Fix the Culture, spread awareness to get awareness
Adoption of analytics tools and capabilities has not yet caught up to industry standards. Talent has always been the bottleneck towards achieving the comparative enterprise adoption. One of the primal reason is lack of understanding and knowledge within the stakeholders. To facilitate wider adoption, data analytics leaders, users, and community members needs to step up to create awareness within the organization. An aware organization goes a long way in helping get quick buy-ins and better funding which ultimately leads to faster adoption. So be the voice that you want to hear from leadership.
[ DATA SCIENCE Q&A]
Q:What is your definition of big data?
A: Big data is high volume, high velocity and/or high variety information assets that require new forms of processing
– Volume: big data doesnt sample, just observes and tracks what happens
– Velocity: big data is often available in real-time
– Variety: big data comes from texts, images, audio, video
Difference big data/business intelligence:
– Business intelligence uses descriptive statistics with data with high density information to measure things, detect trends etc.
– Big data uses inductive statistics (statistical inference) and concepts from non-linear system identification to infer laws (regression, classification, clustering) from large data sets with low density information to reveal relationships and dependencies or to perform prediction of outcomes or behaviors
An application can live or die by its embedded analytics. It doesnât matter if the rest of your product is perfectly designed. If your dashboards and reports have a disappointing user experience (UX), user adoption and customer satisfaction can plummet.
âUser experience matters,â writes Gartner in their recent report, 5 Best Practices for Choosing an Embedded Analytics Platform Provider. âEmbedded analytics [should] not only support embedding of charts and visualizations, but also go deeper and integrate the data and analytics into the fabric of the application. This âseamlessâ approach means that users donât even know they are using a multiproduct application.â
How solid is your analytics UX? Ask yourself these three questions to gauge where and how you can improve your analytics experience:
#1. Do you have a deep understanding of your users?
A lack of understanding about what users need from their dashboards and reports is a challenge that plagues product teams. Many companies fill their analytics with data they think their users want, and never do the due diligence to find out what users actually need.Donât assume what your business intelligence users want. Take the time to research how users will interact with your application, so you can build it with them in mind. Itâs a seemingly obvious but often-missed point: Different end users want to use your applicationâs embedded analytics in different ways.
#2. Does your embedding stop at the visualizations?
Embedded analytics involves more than white-labeling some charts and graphs. Application teams need to look at the complete experienceânot just the visualsâto ensure end users canât tell where your application ends and the embedded analytics begins.A truly seamless experience âallows users to take immediate action from within the application, without shifting context,â notes Gartner in their report. Ideally, you want to integrate the analytics into your usersâ workflows by letting them take action from the analytics, write-back to the database, and share insights in context.
#3. Do the visualizations match your data?
Another common problem is choosing the wrong data visualizations to illustrate your datasets. Most visualizations are good for some types of data, but not every type. For example, a scatter chart works well to display two variables from a dataset, but itâs only useful when there is a number value on each axis; without that, it will appear to be a line chart without the line. Or consider the common pie chart, which is great for four or five valuesâbut completely breaks down when sliced into dozens of sections. These are just two examples of how poor UI/UX can make information difficult for a user to understand.
If youâve answered âyesâ to any of the questions above, itâs time to update your analytics before your customers start abandoning your product for the competition. Learn how to take the next steps in our Blueprint to Modern Analytics guide.
Visualization is not a straight path from vision to reality. It is full of twists and turns, rabbit trails and road blocks, foul-ups and failures. Initial hypotheses are often wrong, and promising paths are frequently dead ends. Iteration is essential. And sometimes you need to change your goals in order to reach them.
We are as skilled at pursuing the wrong hypotheses as anyone. Let us show you.
We had seen the Hierarchical Edge Bundling implemented by Mike Bostock in D3. It really clarified patterns that were almost completely obfuscated when straight lines were used.Â
We were curious if it might do the same thing with geographic patterns. Turns out Danny Holten, creator of the algorithm, had already done something similar. But we needed to see it with our own data.
To start, we simply put a single yearâs (2014) migration data on the map. Our first impression: sorrow, dejection and misery. It looked better than a mess of straight lines, but not much better. Chin up, though. This didnât yet account for how many people were flowing between each of the connections â only whether there was a connection or not.Â
Unweighted edge bundled migration
With edge bundling, each path between two points can be thought to have some gravity pulling other paths toward it while itself being pulled by those other paths. In the first iteration, every part of a path has the same gravity. By changing the code to weight the bundling, we add extra gravity to the paths more people move along.
Weighted edge bundled migration
Alas, things didnât change much. And processing was taking a long time with all those flows. When the going gets tough, simplify. We cut the data into two halves, comparing westward flows to eastward flows.
East to west migration
West to east migration
Less data meant cleaner maps. We assumed there would be some obvious difference between these two, but these maps could be twins. We actually had to flip back and forth between them to see that there was indeed a difference.
So our dreams of mindblowing insight on a migration data set using edge bundling were a bust. But, seeing one visualization regularly leads to ideas about another. We wondered what would happen if we animated the lines from source to destination? For simplicity, we started with just eastward migration.Â
Lasers
Cool, itâs like laser light leisurely streaming through invisible fibre optic cables. But thereâs a problem. Longer flows appear to indicate higher volume (which is misleading as their length is not actually encoding volume, just distance). So we tried using differential line lengths to represent the number of people, sticking with just eastward flows.Â
Star Wars blasters
Here we get a better sense of the bigger sources, especially at the beginning of the animation, however, for some paths, like California to Nevada, we end up with a solid line for most of the loop. The short geographic distance obscures the large migration of people. We wondered if using dashed lines would fix thisâparticularly in links like California to Nevada.
Machine gun bursts
This gives us a machine gun burst at the beginning with everything draining into 50 little holes at the end. We get that sense of motion for geographically close states, but the visual doesnât match our mental model of migration. Migrants donât line up in a queue at the beginning of the year, leaving and arriving at the same time. Their migration is spread over the year.
What if instead we turn the migration numbers into a rate of flow. We can move dots along our edge bundled paths, have each dot represent 1000 people and watch as they migrate. The density of the dots along a path will represent the volume. Â This also has the convenience of being much simpler to explain.
Radar signals
We still have a burst of activity (like radar signals) at the beginning of the loop, so weâll stagger the start times to remove this pulsing effect.
Staggered starts
Voilà . This finally gives us a visual that matches our mental model: people moving over the period from one state to another. Letâs add back westward movement.
Ants
Very cool, but with so much movement itâs difficult to tell whoâs coming and whoâs going. We added a gradient to the paths to make dots appear blue as they leave a state and orange as they arrive.
Coloured ants
Letâs be honest, this looks like a moderately organized swarm of ants. But it is a captivating swarm that people can identify with. Does it give us any insight? Well not any of the sort we were originally working for. No simple way to compare years, no clear statements about the inflows and outflows. If we want to make sense of the data and draw specific conclusions⦠well other tools might be more effective.
But it is an enchanting overview of migration. It shows the continuous and overwhelming amount of movement across the country and highlights some of the higher volume flows in either direction. It draws you in and provides you with a perspective not readily available in a set of bar charts. So we made an interactive with both.
Each dot represents 1,000 people and the year’s migration happens in 10 seconds. Or if you’d prefer, each dot can represent 1 person, and you can watch the year play out in just over 2 hours and 45 minutes. If youâre on a desktop you can interact with it to view a single state’s flow. And of course for mobile and social media, we made the obligatory animated gif.
And just when we thought we’d finished, new data was released and were were obliged to update things for 2015.
Glowing ants
Building a visualization that is both clear and engaging is hard work. Indeed, sometimes it doesnât work at all. In this post weâve only highlighted a fraction of the steps we took. Â We also fiddled with algorithm settings, color, transparency and interactivity. Â We tested out versions with net migration. We tried overlaying choropleths and comparing the migration to other variables like unemployment and birth rate. None of these iterations even made the cut for this blog post.
An intuitive, engaging, and insightful visualization is rare precisely because of how much effort it takes. We continue to believe that the effort is worthwhile.
Fix the Culture, spread awareness to get awareness
Adoption of analytics tools and capabilities has not yet caught up to industry standards. Talent has always been the bottleneck towards achieving the comparative enterprise adoption. One of the primal reason is lack of understanding and knowledge within the stakeholders. To facilitate wider adoption, data analytics leaders, users, and community members needs to step up to create awareness within the organization. An aware organization goes a long way in helping get quick buy-ins and better funding which ultimately leads to faster adoption. So be the voice that you want to hear from leadership.
[ DATA SCIENCE Q&A]
Q:How to detect individual paid accounts shared by multiple users?
A: * Check geographical region: Friday morning a log in from Paris and Friday evening a log in from Tokyo
* Bandwidth consumption: if a user goes over some high limit
* Counter of live sessions: if they have 100 sessions per day (4 times per hour) that seems more than one person can do
With analytics taking a central role in most companiesâ daily operations, managing the massive data streams organizations create is more important than ever. Effective business intelligence is the product of data that is scrubbed, properly stored, and easy to find. When your organization uses raw data without proper management procedures, your results suffer.
The first step towards creating better data for analytics starts with managing data the right way. Establishing clear protocols and following them can help streamline the analytics process, offer better insights, and simplify the process of handling data. You can start by implementing these five rules to manage your data more efficiently.
1. Establish Clear Analytics Goals Before Getting Started
As the amount of data produced by organizations daily grows exponentially, sorting through terabytes of information can become problematic and reduce the efficiency of analytics. Such large data sets require significantly longer times to scrub and properly organize. For companies that deal with multiple streams that exhibit heavy bandwidth, having a clear line of sight towards business and analytics goals can help reduce inflows and prioritize relevant data.
Itâs important to establish clear objectives for data and create parameters that filter out data points that are irrelevant or unclear. This facilitates pre-screening datasets and makes scrubbing and sorting easier by reducing white noise. Additionally, you can focus even more on measuring specific KPIs to further filter out the right data from the stream.
2. Simplify and Centralize Your Data Streams
Another problem analytics suites face is reconciling disparate data from multiple streams. Organizations have internal, third-party, customer, and other data that must be considered as part of a larger whole instead of viewed in isolation. Leaving data as-is can be damaging to insights, as different sources may use unique formats or different styles.
Before allowing multiple streams to connect to your data analytics software, your first step should be establishing a process to collect data more centrally and unify it. This centralization makes it easier to input data seamlessly into analytics tools, but also simplifies the methodology for users to find and manipulate data. Consider how to set up your data streams best to reduce the number of sources to eventually produce more unified sets.
3. Scrub Your Data Before Warehousing
The endless stream of data raises questions about quality and quantity. While having more information is preferable, data loses its usefulness when itâs surrounded by noise and irrelevant points. Unscrubbed data sets make it harder to uncover insights, properly manage databases, and access information later.
Before worrying about data warehousing and access, consider the processes in place to scrub data to produce clean sets. Create phases that ensure data relevance is considered while effectively filtering out data that is not pertinent. Additionally, make sure the process is as automated as possible to reduce wasted resources. Implementing functions such as data classification and pre-sorting can help expedite the cleaning process.
4. Establish Clear Data Governance Protocols
One of the biggest emerging issues facing data management is data governance. Because of the sensitive nature of many sourcesâconsumer information, sensitive financial details, and so onâconcerns about who has access to information are becoming a central topic in data management. Moreover, allowing free access to datasets and storage can lead to manipulation, mistakes, and deletions that could prove damaging.
Itâs vital to establish clear and explicit rules about who can access data, when, and how. Creating tiered permission systems (read, read/write, admin) can help limit the exposure to mistakes and danger. Additionally, sorting data in ways that facilitate access to different groups can help manage data access better without the need to give free rein to all team members.
5. Create Dynamic Data Structures
Many times, storing data is reduced to a single database that limits how you can manipulate it. Static data structures are effective for holding data, but they are restrictive when it comes to analyzing and processing it. Instead, data managers should place a greater emphasis towards creating structures that encourage deeper analysis.
Dynamic data structures present a way to store real-time data that allows users to connect points better. Using three-dimensional databases, finding methods to reshape data rapidly, and creating more inter-connected data silos can help contribute to more agile business intelligence. Generate databases and structures that simplify accessing and interacting with data rather than isolating it.
The fields of data management and analytics are constantly evolving. For analytics teams, itâs vital to create infrastructures that are future-proofed and offer the best possible insights for users. By establishing best practices and following them as closely as possible, organizations can significantly enhance the quality of the insights their data produces.
Traditionally, Office Depotâs European sales were primarily sourced through an offline, mail-order catalog model drive by telemarketing activities. The company has since moved to a hybrid retail model, combining offline and online shopping, which required a data consolidation strategy that optimized the different channels. Additionally, the companyâs myriad of backend systems and disparate supply chain data collected from across Europe had become difficult to analyze.
Using Talend, Office Depot can now ingest data from its vast collection of operational systems. The architecture includes an on-premise Hadoop cluster using Hortonworks, Talend Data Integration, and Data Quality to perform checks and quality control on data before ingesting it into the Hubâs data lake.
Powering Use Cases from Supply Chain to Finance
Integrating online and offline data results in a unified, 360-degree view of the customer and a clear picture of the customer journey. Office Depot can now create more-specific audience segments based on how customers prefer to buy, and tailor strategies to reach the most valuable consumers whether they buy online or in-store. They can compare different offline customer experiences to see how they are influenced by digital ads. Customer service operators have complete information on a customer, so they can talk to them as they know their details.
Office Depotâs data hub approach also provides high-quality data to all back-office functions throughout the organization, including supply chain and finance. Office Depot can now integrate data from the range of supply chain back-end systems in use in various countries, and answer questions such as which distribution center has the most efficient pick-line and why; or which center is in the risky position of having the least amount of stock for the best-selling products.
Data aids, not replace judgement
Data is a tool and means to help build a consensus to facilitate human decision-making but not replace it. Analysis converts data into information, information via context leads to insight. Insights lead to decision making which ultimately leads to outcomes that brings value. So, data is just the start, context and intuition plays a role.
[ DATA SCIENCE Q&A]
Q:Is it better to spend 5 days developing a 90% accurate solution, or 10 days for 100% accuracy? Depends on the context?
A: * premature optimization is the root of all evils
* At the beginning: quick-and-dirty model is better
* Optimization later
Other answer:
– Depends on the context
– Is error acceptable? Fraud detection, quality assurance
Buying and selling stocks has dramatically changed since the advent of the web.
Online brokerages have made trades more accessible, faster, and dramatically cheaper for the retail investor.
Prior to the web, it was common for a full-service broker to charge 2.5% for a stock trade. Now transactions are low-cost commodities with many trades costing less than $10 (or even free).
But the ease of executing trades, access to account data, and the tsunami of financial information available online hasnât taken away some common challenges for the typical investor.
To understand this experience, we benchmarked the user experience of six top US-based brokerage websites to understand how people are using these services and where the process can be improved.
We benchmarked the desktop and mobile user experiences of the following six brokerage websites:
Charles Schwab (schwab.com)
E*Trade (us.etrade.com)
Fidelity Investments (fidelity.com)
Merrill Lynch (ml.com)
TD Ameritrade (tdameritrade.com)
Vanguard (investor.vanguard.com)
We collected SUPR-Q data, including NPS data, and investigated reasons for using the website, usersâ attitudes toward the website, and how well people understood key terms and features. We also supplemented the empirical data with a guideline review using the Calibrated Evaluatorâs Guide (CEG) where evaluators score the websites using 107 standardized criteria that have been shown to impact the user experience.
Benchmark Study Details
We recruited 199 participants in August 2018 for a retrospective study where we asked current brokerage account holders to reflect on their most recent experiences using their brokerage companyâs website.
Participants in the study answered the 8-item SUPR-Q (including the Net Promoter Score) and questions about their prior experiences.
Quality of the Brokerage Website User Experience: SUPR-Q
The SUPR-Q is a standardized measure of the quality of a websiteâs user experience and is a good way to gauge usersâ attitudes. Itâs based on a rolling database of around 150 websites across dozens of industries, including brokerages.
Scores are percentile ranks and tell you how a website experience ranks relative to the other websites. The SUPR-Q provides an overall score as well as detailed scores for subdimensions of trust, usability, appearance, and loyalty.
The scores for the six brokerage websites are well above average, which isnât too surprising given the financial incentive to make transactions easy and frequent. The average SUPR-Q is at the 92nd percentile (scoring better than 92% of the websites in the database). Merrill Lynch has the lowest SUPR-Q of the group with a score at the 88th percentile. E*Trade and TD Ameritrade lead the group with scores at the 95th percentile. Brokerages as a group also scored higher than the banking websites we recently evaluated.
Usability Scores & Trust
We asked participants to reflect on how easy they thought it was to use and to navigate through their brokerage websites. Fidelity has the highest score in the group (at the 93rd percentile) and Vanguard has the lowest usability score at the 77th percentile.
Not surprisingly, trust scores for this group are also high as users are confident in the companies that handle their money. Participants rated their brokerage companies very high in trust and credibility, and all sites in our study scored between the 92nd and 99th percentiles. This is higher than our group of retail banking websites, which had trust scores between the 76th and 95th percentiles.
Loyalty/Net Promoter Scores
The brokerage websites have an average NPS of about 41%. While weâve seen that users are more reluctant to recommend their banking websites (with an average NPS of 16%), the same was not the case with other financial websites. This relatively high likelihood to recommend could also be from the effects of a long-term bull market and 401k retirement accounts are at all time highs[pdf]. U.S. equities are up about 40% in the prior two years leading up to the data collection period in August 2018. That is, people may be more interested in purchasing stocks, viewing mutual fund balances, and recommending to friends because of positive financial experiences (as opposed to actual differences in website changes).
Use of Brokerage Sites & Mobile Apps
As a part of this benchmark, we asked participants how they accessed their brokerage site and the activities they attempted on their last visit and in the last year. Not surprisingly, most participants use a desktop or laptop computer to access their brokerage site. However, on average, about 35% of participants reported using their mobile devices (about equally between the mobile website or mobile app).
The mobile app is most popular for checking portfolios while the desktop site is used for tasks such as investment research and transferring funds. The brokerage mobile app usage is lower than the banking app usage, which showed significantly more mobile app and mobile website usage (46% and 64%, respectively).
Overall, checking account balances is the top reason to visit the brokerage sites (27%), followed by looking at the recent performance of stocks and investments (24%). The top tasks were the same across desktop and mobile. More details on the mobile app experience are available in the report.
The Learning Curve & Understanding Jargon
Users on some of the brokerage sites (Charles Schwab, E*Trade, and Vanguard) mentioned a steep learning curve at the start of use, especially with complex jargon.
âTo the uncertain investor they have a lot of information that is complex and it can be very overwhelming looking at all of the investing lingo they have on their website. Such as looking at their EFTS, bonds, stocks, 401k, and IRA optionsâthere are a lot of things that you can choose from and getting started is complex.ââVanguard user
âHonestly, my only problems were at the beginning, when I was still learning their siteâs features and layout.ââCharles Schwab user
âI was a little lost and there wasnât a lot of guidance. I had to do a lot of my own research to get started.ââE*Trade user
To further assess how participants understood common jargon, we asked participants to describe three common brokerage terms in their own words. A YTD return was the most widely known term (91%), whereas only 64% of participants knew what a prospectus was, and 73% knew what an ETF was. This comprehension was generally higher than the health insurance websites.
Calling Customer Service
As with banking websites, taking calls is expensive and time consuming. On average, 13% of brokerage website respondents reported contacting customer service in the last year (lower than the 20% on banking websites and 30% for health insurance websites). Being able to accomplish the task without calling customer service was one of the key drivers of UX quality; it explained 8% of the variation in SUPR-Q scores.
Also, like the banking websites, the top reason for brokerage respondents to call customer support is login troubles. However, this only accounted for 5% of participants calling in (compared to 29% for banking and 21% for health insurance). Other common reasons for calling in were researching particular products (stocks, bonds, ETFs) (5%), and transferring funds (3%).
Having an online chat system can help reduce the number of calls to customer support and itâs something our respondents were looking for.
âI wish they had a way to contact customer service other than by phone. I prefer using email or chat to contact customer service so that I can have text to refer back to if Iâm trying to get help with something, but they only seem to allow phone contact.ââVanguard user
In the guideline review (more info can be found in the report), Vanguard has the lowest scores on Help & Information partly because it doesnât have an online chat option.
Product Pages Challenging
While we didnât conduct a usability evaluation for this industry, to identify problem areas, we did conduct a detailed guideline review using the CEG. The top weakness identified by evaluators on both Merrill Lynch and Fidelity was the organization of the product pages.
For Merrill Lynch, the website lacks a clear path to finding investment products as the information is not organized effectively. The page seems to lack an element hierarchy or sidebar to narrow choices into products like ETFs, Retirement Accounts, Stocks, or Mutual Funds.
 âItâs a little hard to navigate unless you know exactly what youâre looking for. Otherwise, it takes time to navigate and look at all of their options.ââMerrill Lynch user
Image 1: Product pages form the Merrill Lynch website shows a lack of hierarchy on the page.
A similar challenge is on the Fidelity website. While it has a better visual arrangement, the vast number of products are hard to differentiate for the typical retail investor as Fidelity has no apparent overall product comparison page to understand and differentiate the financial products.
Image 2: Thereâs no overall investment products page where users can look at products with descriptions from a high-level view.
Image 3: Users must navigate to each individual product page to get any information, which can seem tedious.
The websites that performed well in the product pages category, like E*Trade, have an overview of the investment products with short descriptions and an option to learn more on a detailed product page.
Image 4: E*Trade makes it easy for users to quickly digest the different product options and see where they want to dive deeper.
Key Drivers
To better understand whatâs affecting SUPR-Q scores, we asked respondents to agree/disagree to the following attributes of their brokerage website on a five-point scale from 1=strongly disagree to 5=strongly agree.
My portfolio dashboard is clear and understandable.
I can buy or sell stocks and bonds easily.
The [Brokerage] site is consistent however I use it.
I can find and understand my investment fees and taxes.
I can accomplish what I want to do without calling customer support.
The [Brokerage] site helps me to research and understand investment products.
The [Brokerage] website keeps my information secure.
I can find my account details quickly and securely.
Information Security & Dashboards
The top key driver for the brokerage website user experience is keeping user information safe and secure. âThe brokerage website keeps my information secureâ explained 16% of the variation in SUPR-Q scores.
The second biggest driver of the perception of the website UX quality is having a clear portfolio dashboard.
Clear dashboards explain 12% of the variation in SUPR-Q scores. Checking portfolio balances is also a top task for this group, so itâs not surprising this is found to be so important to the quality of the brokerage website experience.
The six sites we tested all scored high for the item âMy portfolio dashboard is clear and understandableâ (Item 1). The lowest score is on TD Ameritrade where only 88% agreed or strongly agreed compared to the highest for Vanguard and Charles Schwab (both 94%).
Summary
An analysis of the user experience of six brokerage websites found:
1. Brokerage websites offer above average UX across devices. Current users find their brokerage website experience well above average, with SUPR-Q scores falling at the 92nd percentile. Merrill Lynch scored the lowest but still above average at 88% while E*Trade and TD Ameritrade lead the pack at the 95th percentile. Mobile apps are widely used in this industry with 35% of respondents reporting using the app or website within the past year.
2. Users need to overcome a steep learning curve and overwhelming product pages. Respondents on half of the brokerage sites (Charles Schwab, E*Trade, and Vanguard) mentioned a learning curve when first using the website. Many said the brokerage sites didnât lend themselves to novice investors and some jargon was confusing to learn at first. Improved on-boarding and more common language could improve the experience on these sites. The product pages, in particular, may be difficult for less experienced users to differentiate and select.
3. An understandable dashboard and security are key drivers of UX. The top two key drivers of the brokerage website user experience were an understandable portfolio dashboard (12%) and information security (16%). Users want a quick and easy view of their portfolio and they donât want to worry about their personal information being compromised (they probably also want their portfolio balances to increase).
From Joe Biden’s smirks, smiles, laughs, sharp elbows and impolite interruptions to Paul Ryan’s focussed and serious attitude, 2012 VP debate was all fun and informative. Both the sides were trying to be reasonable and focussed. Along the way we got an ocean of hilarious tweets that emerged and made us laugh.
Here is my list of 12 tweets that stood out in yesterdayâs debate:
Patton Oswalt — Ryan is a nervous Walmart manager. Biden is an irate customer with the receipt, the warranty & he’s friends w/ the store owner. #debate
Josh Branson — BREAKING: Post-debate poll has … Biden interrupting the results. #vpdebate
Bill Maher â⪗ Hello 9 1 1? There s an old man beating a child on my tv
GuyEndoreKaiser — Tonight’s debate is scheduled for ninety minutes, but Paul Ryan is already claiming he can finish in fifty something.
Kelly Oxford — While Ryan speaks, Biden looks like he’s trying to order a drink at the bar and the bartender is ignoring him. #vpdebate
Jill Morris â⪖ The VP candidates get to sit because they’re exhausted from standing up for our values. âª#VPDebate
Paul Ryan Gosling â⪖ Hey girl, I’m not taking nervous sips of water, I’m drinking every time Biden laughs at me. âª#vpdebate
Jose Antonio Vargas â⪖ Before this âª#VPDebate, âª@JoeBiden had a Venti macchiato and two cans of Red Bull.
Morgan Murphy â⪖ Biden’s teeth are so white they’re voting for Romney. âª#VPDebate
Indecision â⪖ Watered-down sanctions are the worst. You need four just to get tipsy. âª#vpdebate
James Garrett â⪖ I kind of feel like Joe Biden is Kanye and Paul Ryan is Taylor Swift. âª#VPDebate
National Review â⪖ Wait, is Biden yelling at Martha Raddatz right now? I thought he was debating Paul Ryan…âª#VPDebate
Jeffrey Wisenbaugh â⪖ Biden is yelling less. I think it’s getting closer to his bed time. âª#sleepy âª#VPDebate
Chad Schomber â⪖ To think, all this just to sway 4-6% of undecided voters. And those folks are not watching the âª#VPdebate
Seth Masket â⪖ Actually, this is like if Aaron Sorkin wrote an exchange between the Skipper and Gilligan. âª#vpdebate
Data Have Meaning
We live in a Big Data world in which everything is quantified. While the emphasis of Big Data has been focused on distinguishing the three characteristics of data (the infamous three Vs), we need to be cognizant of the fact that data have meaning. That is, the numbers in your data represent something of interest, an outcome that is important to your business. The meaning of those numbers is about the veracity of your data.
[ DATA SCIENCE Q&A]
Q:What do you think about the idea of injecting noise in your data set to test the sensitivity of your models?
A: * Effect would be similar to regularization: avoid overfitting
* Used to increase robustness
Data is growing faster than ever before and by the year 2020, about 1.7 megabytes of new information will be created every second for every human being on the planet.