CIOs and CTOs like it; Enterprises like it; Startups like it; Investors like it; Researchers like it; Users like it. They invest in it, promote it, deliver it, research in it, refine it, and use it [ex. 1-12]. It is our time’s most profound machine to explore with; to explore everything. Now, why should the biometrics market and industry take a closer look at it? Read on and Welcome to The Era of Big Data and Metrics!
Hype, Hope, or History goes Mainstream
The hype aside, it has been there for years: Big Data and its use cases. For ages, the diversity and scale, not to mention the necessity, of its applications and advantages have been tremendous: Aerospace industry has used it to monitor and protect spacecrafts and crews; Formula One teams have applied it to optimize car performance and racing strategies; Rotating machine industry has implemented it to predict and prevent machine part failures. Limitless examples of Big Data can be given in multiple sectors, markets, and fields with totally different kinds of business objectives.
It has become increasingly obvious that applications of Big Data are expanding immensely: from mining the social web to maximizing efficiency of green-tech’s smart-grids; from ocean observation to The Human Brain Project and to human genome projects; from personal fitness and health monitoring to Hans Rosling’s moving bubbles that visualize global trends… the list can go on and on. Primary drivers for this transformation to mainstream, according to IDC’s annual study, are decreasing cost in storage, increasing connectivity (e.g. cloud), sensors, data capture and analysis tools. And then, there is the vastness and variety of data: 1.8 zettabytes in 2011 and growing exponentially; see IDC’s report “2011 Digital Universe Study: Extracting Value from Chaos” or see its infographic. Last but not least, what I would add to IDC’s analysis is birth of new metrics on daily-basis and in everywhere.
Big Data in the Big Picture…
First of all, allow me put forward my hypothesis: I think Big Data may very well be our single most meaningful IT activity and that it is maturing alongside the human evolution set to approach our most fundamental questions; in particular, the biggest of them all: “The Mystery of Consciousness” (via e.g. modeling the human brain; metrification of consciousness in systems). In a philosophical sense, everything is Big Data because everything is information/signal processing. In a technical sense, IDC’s “Digital Universe study” defines it clearly (which is by the way a good candidate for consensus-building across fields):
“Big Data Is Not the Created Content, nor Is It Even Its Consumption — It Is the Analysis of All the Data Surrounding or Swirling Around It.”
…and in Biometrics
Thus, let’s agree on the following: biometric features, templates, or even any kind of transactional data (whether synthetically generated or as result of any application usage) by themselves are not Big Data.
In the broadest sense, Big Data in Biometrics is the collection and analysis of biometrics-related data of many sorts for completely diverse purposes and properties such as reliability, availability, maintainability, security, usability, performance, prediction, prevention, detection, visualization, and so forth and so on.
In other words, in any biometric-based application, Big Data is about extracting streams from one or more data points and making sense of those data streams; e.g. drive valuable metrics to support business goals. Examples of biometrics-related data are general or vendor-specific system-events or diverse types of performance and environmental metrics/data (e.g. a verification reject, transaction time, temperature, noise) and many more.
Value-Add Expressed in ROI
Biometric devices, systems, and applications are no exception for Big Data. Quite the contrary, they are perfect candidates to use it. Here is why:
- Their fundamental properties get affected by real-world factors such as 1) Human factors, 2) External environmental conditions, 3) System related issues [ref. 1]
- They can generate real-world transactional data (i.e. events, data, metrics) that can be fed into Big Data tools
- In return, they need their fundamental properties improved, optimized, preserved, etc (e.g. to have their reliability, availability, maintainability, or performance maximized)
Now, in order to motivate and engage your customers or your team to understand value-adds of Big Data in your business case you should be able and confident to present to them tangible numbers via Return on Investment (ROI) calculations. See for example this simple yet advanced ROI calculator that my company has released. It looks at different types of operational costs to then produce ROI numbers by implementing a Big Data tool and Best Practices in order to reduce those costs.
I will give you a few simple and practical examples with regards to the properties of Big-Data-in-Biometrics that I mentioned above. For these examples, let’s assume this scenario: you either have just a few or tens-of-thousands biometric capture devices spread over several geographically remote sites and you have the data collection and analysis tool needed for the job.
- Availability and Visualization: Your goal is to minimize downtime. The metric requirement you want to meet is Operational Availability (Ao) set in your SLA contract [ref. 2]. You also want your Operation and Support Center to easily and instantly see what devices are down. You have your devices to send regular heartbeats to the Big Data tool and it elegantly visualizes availability of your devices on a smartphone, a monitor, a big screen, or all of them.
- Maintainability: Your goal is to minimize service needs and maintenance costs. The metric requirement you want to meet when resolving issues is Maximum-Time-to-Recovery. You have the vendor-claimed device reliability metric Mean-Time-To-Failure (MTTF). You track operational hours of all your devices by the Big Data tool and set it to notify you before the MTTF has reached so that you can optimize your preventive or schedule maintenance. This way, you not only can better choose and implement your maintenance strategy (e.g. skip the costly corrective/reactive maintenance type) but you also can minimize potential downtime.
- Reliability: Your goal is to measure operational reliability of your devices. The vendor-claimed metric (and stated in SLA) you want to verify is either Mean-Time-Between-Failure (MTBF) or Mean-Time-To-Failure (MTTF). You are a group of customers using this type of device and you all need this verification for your SLA contracts. You have your devices to send regular heartbeats to the Big Data tool that registers device failures and calculates the metric.
- Performance and Usability: Your goal is to measure and compare operational performance/usability between your devices distributed over different locations and user groups. Your metrics are reject rate, failure-to-acquire rate, and transaction time and you have set their corresponding baseline levels using obtained/recommended by the vendors. Your devices stream their transactional data to the Big Data tool. You can 1) view the metrics in real-time 2) have alerts sent to you when thresholds are exceeded 3) get on-demand or periodic device comparison reports.
- Prediction and Prevention: Your goal is to predict and prevent template aging (or an escalating false reject rate). The Big Data tool is equipped by a specialized prediction algorithm that uses diverse streams of transactional data to pinpoint the issue as an estimate and formatted in time or nr. of transactions.
- Detection and Security: Your goal is to detect a security breach type (e.g. imposter attacks). You have identified a number of potential vulnerability points that you monitor by your Big Data tool. Configurable alerting rules and messages can be either sent to the system admin or directly to the user.
- Product Improvement and After-Sales Offerings: since 2003 we have been promoting [also ref. 2] the concepts of making use of real-world feedback (data and stats) to help R&D teams making better products and also to enhance after-sales support. Now in 2012, I am delighted to quote McKinsey’s recent report:
Finally, big data can be used to improve the development of the next generation of products and services. For instance, manufacturers are using data obtained from sensors embedded in products to create innovative after-sales service offerings such as proactive maintenance (preventive measures that take place before a failure occurs or is even noticed).
Market Drivers and Challenges
Although not a complete list (everyone can contribute by leaving comments), here is how I divide it:
Customer demands: Let’s coin this term: There’s A Metric for That. For, if there is a metric for that then not only there is a customer need for that but also there is an industry-wide language for that to be communicated and used between the market and industry (via e.g. Service Level Agreements, SLAs).
Rise of the Big Data tools:
- Product category definition: Here is how I define it: Big Data tools for biometrics collect and analyze varying types of biometrics-related data to help to achieve diverse business objectives.
- Standardization: There are parts such as terminology and data capture interface that when standardized simplify for everyone.
- Diversification: tools from different vendors can coexist, be utilized (even in complementary roles in certain deployments), and constantly evolved.
Now they’re talking, biometric devices go TCP/IP…: A walk around the recent security show ISC West 2012 in Las Vegas clearly demonstrated that majority of biometric products for physical access control and time and attendance applications are now network-based (see my pictorial coverage).
…and, there is a new standard for connectivity: Speaking of increasing connectivity of biometric sensors, if disseminated well, we are going to see lots of WS-BD enabled devices in the marketplace. The NIST WS-BD Team (Web Services – Biometric Devices) recently published NIST Special Publication 500-288 – Specification for WS-Biometric Devices (see: http://bws.nist.gov).
Staffing: Personnel with Big Data and deep analytic know-how will be needed. McKinsey’s report forecasts:
By 2018, the United States alone could face a shortage of 140,000 to 190,000 people with deep analytical skills as well as 1.5 million managers and analysts with the know-how to use the analysis of big data to make effective decisions.
‘Champions of Big Data’, dedicated teams within organizations: In the age of Big Data and Metrics every organization should build a Big Data team to champion the culture of “Think in Metrics”. As ComputerWorld article indicates, role of CIOs and IT managers will advance more towards driving business rather than only keeping infrastructure up and running. I have, in my reference portal for metrics WhatMetric.com, proposed Chief Metrics Officer, CMO, as a key function for any organization and enterprise.
Privacy, security, and more: Big Data like any other tool is neutral and can be misused, altered, hacked, illegally accessed and shared and so forth and so on. Regulations and policies have to be in place for successful implementations.
What to Envision
Applications of Big Data in Biometrics like the few I described above (whether or not new or already deployed in some projects) will grow in flavors and numbers. Beyond what already exist, here is what I expect to discover in near-term:
- 5 published case studies/use cases of Big Data for Biometrics
- 5 competing/complementing vendors with Big Data know-how and commercial-off-the-shelf (COTS) tools
- 5 new useful metrics for the Operations and Maintenance Center
Can the question “How ‘Big Data’ impacts biometrics market and industry” be answered in established metrics? most certainly, and those well-known metrics should be compiled by their specialists, the market and industry analysis companies. I sincerely hope and look forward analysts soon see the logics to publish the reports.
What do you expect to see? Your thoughts and comments are most welcome. By the way, Happy World Metrics Day (WMD), its June 16 every year [via ref. 5].
- Best Practices in Biometrics Performance Monitoring (BPM) Programs, Optimum Biometric Labs
- Reliability, Availability and Maintainability (RAM) in Biometric Applications – Delivering Quality of Service that customer wants, Optimum Biometric Labs
- IDC’s annual study “2011 Digital Universe Study: Extracting Value from Chaos”
- WhatMetric.com, a reference portal for metrics, a hobby project by Babak Goudarzi Pour
- Are Metrics Blinding Our Perception?, The New York Times, By Anand Giridharadas
- ‘Big data’ prep: 5 things IT should do now, Computer World, by Beth Stackpole
- Big data: The next frontier for innovation, competition, and productivity, McKinsey Global Institute
- Taking Small Steps to Big Data, The Wall Street Journal, CIO Journal, by Michael Hickins
- Tech’s Next Billion Dollar IPOs, Forbes, by Dave Feinleib
- Predictive Startup Recorded Future Raises $12M From Balderton And Google Ventures, TechCrunch.com, by Anthony Ha
- Microsoft strikes deal with 24/7, promises to ‘redefine’ customer service, Engadget.com, By Donald Melanson
- CIOs Should Know: H-P Investing in Big Data, The Wall Street Journal, CIO Journal, by Clint Boulton
- Apptegic Uses Big Data Analysis To Help Companies Retain And Upsell Their Customers, TechCrunch.com, by Ryan Lawler
- America’s CTO Todd Park is Giving Away Really Big Data, TechCrunch.com, by Semil Shah
- This Man Could Rule the World, Popular Science, by Gregory Mone
- A “Complex” Theory of Consciousness, Scientific American, by Christof Koch
- Ocean Observatories Initiative (@oceanobserv)
- Gapminder, a fact-based worldview project for sustainable global development, Hans Rosling
- The Human Brain Project, A Countdown to a Digital Simulation of Every Last Neuron in the Human Brain, Scientific American, by Henry Markram
- The Brain, The Connections May Be the Key, Discover Magazine, by Carl Zimmer
Want to broaden your reading on Big Data, here are some recent articles: