Ii) Some Examples Of Digitalisation And Big Data

i) The Sloan Digital Sky Survey (2000) in a couple of weeks collected more than that amassed by the entire history of astrology. By 2010 it had accumulated 140 terabytes. Yet its successor (Large Synoptic Survey Telescope - starting in 2016) will generate 140 terabytes of data every 5 days

ii) The first decoding of the human genome (2003) took a decade of work to sequence 3 billion base pairs. This can now be done in 1 day

iii) Of the 7 billion shares traded daily on US equity markets, 2/3 are traded by algorithms using computer models that go thru mountains of data to predict gains & reduce risk

iv) Retailers (like Woolworth, Country Road, Tesco, etc) are using big data and customer relationship management to study customer shopping habits and predict future buying behaviours so that they are able to offer what the customers want to buy. For example, Country Road has a 2.2 m customer data base that includes information about when customers shop and how much they spend. Some results include

- A US retailer doubling sales in 5 years on a 5% increase in customer numbers

- Country Road aims to increase sales by 50% over a 3 to 5 year period by encouraging consumers to shop across 4 brands, ie Country Road, Trenery, Mimco & Witchery. Currently, there is a 38% crossover between Witchery and Country Road and this figure could be increased considerably and provide a major impact on profit, ie by shopping over the 4 brands it is worth 10 times the value as shopping with I brand

- In UK, Tesco used its close relationship with its customers, ie 13 million loyalty cardholders, to pioneer the use of Big Data. It is claimed that Big Data helped grow the business from $8.5b. in 1992 to market leader with $50+b. in 2011. Since 2011 Tesco has lost its dominant position to Sainsbury. In April 2012, Tesco reported its first fall in earnings in 2 decades. Some explanations for the decline in Tesco include

i) the claim that they have been too caught up in the science of retailing rather than the experience, eg instead of being greeted with hello at the checkout, shoppers would be asked if they have a loyalty card!!!!

ii) better value for their money elsewhere (James Chessell, 2014a)

v) Health

- Twitter updates were as accurate as official reports on tracking the spread of cholera in Haiti after the 2010 earthquake but were available 2 weeks earlier

- use of Google Flu Trends accurately predicted surges in US hospital visits a week before official government warnings

- a digital data collection network was found to be able to detect the SARS outbreak more than 2 months before the first publications by the World Health Organisation

- using now-available digital surveillance, the 2005-06 avian influenza outbreak "Bird Flu" could have been detected between 1 & 2 weeks earlier than official surveillance reports allowed at the time

vi) using Hadoop, Visa (credit card firm) reduced processing time of some 73 billion transactions from 1 month to 13 minutes

vii) Collecting & analysing 0.5 m daily prices of products sold in the USA, MIT was able to detect changes in CPI 2 months ahead of the official government CPI figures

viii) ZestFinance

- Uses technology to determine credit worthiness of small loans to people who have poor credit ratings

- Traditionally credit scoring is based on a handful of strong indicators like history of late payments

- ZestFinance use a large number of weaker variables in a matrix format

- in 2013 its loan default rate was 1/3 of the industry average

ix) Predictive Analysis (where you take all the available information in the system and analyse it). For example, in firefighting using predictive analysis to estimate the level of risk to each property and what will happen next when a fire occurs, ie what going to happen days in advance. It enhances weather forecasts of variables like wind direction, temperature, etc by factoring in other conditions like

- how close a property is to a national park

- what vegetation surrounds it including likelihood of nearby plants to burn
- what material the property is made of
- the slope of the ground
- the history of the area, etc

It can model the relationship between resources and community risk, and provide a link between government spending and community risk, ie it will reduce the chance of over-servicing some areas and under-servicing other areas.

It has application is beyond bushfires to other emergencies

Additional tools are the use of drones and ground sensors to fight fires. They are used to capture a more accurate, earlier indication of the risks and fires than satellite images and/or human eyes on the ground. The use of ground sensors placed on trees provides the primary indicator of fire conditions; then drones are used to fly over the location to relay video and information back to decision-makers.

x) Some Australian examples

- Australian Valuation Office (started in early 1900's) - competitors use Google Maps (under 10 years old) to gather information more quickly, cheaply, efficiently & accurately, ie

"...Google maps has already contributed to the demise of the 113-year-old Australian Valuation Office, and the ABS's long-standing house price index, which has been largely obsolete by the more frequent and accurate surveys produced by RP Data and the Housing Industry Association..."

Fleur Anderson, (2014a)

- Aust. Bureau of Statistics ($300 m budget with 3000 staff) - via the Internet, the relevant data that ABS collects & supplies is readily available sooner, eg inflation rates can be supplied in real time rather than released quarterly with a 2 month lag; similarly for housing market statistics, ie

The Reserve Bank already looks to the monthly inflation indices produced by Perpetual and TD Securities while it waits for the quarterly ABS official figures which comes with a two month lag, a lifetime in real time digital economy.

"...Now the Reserve Bank economists and the private sector will put jobs surveys produced by the ANZ and NAB ahead of the official jobless rate which comes to relying on credible information..."

Fleur Anderson, (2014a)

- Australia Post - being a courier for Internet shopping purchases is more profitable than the 'snail mail' services (the latter makes a loss)

- Cancer survey - normally takes 2 years to find a suitable sample; with registration on the Internet, it took 2 weeks

Understanding the statistical phenomenon called "regression to the mean", ie an extraordinary period of poor performance will auto-correct. In sports it is related to players returning from injury, shots stop hitting the post, misses become goals, etc so that the shine of good fortune returns

 

designed by: bluetinweb