Big data. Big bucks or big bubble.

Big data is the buzzword of technology companies of today. Gaming companies, marketing companies, banks and governments seek to jump on the bandwagon and explore this murky field and monetization potential. OCBC, Ten-cent are companies that seek to utilize / profit from aggregated / segmented data groups (Demand side) and companies like Singtel and Silver Lake axis seek to provide analytical solutions (Supply side) . I did some research to distill my understanding of big data, and whether it is a big bubble or potential big bucks. 

Macro view
Western societies generally have strong data protection and privacy laws, while Asian / Chinese societies are not as uptight on the privacy and sharing of such information.

Facebook is currently facing a sharp correction in its share price from the failure of its data protection schemes, which is worsened by the upheaval from management fallout. It's profit margins are severely eroded due to data protection policies and increasingly saturated advertising market. Data privacy acts are written into legislation in western societies and can be enforced by fines, whereby Asian societies are unconcernedly laid back about it. 

Compared to Tencent, the Chinese willingly accept social profiling and <loyalty to CCP> profiling which is absurd by western standards. This is also apparent in Singapore,  whereby security cameras are situated throughout Singapore, and there is active hate / racial / controversial comments surveillance by government authorities on social media platform.

SWOT

Pros (potential)

1) Client profiling and targeting to sell financial products that client needs (risk, return, liquidity). Wealth management needs.

2) Risk profiling for credit and insurance purposes
Quote
credit card company had a large amount of data on people and normal analysis segregated the people to risk level to default. They found a certain attribute which was ranked low by the normal analysis was in correct and by re-evaluating by changing the ranking of this attribute, a whole new group moved into a lower risk group. They were offered credit cards (note: no other company offered them credit cards or turned them down). This credit card company had a large group of new customers using their card. History has shown this group had a lower default rate than the average of the total population.
Unquote

3) Generate high level report for client to summarize his financial health (cash flow, expense profile, balance sheet, income statement) and usage needs / trends / volume / appetite

4) Price discovery for customized financial product and other products.
First degree price discrimination in capturing consumer surplus. 

5) Transaction history profiling
=>allow cross selling of related vendor non financial products,
=>fraud detection.
=>generally high quality data as they are transaction backed.
=>Use banking platform to display third party ads. 

6) Allows organizations to internally review and remove unnecessary and unprofitable products and services 

7) Passive client feedback according to his usage habits and spending profile. Detect likelihood of client retention and attrition to maximize lifetime value.Explored by gaming companies, which is dependent on consumer sentiment and employs a software-as-a-service model 

Quote
Phone companies will analyze all the call detail records, billing records and customer profile information. They will also look at customer who cancelled their service and they will use that data to identify patterns of behavior that preceded a cancellation. So this is a pretty obvious predictive model.

However, what we can also observe from the call detail records is who calls who, especially if both numbers are in the same phone network. We can see the caller, the recipient, the time of call and its duration. Over a long period, we can see occasional calls, frequent calls, regular calls and scheduled calls. From that we can figure out who is family, friend or business contact. As we look across everyone, we can work out the real social networks and the relative strengths of relationships.

Now adding these two insights together, being able to predict a cancellation of service and the social network of the customer, we also see that when a customer cancels, the likelihood that any of his social network will cancel also increases for a certain period of time. 

One more thing we say is that an INSIGHT IS THE CHANGE OF BEHAVIOR CAUSED BY A RECEIVING A PREVIOUSLY UNKNOWN PIECE OF INFORMATION. Very important - without the change in behavior, it's just another piece of information. 

So the results of this new information leads the phone company to reduce the number of cancellations across a person's network if they suspect a cancellation of service. 

So if you receive an offer for a discount or an upgrade, seemingly randomly, it is highly likely that someone in your network will be cancelling their service soon. 

The monetization part? Well, this is the future lifetime value of customers who were saved.
Unquote

Cons
1) GIGO. 

Not everything that counts can be counted. Not everything that can be counted counts.

Fools gold. Underlying cash flow comes not from data but from advertising revenue. Theoretical finance does not transit to actualized profits. Actual profit margin may be negligible. 

Quote
Data is not the new gold, but – at best - a new shovel to dig up gold. That difference matters: If there is no gold to begin with, i.e., no willingness to pay from anybody, no data in the world will help you make money. Therefore, with very few exceptions, data is not a business model in itself, but potentially an enhancer. If you forget this basic principle (and many companies do) your new business model is built on sand.
Unquote 

2) Cost of maintaining big data.
Server, storage and infrastructure costs.
Collection and data cleaning costs.
Hiring of high cost professionals

3) Accuracy and reliable of input and output
Conclusions and predictions are based on historical information. Cannot account for changing trends, tastes and preferences. Does it reflect market reality? 

4) Unproven monetisation benefits

Direct =>packaged to products / services to sell to data buyers. Are people actually willing to pay for them? 

Indirect =>analyse data to extract insights to make certain conclusions. Are these conclusions valuable enough to be sold? 

Initial thesis
The insights from big data is only as useful as the interpreter. Similar to quantitative finance, one can create elaborate charts and graphs to spot correlation and trends and look hip and sophisticated. However, correlation does not imply causation and it could be GIGO. 

Without qualitative understanding of underlying cause and effects, conclusions generated will not reflect market reality. To a man with a hammer everything looks like a nail. What gets measured gets managed. Line of best fit can be forcibly created and outliers can be conveniently ignored. 

Drawing lessons from another technology mania (cryptocurrency), Bitcoin is another technology innovation that gone south. The underlying technology may be incredible but ultimately the underlying monetisation model is uncertain. The monetization potential of big data, and the type of data collected may be industry specific, and the profit margin is widely uncertain. For companies using this as an unique selling point, I need to analyse it from a cash flow perspective and the supply demand agents for this unventured market. 

Links
https://www.globalbankingandfinance.com/monetising-big-data-in-retail-banks-starts-with-a-better-customer-experience/amp/

https://www.forbes.com/sites/forbestechcouncil/2017/05/03/key-challenges-for-monetizing-big-data-powered-ai-an-overview/amp/

https://www.quora.com/What-is-big-data-and-data-monetization