Bullpen Bulletins

Designing for Humans or Robots

Imagine you are driving in an Alexa-enabled car. While traveling, you issue verbal commands to set your home temperature to 62 degrees and preheat your oven to 350 degrees. While you feel your instructions are clear, this can be a massive challenge from a design standpoint. Hopefully, the system understands and correctly translates your voice commands, considering no one wants a 350-degree home and an oven baking at a mere 62 degrees.

In a separate scenario, let’s say you use Google to send out my robot vacuum to clean your floors. The vacuum could be activated by voice or by text. Using the correct parameters of commands becomes essential. If you ask Google to command my vacuum, it doesn’t respond. However, if you tell Google to do the same function, my vacuum is activated, and my floors are swept clean.

In both cases, we design the user experience based on our human perceptions. But computers, robots, and logic systems might not respond how we want if we only design for human interaction, especially if one system is talking to another system.

We often cite Isacc Asimov’s Three Laws of Robotics which command a robot to follow our instructions but prevent them from harming humans or itself. However, these are not always feasible. Think of something like a self-driving car. What would happen if that driverless vehicle were about to crash into something or someone? It’s likely to harm the driver or harm the other party. Does this suggest that we need more from a design standpoint? Should software and hardware companies ensure that voice and visual interactions are more human-like? How might that change when robots talk amongst themselves?

While the non-human interactions can be with computer networks, artificial intelligence (AI) systems, algorithms, and robot systems, in this blog, I will collectively refer to all these as robots for brevity.

Every day, we are using the interaction with these robots to complete sophisticated tasks. To continually improve those interactions, we demand that solutions designers provide a user experience and user interface that incorporates our human ways of thinking and performing. But as more machine learning (ML) and deep learning algorithms are used for these interactions, designers must consider how these networks, systems, and robots will want to interact with each other. And careful consideration should be given to how human developers might keep a hand on the wheel for command and control.

Should we design only human-oriented interaction designs for humans to maintain control? If AL/AI interfaces and robot application programming interfaces (APIs) require a different type of experience and ethical design, are humans the ones to do it? Could robots design their own interfaces?

Read the rest of the article here: https://www.extremenetworks.com/extreme-networks-blog/designing-for-humans-or-robots/

Four Ways to Add Analytics Value to Your Business

All data tells a story. The challenge is to find that story by analyzing business data in dozens of ways, using spreadsheets, applications, desktop tools, long-standing data warehouses, and business intelligence software. That means you are probably managing a variety of different dimensions—specific requirements for accessing various data sources, complexities around understanding visualizations, and of course, the associated costs. At the same time, everyone wants more innovation faster—all without sacrificing correct and consistent results.

Companies just like yours seek solutions that offer easy access to any data, enabling you to make the best use of your entire data story, no matter where it is. What’s even more important is having a variety of options for intelligent analysis that isn't overwhelming to deploy and manage. The result creates an effective way to engage more people in analysis and extend your organization's expertise. After all, you can have all the bells and whistles a robust platform should offer—but if only a select few can figure out how to use them, what are you really gaining?

So, the question you face is how to find and tell that story. How will you get the flexibility you need together with the structure to drive accuracy, and the speed to disrupt your competitors? Here are four values that can apply to your business analytics strategy.

1) Visualize Your Business

Every company has a story to tell with its data. There are customer transactions, marketing metrics, collaboration with partners and suppliers, innovations in business processes, and the enablement of growth and personnel efficiencies. And while the data is abundant, useful insights are sometimes difficult to develop.

Data visualization software makes these complex ideas engaging, meaningful, and easy to understand. With just a few clicks, business managers can add, analyze, and share new insights. Savvy executives use this technology to improve their strategies with facts and get action from their ideas. The important thing is that data visualization can help us see information differently—and by doing so, we make better decisions in ways that spreadsheets or traditional static business intelligence cannot deliver.

2) Make 'What-If' Simple

One of the benefits of modern business analytics software is that more people can gain a deeper understanding of the problems they are trying to solve without having to be data scientists.

Imagine if workers at a manufacturing plant could know when and why a machine would fail before it happened. How much would that impact production? Data modeling and predictive analytics programs can help answer these types of questions. Previously, this type of insight could only be delivered by data scientists working with sophisticated algorithms. Now, business managers with no data science experience can easily look at data, ask "what if" questions, identify hidden relationships, and convey these insights to the rest of the business.

Simplifying these forecasting and modeling scenarios enables business users to quickly model complex business questions. Organizations can then define a dimensional view of their business—and provide business users with new levels of self-service to access, navigate, and gain actionable insight into critical business issues. All of this is enabled through cloud-based analytics platforms.

3) Choose Flexibility

Analytics implementations have matured in uneven ways, making it clear that there's no single formula for success, especially with hybrid-cloud and on-premises deployments. In some cases, a turnkey cloud service is a right answer; in other cases, you may need a cloud environment where you can turn all the knobs and cranks like you did on-premises. Either allows you to lift and shift back and forth between on-premises and the cloud. The result is a sense of balance, with organizations working smarter—not harder—to bring these elements together.

The growing tension between governance, risk, and compliance is also relaxed by this flexibility to choose. Financial, operational, and regulatory policies and mandates overwhelm your ability to manage the associated risks. Challenges are compounded by a lack of enterprise-wide visibility. What could go wrong? Have you ever asked yourself what it would take to spend less time worrying about risks so you could focus on pursuing growth opportunities?

4) Add Augmented Analytics, AI, and ML

What if businesses gained the power to identify your needs and fulfill them long before you even thought of them?

Sure, websites already track our browsing patterns and make suggestions, but what we're talking about here is something much deeper. Business intelligence is evolving into augmented analytics.

Augmented analytics, along with the power of artificial intelligence (AI) and machine learning (ML), address how we interact with information and transform how we work and live. For example, businesses are preventing and detecting credit card fraud, automating supply chains through checks and balances, optimizing supply chain finance based on market swings, and boosting data center efficiency. With the help of artificial intelligence and machine learning, these abilities are now at the intersection of people's judgment, and machine automation and augmented analytics make them possible.

What could be possible in your organization if insights came to you when you needed them most? What might happen if half the content for your next operations review was generated by machines? How prepared would you be for your next meeting if you were notified of changes based on your location?

Augmented analytics applications are a new category of continuously adapting, self-learning applications powered by enterprise data from a variety of data sources, including transactional business apps (such as customer experience, enterprise resource planning, supply chain management, and human resources). These are solutions that operate without human bias and deliver a very high degree of confidence—on a very large scale.

The Value of Data-Driven Innovation

Take a moment to think about all the data generated during a single business day. New customer profiles. Sales logs. Marketing touches. Payroll allotments. Service records. Inventory rolls. No doubt that in a single 24-hour period, a business retains specific details of a massive number of transactions. But what does it all mean?

Simpler big data discovery tools will let business analysts shop for datasets in enterprise Hadoop clusters, reshape them into new mashup combinations, and even analyze them with exploratory machine-learning techniques. Extending this kind of exploration to a broader audience will improve self-service access to large data repositories and provide richer hypotheses and experiments that drive the next level of innovation.

Many successful companies have already adopted a data-driven approach to innovation. They connect their data, automate their processes, understand the business drivers, align corporate finance planning with operations planning, and build their team's analytics expertise.

Easing Massive Supply Chain Disruption with Analytics

When world events disrupt business supply chains, smart companies weather the storm with flexibility, contingency planning, and dual sourcing. Even smarter companies adopt data analytics in their best practices to spot opportunities amid simultaneous disruptions in the supply network, spikes in consumer demand, and interruptions in transportation.

Supply chain disruption in the market is not new. Silk and spice traders in the 17th Century faced extreme challenges in their efforts to maintain a consistent supply chain. Treacherous weather, greedy pirates, regional conflicts, trade disputes, and the like continually disrupted trade.

In the modern era, maintaining an optimal supply chain requires preparing for even more unknown risks and the trade-offs of using data from multiple sources to manage your plans. Unfortunately, fewer than half (47 percent) of companies surveyed in a Ventana Research Next-Generation Business Planning Benchmark survey said their supply chain plans are accurate. Even more concerning is that a majority (53 percent) of respondents say they have either a limited or no ability at all to understand the trade-offs.

Analytics can dramatically improve visibility across the end-to-end supply chain, increase understanding of inventory status and location, and provide predictive analysis that enables companies to withstand future shock events. By drawing in data from multiple sources across the supply chain network, companies enable their decision-makers with timely and complete insight. They can analyze supplier status, manufacturing capacity, inventory, supply/demand balance, and logistics datasets. The idea is to cover the entire supply chain, providing insightful analytics for quick visibility into changes in demand so that you and your network partners can adjust production and fulfillment.

Three Areas Analytics Supports Supply Chain

Analytics helps supply chain managers evaluate the challenges and monitor the benefits of real-time insights to optimize actions such as:

• Procurement
• Inventory
• Logistics

Modern procurement starts with standardizing, streamlining, and automating the overall procurement process. Analytics provides visibility into the source-to-pay process. For example, you might need visibility into the supply of direct and indirect material and the resulting impact on order fulfillment, revenue, and cash flow. That requires cross-functionality, or the ability to share data seamlessly between departments. Your finance division should be able to assess the cost of reducing contractual procurement due to a rapid decrease in orders or in demand. And when supply drops and orders cannot be filled, analytics can minimize the effect of disrupted supply on financials.

Maintaining inventories is critical and involves comprehensive visibility and management of material, inventory status and turns, and product costs across the extended supply chain. By using analytics alerts to rapidly identify tasks requiring an immediate attention, managers can rebalance inventory by location due to sudden demand spikes and stockouts. Improving the stocking and handling of different types of materials is enabled by using analytics to monitor and manage material status and trends. Managers who need to quickly understand the cost impact of expediting shipments to meet demand or the costs of holding unexpectedly slow demand can use analytics to combine material status with finance.

Of course, products don't just materialize from one place to another. Shippers and buyers need immediate visibility into available shipment options and associated costs. Given industry volatility, analytics solutions combine data sources to enable determining the best way to fulfill transportation needs, from simple point-to-point to complex multimodal, multileg, and cross-dock operations. And if air freight is too costly, or unavailable, analytics helps managers quickly analyze scenarios to evaluate the impact of alternative shipment options.

The Value of Cross-Functional Departments

Perhaps the biggest value of analytics is to create cross-functional departments—those that integrate the view of multiple supply chain functions into a single perspective to enable quick insight and decisions in a volatile market environment. Modern supply chains are a network of suppliers, logistics firms, warehouses, and customers. Sudden disruptions at any stage in the chain affect all parts, but getting visibility on any one part can be difficult.

Let's say there is a sudden spike in demand for things like food, or medicine, or laptops. Supply chain managers will need to work with finance, sales, and others to determine the costs of overtime production, expedited material supply, and shipping.

In another scenario, the demand for a product stops abruptly. Managers will need to gain valuable insight into the cost of shutting down production and the cost of delaying supply (think purchase contracts). Then there is the cost of terminating marketing campaigns, the cost of carrying inventory, and the impact on cash flow.

One of the hardest situations for supply chain managers is to balance and rebalance demand. In this case, the product might be flying off shelves in hard-hit areas one quarter only to be overstocked in the next quarter. This volatile demand can generate extraordinary measures to secure supplies on time. And if data analytics is not part of the process, companies may spend too much time assigning costs to different parts of the business. Often, no one asks: Why was expediting or paying a premium necessary in the first place?

What's needed in each of these scenarios is the capability to ingest data from multiple sources, create a cross-functional view among supply chain, finance, and executive divisions, and conduct scenario and predictive analysis. 

Analytics, Big Data, and The Innovator's Dilemma

I saw Clayton Christensen speak at a Harvard Business Review-sponsored event in the mid-2010s. He was hard to miss. Towering at 6-feet 8-inches, Christensen was reflecting about his acclaimed work, The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail.

As a response from the moderator's question, "What should people remember about this book?"—Christensen peered out into the audience and gently reminded everyone that, "innovation boils down to two levels – sustaining innovation and disruptive innovation. Both have implications for both large, well-funded companies as well as start-ups."

Subscribe to the Oracle Analytics Advantage blog and get the latest posts sent to your inbox

Well said from a thought leader who foresaw a business revolution disrupted by technology. Christensen's follow up "Innovation" books such as The Innovator's Solution, The Innovator's Prescription, and The Innovator's DNA, were all bestsellers. He also wrote dozens of journal articles for Harvard Business Review and MIT Sloan Management Review.

What these books did is map out the past 20 years—a huge disruption for many business models. Few expect this to change. Publishing and media, retail, entertainment, banking, and others have all faced disruption; they will continue to do so, and new ones will join the list. What has been a constant, however, is the need for data collection and analytics insight. But are companies using their data in constructive ways?

(Clayton Christensen - photo from Twitter)

 

In a 2019 interview with MIT Sloan Management Review, shortly before he died in January of 2020, Karen Dillion asked about causal theory; where data from multiple sources can be combined to answer deep questions. Christensen warned that data for data's sake is unnatural.

"…Big data tends to be far more focused on correlation rather than causation and as such ignores examples where something doesn't follow what tends to happen on average. It's only by exploring anomalies that we can develop a deeper understanding of causation. If you think about it, following a big data approach is what powered our understanding of the sun, moon, stars, and Earth for years, but it was only when Galileo peered through a telescope that we could start to understand more deeply how these celestial bodies moved in relation to one another."

Let's assume that the tidal wave of data will continue. I argue that to find deeper understanding requires not only advanced business intelligence and data analytics but innovation at the customer level.

Know Your Customers

Let's take Christensen as a customer example. We already stated he was tall. He was a writer and contributed to business publications. But would those data points indicate he would be interested in subscribing to The New York Times?

His colleagues at Harvard Business Review suggest, probably not.

"His reasons for buying the paper are much more specific. He might buy it because he needs something to read on a plane or because he's a basketball fan and it's March Madness time. Marketers who collect demographic or psychographic information about him—and look for correlations with other buyer segments—are not going to capture those reasons."

Christensen and his associates know there is nothing that can make innovation easy, painless, or foolproof. What is possible is that an organization can implement processes intended to promote the development of innovative ideas, rather than making them a serendipitous event; implement those ideas in a systematic way; and learn from them, no matter how much – or how little – success they had.

"I believe that in order for our scientific understanding of the world to progress, we must continually crawl inside companies, communities, and the lives of individuals to create new data in new categories that reveal new insights," Christensen said in his last interview.

The fundamental problem is, most of the masses of customer data companies create is structured to show correlations: This customer looks like that one, or 68 percent of customers say they prefer version A to version B. While it's exciting to find patterns in the numbers, they don't mean that one thing actually caused another. And though it's no surprise that correlation isn't causality, we suspect that most managers have grown comfortable basing decisions on correlations.

By contrast, using analytics tools allow you to move beyond correlations. The data determines everything—not users' assumptions. Machine learning can also parse data to identify which datasets to access in response to a query about future outcomes. The model also rephrases the natural language query from the human into machine language that is impartial. The predictions then typically are more accurate and, in many cases, much faster.

Keep in mind that innovation is required from the top down, the bottom up, and everywhere in-between. Don't assume that only certain types of data will unlock the insight you seek. Innovation can strike anywhere. It's up to you to recognize it, grab it, and leverage it when it does.

10 Cognitive Biases in Business Analytics and How to Avoid Them

We like to think that our decisions are based on rational facts and not a guess or a hunch. However, that is not always the case and sometimes our biases influence our thinking. Even at the most data-driven companies, allowing for some predispositions can negatively impact results.

A classic example of cognitive bias is the phrase, sour grapes. In the Aesop fable, a fox sees some juicy grapes but cannot reach them. Because of that, he assumes the grapes must be sour and moves on. Psychologists suggest the fox blindly puts a negative word (sour in this case) on the grapes to avoid the disappointment and pain of being unable or unwilling to pursue them. So too, in our own business, it is easier to change the interpretation of the data or artificially manipulate the data to fit our own agenda and avoid disappointment.

As long as you know that a bias may exist, it's up to you to either use it in your data narrative or eliminate it from your analytics. Here are 10 common cognitive biases that can interfere with your data insight and some suggestions for overcoming those obstacles.

Anchoring Bias—you rely too heavily on, or "anchor", on one trait or piece of information when making decisions. Usually, it's the first piece of information acquired on that subject. How often do the headlines dominate your boardroom discussions? Oracle Analytics allows you to integrate multiple sources of data, so you don't have to limit your decisions to the loudest voices.

Availability Heuristic—you overestimate the likelihood of events and are influenced by how recent the memories are or how unusual or emotionally charged they may be. For example, you consider that your supply chain is immune to global events because you use local suppliers. Those suppliers, however, may be at risk themselves to outside forces. A better approach might be to use analytics data modeling to consider all options.

Choice-Supportive Bias—you remember your choices as better than they were. I have the best dog in the world (says, everyone) … except that they do bite people occasionally. Oracle Analytics allows you to apply a model within a data flow to generate a data set and see those results in a data visualization. In that way, you can spot all the times Spot was a good boy and the times he wasn't.

Clustering Illusion—you overestimate the importance of small runs, streaks, or clusters in large samples of random data (that is, seeing phantom patterns). Why guess? Oracle Analytics has a clustering model that can evaluate groups with similar traits. For example, you might assign your costumers into clusters (such as big-spenders, regular spenders and so on) based on their purchasing habits.

Confirmation Bias—you search for, interpret, focus on and remember information in a way that confirms one's preconceptions. If you were trained to think that customer relationships close sales and losing a sale meant the price was too high, you will have a bias toward data that proves otherwise. It's better to back up a decision with data and becoming data-driven.

Information Bias—you seek information even when it cannot affect action. Is there such a thing as too much data? While some may say, no, there are times when you want to look at specific parameters. Oracle Analytics data visualization feature allows you to identify and isolate specific data attributes, thereby allowing you to see the data you want and when you want it.

Ostrich Effect—you ignore an obvious (negative) situation. Research suggests that people often avoid looking at their debts if they are far behind on their payments. The remedy is to ensure you have a reliable notification system that uses machine learning and artificial intelligence to flag up any problems within the parameters you set. Oracle Analytics Mobile in Analytics has such a notification.

Pro-Innovation Bias—you have an excessive optimism towards an invention or innovation's usefulness throughout society, while often failing to identify its limitations and weaknesses. This is best illustrated in the rash of desktop analytics products that are limited in their usefulness. They can neither scale-up or scale-out. Choosing a cloud-based analytics tool that integrates into all parts of the business and can provide a single source of the truth can help prevent this bias.

Recency Illusion—you hold onto the illusion that a word or language usage that one has noticed only recently is an innovation when it is in fact long-established. Not everyone or everything is "industry-leading." The way to measure performance is to generate detailed information. In this way data is sourced without bias from the top data providers.

Zero-Risk Bias—you prefer reducing a small risk to zero over a greater reduction in a larger risk. Everyone loves a sure thing. The problem is that true progress comes when we make bold moves. Data-driven companies know to rely on augmented analyticstools like Oracle Analytics to uncover deeper patterns and predict trends for impactful, unbiased recommendations.

While even the most seasoned data scientist is susceptible to these cognitive biases, understanding them and identifying them in your decision making can help you improve your overall business analytics.

Driving Consumer Loyalty with Business Analytics

Customer loyalty is everything in the world of consumer-packaged goods (CPG). Products can fly off the shelves or sit unsold for months based on the needs and perceptions of a fickle audience. To keep ahead of the competition, many CPG companies have turned to a coordinated analytics strategy.

The stakes for CPG companies could not be higher. Industry sales in the United States alone are expected to grow to $721.8 billion in 2020 up from the $635.8 billion recorded in 2015, according to compiled data at Statista.com. Much of the growth comes from emerging market economies and the subsequent increase in global consumption.

One tactic in gaining market share and mind share is the use of analytics in customer clustering. More advanced than market segmentation, manufacturers and retailers cluster data into units with similar behaviors that may be managed together to predict desired outcomes.

To help make sense of the impact of data analytics on the CPG industry, we invited Ashish Joshi—senior director of data, analytics, and data science at The Clorox Company—to our Analytics Advantage podcast. A 17-year veteran of Clorox, Joshi leads a team of data management experts, advanced analytics professionals, and data scientists spread across business divisions and centers of excellence. These teams work on different analytics focus areas like business performance tracking and forecasting, advertising, trade promotions, market structures, and marketing personalization.

"The retail landscape is changing, and e-commerce is putting a tremendous amount of pressure on retailers, leading to a greater emphasis on private label and that drives pressure on brand manufacturers like Clorox," Joshi says. "Everyone is in a battle for consumer insights. So, we must get smarter and faster in generating those insights. We communicate this to our retailers make sure we keep the overall value front and center as we are building our products and creating our advertising. That's where our analytics comes in."

Like Clorox, consumer goods companies rely on a limited amount of available transactional data. Manufacturers might know where and when an item was sold and for how much, but nuanced data such as what else might have been in the shopping cart at the time of purchase could lead to deeper insight and brand loyalty.

"We have a lot of data about our consumer that can be used to create consumer clusters," Joshi says. "We are not to a point where we are personalizing to a specific person but we are personalizing to groups of consumers. The number of groups of consumers that we talk to is growing."

In the last three years, Joshi and his data science team have poured over the consumer segments with analytics tools to find out more about how these consumer clusters can be created in a way that can drive the effectiveness of our interactions with the consumers. This is similar to the age-old market segmentation, just order of magnitude more granular.

"We have done segmentation for so many years," Joshi says. "This has resulted in five or six consumer clusters. With the amount of data that is available now, we have our data science team identifying many more different clusters and that can help us serve our consumers better."

Joshi says another key to his team's success is using analytics to see how dynamically the attributes are changing. "Analytics helps us identify those different consumer attributes that really matter," Joshi says. "I could come up with 10,000 attributes, but I have to focus on the ones that are really important so we can supply customers with the right offers and communications."

Often, CPG companies will use promotions to gauge how individual products are performing. These often use trade analytics that looks at trade promotion programs. By working with sales teams and retailers, the analytics helps companies determine how and where the promotions are working and potentially how to optimize them in the future.

"Pricing works off of fundamental measurement of the impact of our pricing and promotions. Joshi says. "Some of the pricing that we do is more tactical, but we are also involved with strategic pricing where we look at our three-year plans, we look at our products, we look at the markets, and we look at our competitive positions. From that we make recommendations and we look at what types of activities can be put in place to keep pricing low."

Building the right data pipelines is critical to companies like Clorox for measuring and analyzing marketing optimization and effectiveness.

"You can look at specific brands and see how much they are spending on streaming videos versus banners versus television and optimizing the overall video spend," Joshi says.

To hear more about Clorox and customer clustering, click on the photo below:

Data Scientist Kirk Borne Discusses Business Impact of AI and ML

Adding artificial intelligence, machine learning, and other cognitive interactions to traditional business processes and applications enable greatly improved user experience and productivity. These technologies are already impacting all levels of the business including finance, marketing, human resources, and sales.

Proponents point to incremental changes in the way we work with AI and ML. From the auto-complete or autocorrect on your phone, to chatbots, and virtual assistants like Siri and Alexa. But larger implementations are also taking hold. The New York Metropolitan Transit Authority (MTA) is using machine learning to understand rider patterns and will be changing color coding of their maps to help commuters and tourists get to their destinations faster.

However, these visible AI and ML projects are more the exception than the norm. Industry experts point to massive interest and investments over the next few years:

• 61 percent of organizations surveyed by O'Reilly Media identify machine learning as the most significant data initiative for 2019
• $57.6 billion in spending on AI and machine learning will happen by 2021 compared to $12 billion in 2017, estimates analysts with Deloitte.
• $3.5 to $5.8 trillion in potential annual AI-derived business value across 19 industries is very possible, according to the McKinsey Global Institute research paper.

To help put these trends in perspective, we invited Kirk Borne to the Oracle Analytics Advantage podcast to discuss the value of AI and Machine Learning for business decision-making. Borne is often touted as a worldwide influencer on data science and one of the leading thinkers on the topic of large scientific databases and information systems.

Borne currently consults clients for Booz Allen Hamilton. He's a researcher, blogger, data literacy advocate, TEDx speaker, and author of several books. He's even a project scientist for NASA as part of its Space Telescope Science Institute.

During his interview, Borne explained that the successful evolution of AI and ML in business should be linked to solving the questions at hand

"I like to remind people that machine learning is an algorithm that learns from experience—it detects and recognizes patterns in data," Borne says. "So, if you are classifying a disease or if you are classifying a customer, or information in your weblog, once you see that pattern, you can take the appropriate action."

While the promise of AI and ML are encouraging, one strong barrier to adoption is a cultural one, according to Borne.

"If you are going to have AI help you in decision making, that means letting go of some authority of your own," Borne noted with a nod to top executives in contrast to mid-level managers who may see AI and ML as a threat to their jobs.

To help mitigate cultural changes, Borne advocates for a group of data experts instead of a single decision-maker when it comes to implementing AI and ML.

"Whether the person is the chief data officer, chief data scientist or chief algorithm officer—which is a term I heard the other day—this person should have the view that people should be empowered to speak up when they see something in the data that can help improve the business they are empowered to bring that forward – it's not just the executive suite or the data scientists."

Thankfully, many companies are already setting themselves up for success by adopting a cloud-based analytics infrastructure, which he says is needed to keep costs down. 

"If all you need are a few minutes to process data and some of these major cloud providers are providing the tools that also use the cloud—then you don't have to incur any more costs," he says.

Borne is also eager to change the meaning behind the acronym of AI. Whereas most people identify AI as artificial intelligence, Borne suggests businesses think of decision-making with the "new AI" where the letter A could mean Accelerated, Actionable, Adaptable, Amplified, Assisted, or even Augmented.

"I think taking full advantage of artificial intelligence is a stretch even for the major companies that announce they are 'AI First,'" Borne says. "Even the ones that are fully using AI tell me they are still on the growth curve. Your average company has typically not even started with AI, which means there are a lot of opportunities to get started."

To hear the full interview, check out the podcast by clicking on the podcast photo below.

Analyzing the Blockchain Advantage

Blockchain is making an impact everywhere around us. From its origins as a cryptography method debuting ten years ago, the practice of securing data blocks in a chain led the way to develop digital currencies, authenticating crowdfunding, and verifying company governance documents, among other uses.

You don't have to look far to see how companies are investing and growing their business with the help of blockchain technology. California recently passed legislation to encompass blockchain technology for electronic signatures and smart contracts. Walmart is using blockchain to track groceries along its supply chain. The related market is forecast to gain revenue worth $20 billion by 2024, up from the $315 million in revenue companies realized in 2015, according to estimates with Transparency Market Research.

But as in all things data, analyzing the information helps keep companies competitive. There's a lot of hype around blockchain thanks to Bitcoin and other cryptocurrencies. However, we know the technology has practical uses in business and is even more valuable when using cloud-based analytics to derive insights.

Vinny Lingham

To help make sense of it all, Oracle Analytics invited Vinny Lingham to our podcast. Lingham is a well-recognized South African Internet entrepreneur and an active technology investor focused primarily on Bitcoin and Blockchain projects. He's also an investor on the South Africa version of the television show, Shark Tank, where they have dubbed him the "Bitcoin Oracle." Lingham is currently the co-founder and CEO of Civic Technologies, a blockchain-powered startup company that specializes in digital identity. Before Civic, he was the co-founder and CEO of Gyft, a mobile gift card company founded in 2012 and backed by Google Ventures.

As part of our conversation, we asked about the balance of enterprise adoption of blockchain versus the consumer adoption of blockchain applications. Lingham recounted his company's work with beermaker Budweiser to make a prototype crypto beer vending machine, showcased at a recent conference (Consensus 2018).

While it sounds like a fantastic idea at a sporting event or music festival, it could be a legal nightmare if someone under the legal drinking age got served a Bud. One solution might be to enforce strict identification rules and monitor the vending machine with a guard. Lingham suggested a different tactic.

"We were in discussions about age verification with Budweiser, but for them, the blind spot is that they can't sell beer in vending machines because you can't tell how old the person is when they order," Lingham said. "We have a technology that verifies a person's identity using their phone and blockchain technology."

And while a vending machine for beer is more proof-of-concept than ready to deploy, Lingham says companies need to invest now into blockchain and learn how to analyze the data to establish a foothold on the future.

"It's going to be a long-term thing in trying to institutionalize blockchain," Lingham said. "This kind of technology could be used for alcohol, sure… but what about vending specific amounts of prescription medication to avoid overdosing, or verifying documents for General Data Protection Regulation (GDPR)."

Click on the photo to listen to the full podcast.

Q&A: James Taylor on Machine Learning for Today's Enterprise

Increasingly, organizations are turning to machine learning to augment their analytics strategies and drive decision-making. But are business leaders using the right tools to make those choices efficient and effective?

James Taylor (@jamet123), CEO at Decision Management Solutions, has never been short of opinions. Recently, Taylor commented on the state of machine learning following a comedic post by XKCD that parodied people's misinterpretation of how to succeed with machine learning.

Taylor's observations included:

1) Just because the data can tell you something, there's no guarantee that the business cares about it, can use it, or will use it for decision-making.

2) Most business decisions require policies and regulations to be applied too, not just the analytic insight from your data. Simply pushing data through machine learning or other analytic algorithms won't tell you this.

In the end, Taylor says there is no substitute for clearly identifying a business problem and knowing what decision (or decisions) the business needs to make more accurately to achieve this.

We spoke with Taylor about machine learning and determining how enterprises can use advanced analytics to improve their outcomes:

Michael Singer: Let's talk a little bit about your recent thoughts on machine learning. This is obviously a very hot topic for enterprises that are trying to get either the vision of what's going on or trying to do some predictive style conversations. What is driving people to think more about this?

James Taylor: There are a couple of things. There are the people who already knew about data science and analytics and predictive analytics, and they're looking at machine learning to process whole new data sets. Then there's another group of people who are looking at machine learning and going, "Gosh, this lets us get advanced analytics without hiring data scientists. We can just pile the data up and run it through machine learning algorithms and we'll be like Google."

That second group is going to be bitterly disappointed because there's still no substitute for knowing what your problem is and having a way to use it.

I think people need to recognize that they've got to tease out the problem more before they apply machine learning to it.

Michael Singer: Can you give some examples of companies that are successful with machine learning that are asking the right kinds of questions?

James Taylor: The best examples are the places where you have constantly changing data, or complex and new kinds of data, or new kinds of data sources, or new kinds of data getting added. It's not always obvious what the best way to analyze that data is going to be to come up with the prediction that you need.

New data keeps coming up. There are existing data sources; you can buy new data; your customers start generating more data; and now you've digitized a new thing. So the ability of machine learning to try lots of different algorithms and say, "OK, if you're trying to predict this, and you've got this data, you know, you should use this combination of things."

I think this is very helpful. Coming up with combinations on your own can be a real slog if you're trying to do that by hand. There's a danger, however. Companies will often keep applying the same algorithms they've always used. If you look at the data mining surveys, regression, classification, and decision trees, those still get used more than all the other algorithms added together.

I think machine learning helps to push on that a little bit. You know, it gives an analytically minded person a way of saying, "OK, we got a bunch of new kinds of data here. I don't know what the best way to do it is. Try different things."

Michael Singer: What would you say are the most critical issues facing companies looking to implement machine learning?

James Taylor: The most important thing is to know what problem you're trying to solve. There is a value for pure research in data, right? Just piling up the data and hiring smart people to look at your data and say, "Tell me an interesting story about this data."

Look at the data and tell me things that will help me do those things in a more profitable way. I don't care how accurate the model is, or how interesting it is, or how cute an algorithm it uses, or how much data it uses. What I want to know is, how is it going to help me make more money?

Pushing back on this idea that you can begin with the data and always work from the data, and always begin with the business problem and work backward to the data. And sure, that's going to drive the data into a data lake, because you're going to say, "To solve that business problem, I need to build these kinds of algorithms."

Those kinds of algorithms, frankly, require access to all the emails people have ever sent us, and the only way I can do that is if I put all the emails in the data lake. OK, fine. Put them in a data lake. But you've got a business problem in mind, right?

The second thing is to think about this as an industrial process, right? It's a little bit like when IT started in most companies. The first few IT things built were handcrafted, and what I do is hire one really smart guy who writes this code in the back, or assembly, or whatever it is.

Unfortunately, you can't just go out and magically hire data science unicorns who are going to solve your problem for you. You've got to find people internally. You've got to retrain people. You've got to come up with processes. You've got to put this infrastructure in place, and you've just got to do the hard work of making this a core business capability.

Michael Singer: It sounds like you have to be a cynic to be able to answer these questions, right?

James Taylor: Right. Many years ago, I worked for a consulting firm. We were merging the IT methodology with the process improvement methodology. And one of the guys there turned to me and said, "James, the thing you IT people forget is that in the end, all projects are organizational change projects." And I think of that often when I see the analytics people. "Oh, it's an analytics project." And I'm like, "No, in the end, it's not." In the end, it's an organizational change problem.

Analytics and Data Prep Create Box Office Magic

Long before Moonlight won the Academy Award for Best Picture or Casey Affleck was chosen as Best Actor in Manchester by the Sea, adaptive analytics helped film producers and investors determine the risks and rewards of these and hundreds of other films last year.

The secret to choosing a winning film lies within the numbers gathered from historical movie metadata (such as genre, cast and crew, MPAA rating, production companies, and distributors) and aggregate financials (including production budget, box office, and DVD and Blu-ray sales, not to mention licensing to streaming platforms such as Netflix and Hulu). Independent filmmakers and their financiers often look at such data and use that as the basis for investment.

The analytics behind these numbers are run by Bruce Nash, who founded Nash Information Services 20 years ago. The company operates The Numbers, which provides free information on box office and video sales, and OpusData, which includes a database of movies, metadata, people, and companies involved in the film industry, as well as all time-period data such as daily, weekend, and weekly box office and weekly video sales.

When we talked, Nash described, Moonlight—which was produced for $1.6 million—as a longshot because the main ensemble was generally unknown. By contrast, he described Manchester by the Sea as a lower-risk film because it had recognizable stars and an $8.5 million budget. However, Moonlight has performed better at the box office, earning nearly twice as much as Manchester since taking home the Best Picture award.

“As well as doing in-depth analysis for individual clients, we license OpusData for use on partner websites and applications. They might merge our information with their own proprietary data to help them identify which projects to invest in and which ones to avoid,” says Nash.

Bruce Nash

We also asked Nash about the effort it takes to prepare data for his clients who may need the information as it happens.

Michael Singer: What type of data prep does your company have to engage in to make sure the information is clean and reliable?

Bruce Nash: Basically, we sanity-check everything! The back end is, in effect, a massive rule set that we expect all the data to abide by. At the very simplest level, this might be something like “today’s box office total for movie X is normally yesterday’s total plus today’s daily box office.” But over the years, we’ve added many rules that collectively keep the data in shape. For example, “movies tend not to have sequels unless they’re profitable” is a rule that can be used across the entire data set to tell us quite a lot about how movie financing works.

Michael Singer: To what extent do your clients know how to set up and use the data to begin their visualizations?

Bruce Nash: That depends on how we supply the data. For data extracts, it’s a matter of loading a .csv file into your favorite analytical tool. Our API comes with a replicator that makes it easy to mirror our database in real-time into another database system. Generally, even clients with high-end needs can be up and running on day one.

Michael Singer: Are there data sources that your company has to deal with that are either proprietary in nature or sensitive in their function? How do you deal with this type of situation?

Bruce Nash: All of the data we sell ultimately tracks back to information that is publicly available, although we might be doing a huge amount of number-crunching between data gathering and estimating such as total video-on-demand rentals for a particular film. We do use proprietary data from our clients [film producers] to build and fine-tune our models, but that data and the model itself are not made public.