Feedback loops: The confluence of DevOps and big data
Across the course of computing, the feedback we need from digital data has always been at our fingertips. Yet for the longest time, only those at the top of the analysis value pyramid—exemplified by Wall Street and the intelligence community—could afford to get to that data and exploit it.
Now, as costs and complexity have dropped dramatically, a broad democratization of digital information has arrived. For businesses, the information received from transactions and interactions provides a variety of new opportunities to learn about customer needs and habits and make educated guesses about what will delight them next.
A DevOps approach to software development is one example: software development organizations use these techniques to quickly release apps and gather feedback on new features in the latest version. For quantitative analysts and data scientists taking a big data approach, the feedback available through log files and click-stream monitoring offers a similar, if not more granular, mode for understanding user behavior.
By hosting the high-caliber analytics engines in the cloud, even small companies and individuals can obtain astonishing insights to aid their weighty decisions. Near real-time feedback loops are transforming slow, batch assessments of essential data into live streaming of reaction and prediction services. Affordable automation of actionable analysis accompanies the democratization trend.
The emergence of feedback loops
One of the first uses of feedback loops occurred in the IT data center. The prodigious volume of log and management data generated by IT systems is now being mined and fed back to the development and new requirements teams. Whether we are pursuing big data analytics or a DevOps approach to application deployment and customer input, this is an opportunity to respond to consumer behavior and deliver competitive differentiation more quickly than we've ever seen in the past.
Today, the more businesses and consumers share, the more we all benefit. The rich, constant data flow from our queries, actions, locations, sensors, and devices easily gets back to big data analytics capabilities. The resulting feedback loops are just now being assembled and greased. But they apply to and from our customers, employees, partners, suppliers, prospects, and communities. The more we use them, the more valuable they grow.
Using a watch, smartphone, tablet or PC, we gain an immediate and ubiquitous ability to interact with a rich and growing set of cloud services, each endowed with its own big data analytics. These services benefit from our collective data sharing and vice versa. Data-driven business and consumer apps made available to us anytime and anywhere are rapidly turning us all into budding data scientists.
Democracy is at hand
The emerging big data feedback loop mechanism in the era of increasingly integrated mobile, Internet of Things, DevOps, and cloud opportunities is only now being defined, still open for innovation and experimentation. At its essence, the latest technology compresses the ability for trial and error testing on just about anything. Refinement cycles go near-instant for mining data-driven truth. Most importantly, data-driven learning extends rapidly from experience to analysis and back to new implementation, for nearly any business problem. Users and machines gain empirical insights and make accurate adjustments. Repeat. Affordably. At enormous scale.
It's still far from full democratization of such analysis, but it's essential that businesses prepare to send and receive real-time data and analysis on just about anything. Feedback loops galore will fuel our data-driven decisions and offer entirely new levels of precision and rapid adjustment to more types of events, variables, and needs. The bold and curious will lead the creation of value around such new feedback-loop empowered analytics.
Businesses are on the cusp of receiving instant analysis of anything worth analyzing, and we're moving swiftly toward the prospect of wider prescriptive analytics, in which the best course of future actions are displayed as a near certainty. Many variables toward the best solution can be tested with little or no pain. And this really does change everything.
Five good examples
To gain a glimpse of what to expect, consider how several businesses are already using rapid feedback loops to gain newfound productivity benefits and develop better decision making:
- adMarketplace in New York, a leading advertising platform for search intent, provides consumer-focused advertisers instant insight into where they should project ads. The idea is to use a big data platform to better match ads with the right user intent, based on data from the users' actions. adMarketplace directly streams all user click data into an analytics engine, which provides predictive analytics on top of the user data. They examine some 15 dimensions, like geospatial analysis, to fit an ad to the consumer to meet their expected needs. The whole process occurs in milliseconds, with half a billion requests daily.
- Unum Group in Portland, Maine has been building a data-rich DevOps continuum and is further exploring the benefits of a better process around cloud-assisted applications development and deployment. At Unum, the infrastructure and application teams yearned for visibility into each other's spaces. Leveraging insight into operations, developers gained visibility into what's happening in the server environment. Now, Unum establishes SLAs in development and then monitors how the apps perform in production—all involved learn as they go.
- Nimble Storage in San Jose, CA uses a tight feedback loop leveraging big data and the cloud, allowing them to provide optimized and on-demand data-storage performance. High-performing, cost-effective big-data processing helps Nimble make the best use of dynamic storage resources by taking in all relevant storage activities data, analyzing the data, and then making the best real-time choices for dynamic hybrid storage optimization. Receiving anywhere from 10,000 to 100,000 data points per minute from each storage array, Nimble creates a feedback loop that allows dynamic cache sizing. Nimble then discovers how much cache each customer needs based on an analysis of their real workloads. Using this method and scale, Nimble determines in near real-time the distribution of workloads—how people truly access their storage. This generates a better understanding of how physical and flash storage work together in the real world. To aid their business, Nimble can do capacity forecasting using predictive analytics to know when customers' storage administrators are going to need to purchase something and then provide it first.
- Auto racing powerhouse NASCAR in Daytona Beach, FL built its Fan and Media Engagement Center using big-data analysis. The result is that NASCAR rapidly adjusts services and responses to keep best connected to those fans across all media and social networks, about 18 million impressions in the first year. By accessing all information and data generated constantly from social media, news media, broadcast media, and traditional media news sites, and then making the analysis instantly available as easy-to-consume and relate via visualizations, NASCAR gains a single pane of glass on the overall conversation about its brand and activities. NASCAR uses it on race day to tweak how they present their live media, and they use it longer term to create new services and to enter new markets at low risk.
- Visible Measures Corp. in Boston delivers impactful metrics on online video use and engagement patterns by determining how much people actually watch a video and how they interact with it while it's playing. Video is among the fastest growing components of the Internet. Video use measurements are much more granular and precise than traditional television. With its massive database of video, Visible Measures provides insights to advertisers on how well their campaigns perform. Using the feedback loop from the data, brands inject other brief video segments across thousands of online publishers. Visible Measures can then measure that engagement, as well as the brand-lift, thereby building trust of the viewer and ultimately gaining better customer allegiance and sales for the brands and advertisers.
Now, more organizations can bolster their business decisions by exploring data in real time—affordably. Like these innovative users described here, you can begin taking the steps now to make decisions better and faster than has ever been possible. After all, hope, guesses, and ill-informed extrapolations have never been particularly good business models.
Not long ago, you could only pine and wait for better business insights. Now, you and your business can define what you want and get it affordably, repeatedly, and rapidly, while remaining on the vanguard of constant, iterative improvement.