Information Playground

EMC Fellow

The Four Principles of Data Value

Over the last few weeks I've been summarizing some of the preliminary findings from the Data Valuation research being conducted by Dr. Jim Short of UC San Diego. In particular I've written three blog posts describing:

Data value can be ephemeral. Instead of holding on to data and attempting to mine it over time (e.g. a "refinery" model), some corporations extract value from data that is temporal or fleeting (e.g. Snapchat data).
Data value can be eternal. Examples of data having eternal value would be genealogical records, or documents related to the history of a nation (or a presidency, as in the case of the John F Kennedy LIbrary and Museum).
Data value can be contextual. It may represent different value to different organizations within a business, such as manufacturing, product line, security and IT team, etc.

In this post I'd like to highlight Jim's final point: data value can be ever-changing. I understood this to mean ever-fluctuating. Insurance companies depreciate automobile value over time, and in many of the traditional "data life-cycle management" models there was an assumption that data loses value as it ages. While this may be true for some percentage of the time, Dr. Short's research is searching to uncover use cases where the value of specific corporate data assets rises as well as falls.

My EMC colleague David Dietrich sent me an interesting Wall Street Journal article the other day that emphasized how the industry's struggle with data valuation. The title of the article was "The Big Mystery: What's Big Data Worth?" The sub-title was also telling: A Lack of Standards for Valuing Information Confounds Accountants, Economists. The beginning of the article paints a good picture:

The problem is that no one really knows what all that information is worth. Data isn’t a physical asset like a factory or cash, and there aren’t any official guidelines for assessing its value.

“It’s flummoxing that companies have better accounting for their office furniture than their information assets,” said Douglas Laney,an analyst at technology research and consulting firm Gartner Inc.“You can’t manage what you don’t measure.”

After listening to Dr. Short and his description of the Four Principles of Data Value, it's clear to me why the industry is struggling with how to approach this problem. Some data is temporal and some is eternal, and its value is not only fluctuating but it is fluctuating conditionally based on who is asking (e.g. data value is contextual).

The article certainly touches on the difficulties of temporal versus external as it refers to the "shelf-life" of data:

"Companies also would have to estimate the shelf-life of their data, figure out its future worth and track and report any changes in its value. Crunching those numbers would be relatively easy for a physical asset like a factory. But in the squishy world of intangibles, there’s little precedent for such calculations."

However, as I read the article it became clear that there are existing data valuation processes, including:

Intellectual property
Selling consumer data
Search algorithms
Data acquired during mergers

One area that the article fails to mention is a trend that Dr. Short has noticed occurring outside of the enterprise: data valuation processes in the insurance industry. As part of the research I've begun exploring the intersection of insurance industry valuation processes and IT infrastructure.

These processes take into account extremely ephemeral data (e.g. packet-level insurance on in-flight data via companies like CloudCover) as well as more traditional corporate data (e.g. data subject to breach).

In future posts I plan on diving into more of Dr. Short's research and highlight several use cases of note that hint at new standards and methods in the data valuation area.

Steve

EMC Fellow

Contextual Data Value

I've been stepping through some of the "Principles of Data Value" shared by Dr. Jim Short of UC San Diego during his recent visit to an internal EMC Technology Summit. In recent posts I've shared that

Some data value can be ephemeral. Instead of holding on to data and attempting to mine it over time (e.g. a "refinery" model), some corporations extract value from data that is temporal or fleeting (e.g. Snapchat data).
Some data value can be eternal. Examples of data having eternal value would be genealogical records, or documents related to the history of a nation (or a presidency, as in the case of the John F Kennedy LIbrary and Museum).

A third principle of data value is that it is contextual.

In other words, data value might be variable based on who is asking. After hearing Dr. Short describe this principle I likened it to a set of business units within a larger corporation that assign different amounts of value to the same data set. In addition, these business units are often grouped into two levels: the business level (the "demand" side of requesting IT resources) and the infrastructure level (the "supply side" of providing IT resources).

For example, consider a data set that contains customer order information. This data set represents different amounts of value depending on the dependencies that each organization has with the data set.

Listed below are several organizations that have a different "value" view, such as:

Manufacturing: this organization has a critical reliance on these purchase orders, and the growth in size of this data set is of particular interest.
Chief Security Officer: the names of these customers, should they be compromised, represent a potential risk in the form of damages and brand value.
IT infrastructure team: this data set will have high availability and reliability characteristics. Smooth operation of the infrastructure increases revenue, failures decrease revenue.
The Line of Business for the product being sold has strong affinity to the data set.
etc.

Given this reality, it is difficult to go to the IT team and ask "what's the business value of the data set"? Specific value is defined differently by key stakeholders in each organization; creating an overall enterprise value of the data set must take each of these stakeholders' views into account.

This is one of the key areas of the research that Dr. Short is conducting. Specific questions of concern are:

Is data value being calculated at all by individual business units?
What is the process whereby this value is calculated?
How can an aggregate value be calculated? Should a value be calculated on the supply side as well as the demand side?
How does the value get communicated across organizations?
How are fluctuations handled over time?

The principle of contextual data value represents one of the meatier areas of the research being conducted this year. I plan on posting several follow-up posts on this topic going forward.

Steve

EMC Fellow

Eternal Data Value

In my last post I introduced the concept of Ephemeral Data Value. The term, as defined by UC San Diego researcher Jim Short, was recently described to the EMC Technical Community as he highlighted the company known as Snapchat. Dr. Short pointed out that the ephemeral, transient nature of data in the Snapchat environment (it disappears within 1-10 seconds) does not mean that the data is without economic value. The reported Snapchat ad revenue model would be proof of this thesis.

On the other end of the spectrum from short-lived data is data with a tremendously long shelf-life. This leads to another principle of data value: some data has eternal value. Dr. Short has begun to interview several different companies in this regard. This will allow the industry to gain more insight into the valuation processes of companies with business models that require longer-term retention of data.

One company involved in the study is Bit Mountain. This company is charged with forever holding the Mormon Genealogical repository. Several years ago the NORAD military ICBM site was purchased for the purpose of maintaining these genealogical records, and they've now become a case study in the Data Value research.

When asked to represent the value of data under their care it became difficult for them to express. The records to some degree represent a portion of the history of the world.

When asked how long they retain the data the answer is indefinitely. A key principle is to never to lose anything ever.

Another variant to eternal data value is generational data value. This principle can be found in some financial institutions. For example, Dr. Short is working with mutual fund companies that have policies to retain data for multiple generations that are often measured in the 100-150 year time frame. Most companies in this space have similar (or longer) retention periods for valuable data.

The eternal data value topic is tangential to an effort launched over ten years ago by SNIA: the 100-year archive Task Force. This task force was concerned with data aging and application/infrastructure compatibility that would continually enable access and availability in the long term. For more information about survey results from the 100-year initiative, please refer to the 100 Year Archive requirements survey.

Dr. Short's work is highly complementary to this survey. In addition to properly planning for infrastructure and application compatibility in the long term, how can the value of such data sets be continually evaluated over time as well? His research will hopefully begin to shed some light on possible approaches.

Steve

EMC Fellow