Gregory Maciag

Standards and BIG Data

McKinsey say “the effective use of Big Data could drive $325 billion incremental annual GDP in retail and manufacturing by 2020 while cutting $285 billion in costs across the government and health care sectors”.

That's “could drive...” Because these benefits aren't automatic. And they're downright impossible without core data governance disciplines, especially data analysis, data standards and metadata management. Of these three must-haves, data standards are the absolute central component.

Pitney Bowes

No Bad Data

The New York Times ran a piece airing concerns about the collection of big data and privacy. George Orwell gets a name check. More to the point is a comment from Microsoft's Craig Mundie: “There's no bad data, only bad uses of data.”

I know this is only a soundbite, but it's kind of chewy. Clearly, there is bad data. There's data that's wrong, and data that is wrongly tagged. There's also data that's badly used from structural causes. That is, rather than some evil genius cackling and using honest data for bad ends, the clean data has been incorrectly related to other data within a storage schema.

If you model your data incorrectly, you create structural misuse of data. And this is where the real dangers lie. The way to guard against structural misuse is to exploit established, trusted data standards wherever possible.

Equally, safeguarding privacy concerns will have to mean the incorporation of privacy markers in the data models for relevant applications. For example, mobile healthcare apps should be using personal data standards that include a privacy marker for age (or date of birth).

Big Data is Opening Doors

Data for Evidence

Thank you, Kristian J. Hammond. Writing in the Harvard Business Review blog, Hammond talks about the value of “evidence-based decision-making (aka Big Data)”. At last – someone with the sense to put the horse back in front of the cart.

You need evidence to run the business successfully, and data can be the basis of evidence. Hammond develops this theme by arguing we need to turn data into narrative; that is, build a story from our data.  I'd put it like this: If big data is the answer, what's the question? Data can help you take decisions, but it won't – on its own – tell you what decisions you need to be taking.

There's more and more data every day. But it doesn't automatically make us any smarter. If we're prepared to ask questions about the business, to construct theories and test them, then big data is a boon.The Value of Big Data Isn't the Data

We don't need entitlement reform - we need data reform!

So says Melissa Threadgill in the Boston Globe. She continues: “State agencies should adopt common data standards, preferably in concert with the federal government, to make data-sharing between agencies easier, and they should prioritize operating on platforms that can easily communicate.”

The lack of data discipline in government organizations is becoming a major issue. Taxpayers have experience of data foul-ups in their roles at work or as consumers. They know that paying a dead person welfare checks is a systems error traceable to poor data management. A generation ago, you might have been able to get away with blaming “computer error”. Not any more.

Ironically, the more public sector organizations highlight the potential benefits of big data, the more they expose themselves to criticism of their performance with small data. If you expect the public to follow and support big data projects, then you are acknowledging they are data savvy. And you can be held to account for inefficiencies stemming from poor data handling.

Welfare Fraud

Big Data Readiness

According to new research, executives feel they are not ready for the age of big data. “Nearly a third (29 percent) of executives gave their organizations a 'D' or 'F' in 'preparedness to manage the data deluge.' Another 31 percent rated preparedness with a grade of 'C'. Not having the right systems in place to gather the information they need is the biggest gripe of 38 percent of executives. Following close behind, 36 percent said their organizations weren't able to provide their business managers access to the information, and 29 percent said they were using systems that weren't designed to meet their industry's unique needs.”

I'm not sure how to disentangle “not having the right systems in place” from “using systems that weren't designed to meet their industry's unique needs”. Also, saying 36% aren't able to provide access to relevant information doesn't automatically mean the relevant information actually exists in the first place. Surveys like this can be difficult to interpret because they take the measure of a current concern and map it to a menu of potential deficiencies, each of which is likely to have widespread effects as well as being somewhat subjective.

To get real picky, show me an organization that is consistently and comprehensively using systems designed to meet the industry's unique needs. The industry's unique needs continually change. Also, the whole point of “the data deluge” is that it's meant to deliver surprises. Big data is going to show us industry needs beyond our expectations.

A positive approach to this question would be to pitch a menu of capabilities that executives could choose from in building their active response to the data deluge. These could include, for example: The ability to gather any and all types of information relating to objects and activities in our industry. And: Using systems that were designed around industry standards for data and standard process models.

Being ready for big data means being ready to handle any item or mass of data as meaningful information. So, your systems and practices need to be based on open standards agreed for your industry – standards whose evolution is actively managed on behalf of the community using them.  Failing

Context for Standards

“For insurers, the goal for big data is to contextualise it – take the data, give it a context and delete the rest,” says CSC's Gary Jackson.  There's something wrong here. Isn't the big hope of big data that it will suggest new contexts? I'd agree that big data is best managed via standard data definitions and relationships, so if that's Jackson's meaning of “context” then I'm happy.

To judge from his subsequent comments about the expense of storing, analyzing and managing data, Jackson and I are on the same page. Keep the information, delete the noise. You can tell information from noise by matching to a standard.  By the way, ACORD is doing plenty with forward thinking forums in Australia.    Insurance Firms Lag

Big Data for Healthcare

The Institute of Medicine is calling for a “Knowledge Network of Disease”. By bringing together diverse data on health, new discoveries can be made. Healthcare can get smarter, faster, and more evenly distributed.

Reporting on developments in The Atlantic, David A. Shaywitz highlights the tension facing all data aggregators: whether it's possible to create a single, coherent and correct data model, or whether it's better to use a more ad hoc approach. He quotes system biologist Eric Schadt: “The grand data model of everything in biology has been tried over and over again and ultimately it comes crashing down because once you lock in a model, you constrain the types of questions you can ask of the data. The information half life is short, and we know and understand so little about what the data can actually tell us that we don't really understand the questions that ultimately will be the most useful.”

How do we build a knowledge structure when we don't know what we want to know? It's a problem we face in business too.   However, the problem can be attacked in an incremental way. There are levels of detail at which healthcare professionals will agree. They'll agree that there's such a thing as a patient, for example. They are also likely to agree on the major biological systems and organs. They will also probably agree on a core set of signs and symptoms.

There will be a point at which interested parties can no longer agree on definitions of items or relationships. But that doesn't mean the modeling approach should be abandoned. We have to tolerate fuzziness and ambiguity at the depths and fringes of our knowledge structures. That's how we learn.

As healthcare data sets become larger, researchers will use analytics to discover new types and constellations of facts. These discoveries will sharpen the resolution of the outlying fuzzy areas. So the model will become more detailed and more effective in response to its use in the field.

The idea that standards need to be all-or-nothing is one of the myths of information management. Standards exist to facilitate  action, not constrain it. You don't have to have a grand data model. But you need a model.   Instead of thinking that a model or standard “locks in” users, we need to manage standards as evolving tools that learn from users as well as supporting them. Standards must lock step with the communities they serve. Database for Medical Knowledge