Month: October 2019

[youtube https://www.youtube.com/watch?v=MYrijdCA0QY]

In this last part of two part podcast @TimothyChou discussed the Internet of Things landscape’s future. He laid out how internet has always been about internet of things and not internet of people. He sheds light on internet of things as it is spread across themes of things, connect, collect, learn and do workflows. He builds an interesting case about achieving precision to introduction optimality.

Timothy’s Recommended Read:
Life 3.0: Being Human in the Age of Artificial Intelligence by Max Tegmark http://amzn.to/2Cidyhy
Zone to Win: Organizing to Compete in an Age of Disruption Paperback by Geoffrey A. Moore http://amzn.to/2Hd5zpv

Podcast Link:
iTunes: http://math.im/itunes
GooglePlay: http://math.im/gplay

Timothy’s BIO:
Timothy Chou has his career spanning through academia, successful (and not so successful) startups and large corporations. He was one of only a few people to hold the President title at Oracle. As President of Oracle On Demand he grew the cloud business from it’s very beginning. Today that business is over $2B. He wrote about the move of applications to the cloud in 2004 in his first book, “The End of Software”. Today he serves on the board of Blackbaud, a nearly $700M vertical application cloud service company.

After earning his PhD in EE at the University of Illinois he went to work for Tandem Computers, one of the original Silicon Valley startups. Had he understood stock options he would have joined earlier. He’s invested in and been a contributor to a number of other startups, some you’ve heard of like Webex, and others you’ve never heard of but were sold to companies like Cisco and Oracle. Today he is focused on several new ventures in cloud computing, machine learning and the Internet of Things.

About #Podcast:
#FutureOfData podcast is a conversation starter to bring leaders, influencers and lead practitioners to come on show and discuss their journey in creating the data driven future.

Wanna Join?
If you or any you know wants to join in,
Register your interest @ http://play.analyticsweek.com/guest/

Want to sponsor?
Email us @ info@analyticsweek.com

Keywords:
#FutureOfData #DataAnalytics #Leadership #Podcast #BigData #Strategy

Originally Posted at: @TimothyChou on World of #IOT & Its #Future Part 2 #FutureOfData #Podcast

Customer loyalty is the leading indicator of business growth. In fact, a main reason why companies implement voice of the customer (VoC) initiatives is to improve customer loyalty. Based on a 2010 study by Gleanster, asking 276 companies about their customer feedback management initiative, a majority of the loyalty leading companies said they implemented their program to increase customer loyalty, increase customer retention and increase customer satisfaction.

There are many different ways customers can engage in loyalty behaviors toward your company or brand. They can remain a customer for a long time. They can recommend you to their colleagues and friends. They can even show their loyalty by purchasing additional products/services from you. These loyalty behaviors, in turn, drive different types of business growth: overall customer growth, new customer growth, and average revenue per customer.

Customer relationship surveys, the foundation of many VoC programs, are used to measure customer loyalty, along with other important customer variables (e.g., satisfaction with their experience).  Including the right loyalty questions in your customer survey is essential to an effective VoC program. Companies use these surveys to understand and diagnose problem areas that, when fixed, will increase customer loyalty.

Not all Loyalty Questions are Created Equal

I have developed a set of customer loyalty questions that measure different types of customer loyalty. These loyalty questions have been shown to be predictive of different types of business growth and can be grouped into three sets of loyalty behaviors: retention, advocacy and purchasing. Each set of loyalty behaviors contains specific loyalty questions. Research shows that questions that fall into the same set are essentially interchangeable because they measure the same thing. Some of these customer loyalty questions appear below.

Retention Loyalty: the extent to which customers remain customers and/or do not use a competitor

• How likely are you to switch to another provider? (0 – Not at all likely to 10 – Extremely likely)
• How likely are you to renew your service contract? (0 – Not at all likely to 10 – Extremely likely)

Advocacy Loyalty: the extent to which customers advocate your product and/or brand

• How likely are you to recommend us to your friends/colleagues? (0 – Not at all likely to 10 – Extremely likely)
• Overall, how satisfied are you with our performance? (0 – Extremely dissatisfied to 10 – Extremely satisfied)

Purchasing Loyalty: the extent to which customers increase their purchasing behavior

• How likely are you to purchase different solutions from us in the future? (0 – Not at all likely to 10 – Extremely likely)
• How likely are you to expand the use of our products throughout company? (0 – Not at all likely to 10 – Extremely likely)

Using Different Types of Loyalty Questions

Selecting the right customer loyalty questions for your survey requires careful thought about your customers and your business. Think about how your customers are able to show their loyalty toward your company and include loyalty questions that reflect those loyalty behaviors you want to manage and change. Additionally, consider your business growth strategy and current business environment. Think about current business challenges and select loyalty questions that will help you address those challenges. For example, if you have a high churn rate, you might consider using a retention loyalty question to more effectively identify solutions to increase customer retention. Additionally, if you are interested in increasing ARPU (average revenue per customer), you might consider including a purchasing loyalty question.

Using a comprehensive set of loyalty questions will help you target solutions to optimize different types of customer loyalty of existing customers and, consequently, improve business growth. Including a “likelihood to quit” question and a “likelihood to buy different” question can help you understand why customer are leaving and identify ways to increase customers’ purchasing behavior, respectively.

Customers can engage in a variety of loyalty behaviors. Companies need to think about customer loyalty more broadly and include different types of loyalty questions that meet their specific business needs and comprehensively capture important loyalty behaviors.

Source: Try these Customer Loyalty Questions for your Relationship Survey

As embedded analytics become increasingly prominent in the business intelligence (BI) landscape, the question of whether companies should build or buy embedded BI applications seems to be more relevant than ever. The numerous attempts to answer this question ignore the basic fact that the question itself is misleading since for most organizations there is not a simple yes-or-no answer. Instead, best practices for embedded analytics are neither “build” nor “buy” — but is, in fact, more akin to partnership.

Understanding the Question

“Embedded analytics” is a blanket term that describes the integration of various features of business intelligence tools into other applications (often, but not exclusively, in SaaS). For example, a company that develops CRM software might want to provide more in-depth insights from the data it collects to either enhance the company’s general value proposition or to sell a premium service. Hence it may look to incorporate features such as data transformation, rapid big data querying or interactive visualizations to its own CRM software package.

Most professionals in the BI industry would agree that embedded reporting has become a major area of focus for both business and technology. Customers are demanding self-service, meaningful access to data, and competition is forcing companies to accommodate these demands, which in turn leads to more focus on building these types of capabilities.

In-House or Out-of-the-Box

The question of “to build or not to build” has become the subject of heated discussions when considering an embedded analytics project. Run a quick Google search for “build vs buy embedded analytics,” and you’ll be bombarded with page after page of articles asking and attempting to answer this exact question. I will briefly present the most common arguments for each side of the debate:

Developing BI features in-house gives companies more flexibility and control over the end product. The original application developer is the most intimately familiar with its product and customers, and so will be able to tailor a solution more precisely. Building BI features in-house, however, requires a significant investment and often yields sub-par results due to the level of investment required and the need for specialized skills.

Buying an “out-of-the-box” solution enables a company to leverage the massive investments already made by the BI provider and gives access to state-of-the-art BI capabilities.
In a majority of cases, companies that seek to provide meaningful data analysis capabilities to their customers would be better off looking to embed an existing product rather than starting from scratch. However, what I would like to stress is that the way this question is posed is in itself misleading: by far, the more common — and preferable — scenario is actually neither build nor buy, but a third solution that could more accurately be described as partnership.

Business Intelligence is Not a Commodity Product (Yet)

When people talk about “build vs buy,” one might get the impression that the option exists to go online and buy a turnkey embedded BI solution, which one can easily plug into an existing product and presto! Instant customer-facing analytics. Sadly, when it comes to more sophisticated needs and products, this is almost never the case.

I do not mean to imply that BI implementations need to be lengthy or difficult affairs, but merely that each implementation is different. A company that typically wants to present a hundred thousand rows of data to its customers does not need the same technological “muscle” as one that works with a hundred million rows; likewise, data that comes from dozens of structured and unstructured sources is quite different than neatly-organized tables in a SQL database. High-level data visualization is one thing (for example, an e-commerce app that displays traffic and sales to sellers), whereas advanced analytics, drill-downs, and customizable reports require entirely different capabilities.

When it comes to these types of more advanced use cases, the notion of a one-size-fits-all solution is unrealistic: the analytical features will need to be integrated into the existing application and customized to meet the exact needs of the specific product and customer base in terms of data modeling, security, management and reporting. Again, this is not to say that these integration efforts need to be overly complicated or require extensive development resources — however, they will require an understanding of the underlying data, and the ability to easily customize and communicate with the BI platform via API access.

Partnership, Not a One-Time Transaction

The decision to use an external provider for embedding analytics is more similar to a partnership than to a “get it and forget it” type of purchase. The developer and the BI provider work together to build the required data product, and continue to collaborate as products mature, new features are added and new needs arise.

Does this mean that the developer will have to rely on the BI provider for every change or customization? Absolutely not — developers should have complete independence and control over their own product. They should be the sole owner of the product, from end to end, and be able to develop it on their own, without having to rely on a vendor’s professional services or external consultants. In order to achieve such an outcome, developers should partner with a BI vendor that is an enabler, always keeping developers in mind. Best practices include maintenance of a comprehensive SDK, with excellent documentation, and designing the BI product as an open platform.

Open platforms enable easy access via commonly used APIs, ensuring the BI software is flexible enough to integrate with the developers’ existing systems seamlessly, and accommodating specific needs and requirements around data sources, security and similar considerations. And for the truly complex, heavyweight implementations — top BI vendors provide the professional resources needed to get customers up and running as fast as possible and to address the various maintenance issues that inevitably arise.

Furthermore, both parties should see their relationship as long term — new features introduced in the BI platform should always be built in an “API-first” approach, enabling application developers to quickly and easily incorporate these features into their own offering; communication between the BI vendor and the application developer needs to be open and frequent so that both can gain a better understanding of the other’s strengths and limitations and adjust development, support and account management efforts accordingly.

Understanding embedded analytics as an ongoing partnership, rather than a one-off purchase, will lead developers to ask more relevant questions before embarking on an embedded BI project; and lead BI providers to make a serious commitment to building truly open platforms, maintaining superb customer service and documentation. In such cases, everyone stands to benefit.

Source: Embedded Analytics: The Build vs Buy Debate is Pointless by analyticsweek

Artificial Intelligence may well be the most influential suite of technologies to impact the enterprise for the next several years. Its learning capabilities and decision support propensity are horizontally acclaimed; many consider machine learning a panacea for data-driven processes in general.

Still, an often neglected aspect of the transformative impact of AI is exactly how to implement it to gain immediate benefits. Most organizations lack the resources to deploy progressive intelligent agents for conversational interactions about their array of products, services, and organizational functions. Other simply don’t know where to start because of the wide range of AI technologies and the variegated nature of their own business needs.

Nevertheless, there are a couple of recurring use cases which are applicable to most organizations in which AI consistently delivers quick wins. The most accessible involve using these technologies for recommendation engines and process automation—both for front offices and back offices.

According to Razorthink CEO Gary Oliver, the tangible business value rapidly gained from these use cases sets the foundation for “opportunities to use AI to impact business processes from the front end with their customers all the way throughout the backend processes.”

Recommendations
Few approaches can match those of AI for creating timely, incisive recommendations. The front office appeal of recommendation engines empowered by AI is apparent since they’re instrumental in personalizing customer interactions. Moreover, this use case leverages two of the core strengths of AI: its incorporation of both predictive and prescriptive analytics. The former ascertains what customers are likely to want, need, or do in relation to the products and services they’re offered. The latter creates action for them in the form of cross-selling and up-selling opportunities. Transactional data—at scale—is ideal for “looking at all the transaction patterns and trying to understand behaviors by microsegmenting customers into groups that act the same,” Oliver explained. Deep learning’s propensity for finding minute patterns in even the largest big data sets facilitates its capacity for microsegmentation. AI can then determine “what’s the right next best thing to do for that client, whether that has to do with customer service, how to make them happy and successful, or it may also be the ability to try and sell them additional products or services,” Oliver commented.

Front End Process Automation
The propensity to automate routine processes is another one of AI’s more celebrated traits. Front office applications of this capability generally revolve around even rudimentary chatbots, which interact with customers about what is typically a set of frequently asked questions. Such mechanisms can be enhanced by more substantial intelligence on the back end, and are often the initial means of directing customers to the appropriate personnel for their needs. Oliver observed that for most automated use cases of AI “a lot of it starts with the front end customer interaction, customer support chatbot kind of thing.” Chatbots, however, are merely the first of many process automation applications of AI. Oliver referenced the plentiful instances in which both traditional machine learning and deep learning are used to scan documents and other forms of text so that digital systems immediately understand their meaning via “intelligent OCR (Optical Character Recognition),” Oliver remarked. Deep learning excels in identifying patterns in massive quantities of unstructured data, which is why it’s frequently used for analyzing both video and image data. According to Oliver, there are instances in automotive insurance in which “they want to be able to snap a picture of an accident, and have AI automatically figure out what the claims are by understanding from the picture, [and] recognizing the issues and all the parts that would be involved to fix them.”

Back Office Process Automation
There is a lengthy history of AI’s use for automating processes in back offices. Elements of machine learning enhance a number of data management pillars such as data preparation, transformation, data discovery, and others. Organizations can also use AI to automate processes such as medical claims or other insurance claims. Oliver described examples “for medical insurers, where today doctors have to review all the claims.” Neural networks can automate the claims review process with near 100 percent accuracy rates. Even better, when the AI is unsure about something, Oliver revealed “it has the ability to reach out to the doctor and say is this good or bad or is this x or y. When the doctor communicates back then AI will take that information and learn and won’t have to ask the question again. It will continuously learn.” It’s important to realize that AI can produce this advantage for virtually any process that involves routine, manual methods, such as most types of on-boarding or other frequently used business forms.

Actionable Outputs
The foregoing use cases not only deliver tangible business value but also are attended by a fair amount of flexibility. In most situations, it’s up to the organization—and its own business processes—to determine the most effective means of implementation. With recommendations, for example, they can choose to either automate those prescriptions or have dedicated sales people broach them to customers. The same adaptability exists with both front end and backend process automation. According to Oliver, organizations “want to insert the deep learning and the AI into their processes, but then they want the outcomes that AI generates to inform action. Sometimes that’s automated action, sometimes that’s triggering off alerts for someone to take action.”

Originally Posted at: Low Hanging Fruit: Seizing Immediate Business Value from Artificial Intelligence by jelaniharper

Just three weeks into 2019, Veeam announced a $500M funding round. The company is privately held, profitable and with a pretty solid revenue stream coming from hundreds of thousands of happy customers. But, still, they raised $500M!

I didn’t see it coming, but if you look at what is happening in the market, it’s not a surprising move. Market valuation of companies like Rubrik and Cohesity is off the chart and it is pretty clear that while they are spending boatloads of money to fuel their growth, they are also developing platforms that are well beyond traditional data protection.

Backup Is Boring

Backup is one of the most tedious, yet critical, tasks to be performed in the IT space. You need to protect your data and save a copy of it in a secure place in case of a system failure, human error or worse, like in the case of natural disasters and cyberattacks. But as critical as it is, the differentiations between backup solutions are getting thinner and thinner.

Vendors like Cohesity got it right from the very beginning of their existence. It is quite difficult, if not impossible, to consolidate all your primary storage systems in a single large repository, but if you concentrate backups on a single platform then you have all of your data in a single logical place.

In the past, backup was all about throughput and capacity with very low CPU, and media devices were designed for few sequential data streams (tapes and deduplication appliances are perfect examples). Why are companies like Rubrik and Cohesity so different then? Well, from my point of view they designed an architecture that enables to do much more with backups than what was possible in the past.

Next-gen Backup Architectures

Adding a scale-out file system to this picture was the real game changer. Every time you expand the backup infrastructure to store more data, the new nodes also contribute to increase CPU power and memory capacity. With all these resources at your disposal, and the data that can be collected through backups and other means, you’ve just built a big data lake … and with all that CPU power available, you are just one step away from transforming it into a very effective big data analytics cluster!

From Data Protection to Analytics and Management

Starting from this background it isn’t difficult to explain the shift that is happening in the market and why everybody is talking more about the broader concept of data management rather than data protection.

Some may argue that it’s wrong to associate data protection with data management and in this particular case the term data management is misleading and inappropriately applied. But, there is much to be said about it and it could very well become the topic for another post. Also, I suggest you take a look at the report I recently wrote about unstructured data management to get a better understanding of my point of view.

Data Management for Everybody

Now that we have the tool (a big data platform), the next step is to build something useful on top of it, and this is the area where everybody is investing heavily. Even though Cohesity is leading the pack and has started showing the potential of this type of architecture years ago with its analytics workbench, the race is open and everybody is working on out-of-the-box solutions.

In my opinion these out-of-the-box solutions, which will be nothing more that customizable big data jobs with a nice and easy to use UI on top, will make data management within everyone’s reach in your organization. This means that data governance, security and many business roles will benefit from it.

A Quick Solution Roundup

As mentioned earlier, Cohesity is in a leading position at the moment and they have all the features needed to realize this kind of vision, but we are just at the beginning and other vendors are working hard on similar solutions.

Rubrik, which has a similar architecture, has chosen a different path. They’ve recently acquired Datos IO and started offering NoSQL DB data management. Even though NoSQL is growing steadily in enterprises, this is a niche use case at the moment and I expect that sooner or later Rubrik will add features to manage data they collect from other sources.

Not long ago I spoke highly about Commvault, and Activate is another great example of their change in strategy. This is a tool that can be a great companion of their backup solution, but can also live alone, enabling the end user to analyze, get insights and take action on data. They’ve already demonstrated several use cases in fields like compliance, security, e-discovery and so on.

Getting back to Veeam … I really loved their DataLabs and what it can theoretically do for data management. Still not at its full potential, this is an orchestrator tool that allows to take backups, create a temporary sandbox, and run applications against them. It is not fully automated yet, and you have to bring your own application. If Veeam can make DataLabs ready to use with out-of-the-box applications it will become a very powerful tool for a broad range of use cases, including e-discovery, ransomware protection, index & search and so on.

These are only a few examples of course, and the list is getting longer by the day.

Closing the Circle

Data management is now key in several areas. We’ve already lost the battle against data growth and consolidation, and at this point finding a way to manage data properly is the only way to go.

With ever larger storage infrastructures under management, and sysadmins that now have to manage petabytes instead of hundreds of terabytes, there is a natural shift towards automation for basic operations and the focus is more on what is really stored in the systems.

Furthermore, with the increasing amount of data, expanding multi-cloud infrastructures, new demanding regulations like GDPR, and ever evolving business needs, the goal is to maintain control over data no matter where it is stored. And this is why data management is at the center of every discussion now.

Originally posted on Juku.it

Originally Posted at: From Data Protection to Data Management and Beyond

An auto-encoder, autoassociator or Diabolo network is an artificial neural network used for learning efficient codings. The aim of an auto-encoder is to learn a compressed, distributed representation (encoding) for a set of data, typically for the purpose of dimensionality reduction. Architecturally, the simplest form of the auto-encoder is a feedforward, non-recurrent neural net that is very similar to the multilayer perceptron (MLP), with an input layer, an output layer and one or more hidden layers connecting them.

An auto-encoder is often trained using one of the many back propagation variants (conjugate gradient method, steepest descent, etc.). Though often reasonably effective, there are fundamental problems with using back propagation to train networks with many hidden layers. Once the errors get back propagated to the first few layers, they are minuscule, and quite ineffectual. This causes the network to almost always learn to reconstruct the average of all the training data.

Piotr Mirowski (of Microsoft Bing London) presented Review of Auto-Encoders to the Computational Intelligence Unconference 2014, with our Deep Learning stream. These are his slides. Slides are a bit old but it does a great job in comparing some of the models relevant today.

[slideshare id=38162984&doc=piotrmirowskiciunconf2014reviewautoencoders-140820024912-phpapp02]
Original link here: https://piotrmirowski.files.wordpress.com/2014/08/piotrmirowski_ciunconf_2014_reviewautoencoders.pptx

He also has Matlab-based tutorial on auto-encoders available here:
https://github.com/piotrmirowski/Tutorial_AutoEncoders/

Originally Posted at: Review of Autoencoders (Deep Learning)