
online advertising revenue with subscription revenue.

I don’t know how accurate his estimates are. There may even be a much larger drop than he predicts given any charge at all. Economically, the conclusion would follow from an assumption of moral hazard; as people consuming news currently don’t have to pay, they’re likely to over-consume news. If you start charging at all, they will stop “over-consuming” to that degree. Of course, perhaps we want citizens of a democracy to “over-consume” on news, so that they are informed – but that’s a philosophical & normative debate.

And, as some commentators have pointed out, his estimates of advertising revenue may be flawed. There may also be greater leeway for premium content, micro-fees, etc one people are signed up and you have their credit card information. Cross-selling and up-selling, in other words.

Regardless, it’s less important that the systematic shifts in “news media”.

First of all, the costs of (1) publication, and (2) distribution have dropped to “effectively zero” for small publishers. This means, basically, that newspapers have dramatically more competition – they are no longer the only game in town.

Secondly, the role of newspapers is the modern age is unclear. Their role pre-internet was clear; that is, they did three things. They (i) aggregated news [local, industry specific], (ii) published news, (iii) distributed news to people interested, and (iv) connected advertisers to an audience.

The problem is that on the internet, there is no value add from (ii) and (iii) [publication and distribution], then that leave only aggregation as a value-add. However, modern technologies have (arguably) “solved” the aggregation problem, and in a much more elegant way than newspapers (who acted as gatekeepers). Online aggregation, based on search technologies, can locate and present related content after it’s been published – from multiple sources, and arranged via topic.

Newspapers lose their gatekeeper power. They devolve into brands – filters, really, who guarantee a certain minimum quality – as well as content production companies. It’s easy for newspapers to pair a photo with a story, because they employ both photographers and writers.

But that’s a much weaker position than their previous one. It’s also one with lower barriers to entry, and thereby more competition. The low cost of publication and distribution means that anyone can set up a blog and start writing; and with Google Adwords and similar technology, the advantage of (iv) – selling advertising – is also much reduced, so anyone can make money.

And, when all your competition is free (or advertising supported), it’s hard to charge for your content. Most industries have learned that it’s hard to fight the “laws” of supply and demand – the newspapers are no exception.

When is a disadvantage not really a disadvantage?

By

Answer: when you’re giving an interview.

Are there disadvantages to continuous analytics?

The main disadvantage of stream-based continuous analytics is the incorrect perception that stream processing is solely for real-time applications, rather than the more general — and much larger — problems facing data analytics today.

I cam across this gem when reading up on Truviso’s Continuous Analytics Platform, a tool which does pretty much what you expect it to, which is the analyze just incoming data. There are advantages to this; for instance, a pretty common task in data analysis is to look at how are things changing? You can look at a trend graph on-demand, which would involve querying a lot of data and take a long time, or you can perform the same query multiple on small batches of incoming data and store the results.

As far as I can see, that’s all Truviso is really doing. They have an efficient, and fast, add-in to a SQL-based server that will perform a set of pre-defined queries on an incoming batch of data. Looks useful for certain cases – mainly when either your queries are really expensive so you don’t run them often, or you need to be sensitive to changes.

Truviso seems to be targeting the former situation in the section of the interview I quoted. The answer, however, is both entertaining and downright wrong: he’s addressing a misconception about what Continuous Analytics is capable of, and not an actual disadvantage (or limitation) of Continuous Analytics. 

In other words, he’s sidestepping the question and answering the question he really wanted to answer – one that shows Truviso in a positive light.

It’s also classical “job interview” technique.

Hume and Critical Realism

By

It is astonishing how much freedom I feel after a rejection of Hume-an Constant Conjunctions; the twin ideas that the limitations in perception in turn limit the knowledge one can acquire to that of observing “constant conjunctions”, and that the weaknesses of induction prevent one from locating real “laws”. Hence the absence of “laws” in science, in favor of “laws which haven’t been disproved yet”. Critical Realism allows a rejection of Hume-an Constant Conjunctions on an ontological basis. Such a rejection re-defines science, from the task of locating constant conjunctions to the task of locating reality’s ontology and subsequently creating theories to explain how they operate. While the result is that I approach science in largely the same way I did beforehand, I know have a considerably firmer philosophical base from which to operate – and one which serves to explain how science both does and should operate. Hume, and his legacy of Positivism, does no such thing.