The 5 Myths of Predictive Analytics

Posted By Piyanka Jain

Mar
19

In this issue, I am borrowing from my recent keynote at Predictive Analytics World on “the 5 Myths of Predictive Analytics” with this SAS interview and a story board.

Predictive Analytics is used by various organizations to analyze current and historical facts to make predictions about health care, financial collection activities, customer behavior, and customer retention. Although Predictive Analytics is a powerful optimization technique, it is often misunderstood, and thus misused.

The top 5 myths of Predictive Analytics, I am going to address today are:

  1. Predictive Analytics is new: In the recent history, the first credit scoring model was built in 1930′s by Fischer and Durand, but predictive modelling technique goes even further back thousands of years – the use of Indian astrological chart in arranged marriages being one such example.
  2. Perfect Prediction:Often while building the model, it is clear to all that model prediction has a probability associated with it, but upon successful use,  there is often a misplaced sense of perfectness in the scores.

    The next 3 myths below with fun illustration by Matter Solutions.

    Myths of Predictive Analytics - Story Board

  3. Good Tool = Good Model: With tremendous development on the tool front with better GUI as well as higher automation,  people new to this field often mistakenly believe, that good model can be built by pressing the “build regression model” button automatically.
  4. Good Model = Good $: This is one of those highly prevalent myths that even experienced analyst fall for, often finding them frustrated that nobody in the business seems to care for the amazing model they have built. Good model generates the business impact, only when right stakeholders are brought into the analytics process at the right time building proper alignment toward actionability.
  5. Build and Forget: Models gets stale over time and if not maintained, often stops delivering the incremental it started with.

If you are looking to learn more, register for our upcoming workshop - Introduction to Predictive and Business Analytics today! You can also download the  5 myths of predictive analytics whitepaper.

Posted in Analytics Methodology, Analytics Training, Analytics Trends, Predictive Analytics | Tags , , , , , , ,
Leave a comment

Analytics training for Big Data to Big Decisions

Posted By Piyanka Jain

Feb
16

Big data, i.e. huge amount of structured and unstructured data flowing at a high velocity, has created storage and access challenges which cannot be handled by traditional data bases.  Software tools like Pentaho, ParStream and many others with support for Hadoop, give faster and better access to data than ever before. Their strength lies in processing and storing Big data in such a way that data queries and drill downs can return answers in sub seconds.

However, these data tools can neither answer the questions that have not been asked nor can tell which business decision is better when one comes to a decision fork. To get from data to decision, one needs a structured analytics approach. This approach cannot be automated as it involves interactions between stakeholders and business issues.  Fundamentally, the success lies in how well a business professional in the field, able to bridge the gap from business decision to data with appropriate interactions with stakeholders.

Even though the Big data tools have solved many of the collection, storage and access issues, the efficiency of coming to a decision based on data is still far from optimal because of the lack of formal training in analytics.  To address this inefficiency, Aryng has developed an analytics course covering a structured Analytics Framework.  This framework enables quick identification of the right business question behind a decision fork, laying out hypothesis, pulling relevant data from Big data tools, deriving insights using appropriate techniques and using the insights for effective decision making. In essence, Aryng’s analytics training marries technical data analysis skills with decision making soft skills in a capsule size analytics class appropriate for any professional with access to data and a need to make better decision.

Posted in Analytics Methodology, Analytics Training, Big Data | Tags , , , , , ,
1 Comment

Data doesn’t speak, it responds

Posted By Piyanka Jain

Feb
7

Data doesn’t speak, it responds. A journey of 1000 miles starts with a single step. The journey from data to decisions is sometimes a long one, funny thing is, it does not start from data… it starts way before, it starts from understanding what you may want from the data; what it is that you want to drive from the data.Real Business Question

Often people think data speaks, it may, more often than not, it doesn’t. Data responds. It responds to questions. Questions we ask which are relevant for us to drive our business. Questions we have arrived based on our strategic priorities and goals. So the first step from data to decisions is identifying the “real Business Question”.

What are the attributes of the real business question? How would you know you have the real business questions? Well, does your question pass this test..

  1. Is the answer to the question, actionable?
  2. Would different answers to that question lead to different actions/decisions for your business?
  3. Has something triggered that question?
  4. Would answering that question make any difference to your business?

So before you start torturing the data to speak, identify the real “Business Question” to ask the data.

Posted in Analytics Methodology, Analytics Training, Business Analytics | Tags , , ,
1 Comment

What is the cost of a bad business decision?

Posted By Piyanka Jain

Jan
30

New York Times reported yesterday that BlackBerry is under siege in Europe where it is drastically loosing share of enterprise market to IPhone and Android. It now has about 1/4th of the market where it used to pretty much own the whole market.

BlackBerry’s maker, Research in Motions’ (RIM) answer to all of its trouble is BlackBerry 10 OS with even stronger focus on security which RIM believes is its edge. However, it may be fast losing its edge on security to companies like Samsung who are now releasing products like Galaxy with extra security features for enterprise customer. To top that, RIM focused so hard on security that it neglected other enterprise features like compatibility with Microsoft exchange server which is the messaging system of choice in most companies. IPhone integrated with Microsoft exchange server back in 2007 whereas Blackberry’s integration is still in beta.

I have been using BlackBerry for years.  My top qualms with BlackBerry are its Microsoft exchange server integration, bad app integration and almost impossible browsing experience (slow!). I still choose to use it because of its qwerty keyboard, great battery life and very reliable messaging service, however many have switched.  What has all this got to do with bad decisions and analytics?

RIM made some bad choices in product focus and feature release over last few years, which is very evident from its loss of enterprise market share in North America and Europe. My only question is: could this have been avoided? A quick Google search on historical BlackBerry issues supports my hypothesis that it possibly could have been. Had the product folks at RIM been listening to the voice of the customer through – internal data, internal and syndicated surveys, social chatter – using effective analytics framework to understand what the customer (not only the CIO’s but also the end users) really wanted, they may have spread their bet on security to improve Microsoft exchange integration and other much needed innovation in their products. Quick look at last year’s earning alone suggests these bad decisions have had impact in $100’s of millions.

On the other hand, I have seen smarter data-driven decisions with direct positive impact to the top line. More on that next week.

Meanwhile, what does a bad decision cost in your organization? To learn smarter and better decision making using data married to gut, register for one of our upcoming “Data to Decisions”™ boot camp today!

Posted in Analytics Methodology, Analytics Training, Stategies, trends | Tags , , , , , , , , ,
2 Comments

Why Analytics Fail

Posted By Piyanka Jain

Jan
24

Why organizations fail to leverage Analytics and our solve!

Thanks to all of you who filled our survey last week. This week we decided to share “The top 5 reasons why Analytics fails in an Organization” using this video. This video summarizes the survey results and my own experience comparing organization who compete on analytics and others who fail to leverage the immense data and resource at their disposal.

Our San Francisco workshop is filling fast, so register today! Early Bird ends this friday…

Posted in Analytics Training, Business Analytics, Predictive Analytics | Tags , , , , ,
3 Comments

Predictive Analytics: 4 Key Constraints

Posted By Piyanka Jain

Jan
10

Predictive Analytics is a powerful statistical technique by which data from past behavior is analyzed to make predictions about future. With this technique, one has the ability to make predictions at the individual customer or event level.

Financial services industry has widely adopted predictive analytics techniques to optimize marketing and fraud detection over the past few decades. Last five years has seen a major adoption of this technique across many more industries. This adoption has been largely driven by increased competitive pressure and, improvements in the systems and tools.   Eric Siegel, Predictive Analytics Expert and, Predictive Analytics World Conference Chair, in his article Predictive Analytics Delivers Value Across Business Applications discusses common business application of Predictive Analytics in  Direct Marketing, Advertising, Product Recommendation, Customer Retention and, Insurance Pricing, to deliver millions in incremental savings and revenue.Predictive Model

In my experience, I have also found it to be a very powerful optimization technique, but it has its limitations. There are four key constraints to consider when investing in predictive analytics.

  1.  Do you have appropriate historical data?
    1. Predictive models are trained on historical data and needs relatively large data set in order to separate signal from noise. So you will need data going back in time or an ability to re-create data at different time periods.
    2. Also you need to consider the minimum occurrence of the behavior you are trying to predict. For example, to weed out spam traps from a targeted email list, you can build a predictive model on past email attributes to identify spam traps. However if spam traps are less than 0.1% of your  past e-mail database, it might get really challenging for the model to pick up signals to identify emails most likely to be a spam trap.
  2. Does the model ROI make sense?Predictive modeling is not only resource and time intensive during the initial build stage, but also need frequent maintenance as these models deteriorate over time. So prior to initiating a project, carefully balance the return from the model with the investment of people and time.
  3. What is the required prediction accuracy? Predictive model can have good prediction accuracy, however it is not perfect. So in the case where you can’t afford any inaccuracy, like in medical diagnosis, predictive model alone is not sufficient. On the other hand, when accuracy requirements are much lower, then simple logic or business analytics techniques like Correlation, Profiling or RFM (Recency Frequency Monetary) models might be sufficient. These simple methods are quicker and cheaper; and if sufficient, can give a better ROI.
  4. Do you have organizational buy-in? Lastly, it’s very important to have stakeholder buy-in before you invest heavily in predictive analytics. Predictive modeling is not a widely understood science yet, and analysts often have a hard time articulating the findings from the model. Some techniques like Neural Networks can have very high accuracy but often impossible to explain. This can lead to people questioning the efficacy of model and, possibly ignoring the model in decision making.

So before starting a new and expensive predictive modeling exercise, make sure you have historical data, the ROI is justifiable, you really do need a predictive model as any simpler logic won’t suffice and that the people around are ready to take action on the model score/recommendation!

For a complete introduction to Predictive Analytics, attend one of our upcoming 1-day workshops on Predictive Analytics.

Piyanka Jain

http://www.Aryng.com | Follow @AnalyticsQueen | Sign up for our Newsletter

P.S: Our February workshop is filling fast, so make sure to reserve your spot today! 1.866.604.3092

Posted in Article, Predictive Analytics, Predictive Modeling | Tags , , , , , , , , , , , ,
Leave a comment

3 Key Analytics strategies for 2012

Posted By Piyanka Jain

Jan
3

Wish you all a wonderful 2012! In this issue I want to share my thoughts on key strategies to compete on analytics for the coming year.

1. Keep it simple: 2011 saw “Big Data” as a “big” trend with new technologies to access and render large mixed data sets, real time, from multiple sources. But, industry jargon aside, the reality of analytics is that, all such “Big data” still needs to be processed and structured the traditional pre-”Big Data” way in order to be turned into insights and actions. In my experience, companies who are competing on analytics, like Zynga, are doing so by using simpler “Business Analytics” and Test-and-Learn on their core “structured/columnar data”. In a recent Forbes article on Zynga IPO, Kord Campbell, CEO of Loggly talks about how Zynga competes on analytics using time series event analysis. This business analytics technique helps Zynga keep users engaged for hours and thus better monetize their user base,

Because analytics on big data is resource and time intensive and because there is still so much untapped in the structured data, the biggest bang for buck still lies in using simple business analytics techniques like correlation and trend analysis on core – structured data.

2012 Analytics strategies

2. Invest in talent development: A Study by Accenture published a few weeks ago, found the top in-demand skill gap for US Workers as problem solving, analytical skills and managerial skills. 55% of the workers report to be under pressure to develop additional skills however only 21% are reported to have acquired it through company-provided formal training. That’s a huge gap. For organizations who want to compete on analytics, the need to train is even more urgent as these skills were not taught in school (at least not when most of us got our formal training). An average analytics project with well-defined scope and quick access to structured data should not take more than 2 weeks. If it takes longer in your organization or if you are not seeing these insights being used in decision making, it’s time to invest in analytics training.

3. Keep a pulse on new “splunky” data tools: Unlike yesteryears, when tools like Oracle database and it’s SQL interface would come in market and stay as the only solution for a while, today technologies are rapidly developing and transforming the market, with lots of platform based hosted solutions, in cloud and often inexpensive or free, yes FREE! 2012 will see even a bigger momentum towards tools like Splunk, Kapow’s Katalyst etc with almost “no” learning curve, accessible to business users, quite often “plug and play” or hosted , and often free. And remember “Google Analytics” core tool is still FREE. So keep a pulse on the new cool and inexpensive tools in the market which your competition may be leveraging to get faster and better product/service to your customers!

Have a great start to your new year and see you next week!

Posted in Analytics Trends, Article, Predictive Analytics, Stategies | Tags , , , , , , , , , ,
Leave a comment

Approach to Analytics: Holmes VS. Columbus

Posted By Piyanka Jain

Dec
27

Let’s face it. Data Ocean is getting deeper and wider. Earlier this month, New York Times published an article on data deluge in DNA sequencing. BGI, world’s largest genomics research institute generates so much data, that it can’t send it to clients over internet as that would take weeks. So they send computer disks containing the data, via FedEx!

The article highlights that data handling is a key constraint now. Data quickly becomes dated or runs into an overwrite issue so making sense of data “quickly” is paramount. Hard to believe but “It costs more to analyze a genome than to sequence a genome.” The article goes on to quote C. Titus Brown, a bioinformatic specialist at Michigan State University. “It’s not at all clear what you do with that data… Doing a comprehensive analysis of it is essentially impossible at the moment.”

What is the solution? In addition to large scale distributed storage system, there needs to be concerted effort towards disciplined approach to analytics. The typical “Christopher Columbus” mentality of sailing into data and hoping to find insights, isn’t going to work anymore. The bio-informatics analyst will do much better by going the “Sherlock Holmes” way. Let me explain.

Imagine you are given a task of finding lost treasures from the Pacific Ocean. Pacific is deep and huge; how would you approach finding the lost treasure/gold. Would you rent a submarine and start cruising (Explorer)? Or would you look for clues to where the gold could be and use that as guide to direct your attention and effort (Detective).

analytics as explorer failsExplorer: If you go the explorer route, you are guaranteed the views: nice corals, beautiful wildlife, emerald-green waters, but the chances of you finding gold is slim. Essentially you are not directing your effort of finding gold. Your actions of exploration are independent of what you are looking for, you could have been tasked with looking for killer whales and you would still take the same action.

Detective: If on the other hand, you choose to be a detective, looking at historical ship routes and ship wrecks, using depth as a way to eliminate, you can start identifying potential areas to explore. Once you have identified a dozen plus potential location, you can prioritize the top 3 areas and then use submarine or deep sea diver to go explore. You will likely find gold and in much shorter time. You will either succeed fast or you will fail faster and then re-strategize to attack the problem again.Analyst with Holmes approach are successful

Good analysts are detectives using Business/corporate goals and priorities as guides to direct their effort to go looking for answers which are relevant to business and thereby finding gold nuggets and driving $ impact in the organization. So which one are you?

Enjoy the rest of your Holidays! And see you in 2012!

Posted in Analytics Methodology, Article, Stategies | Tags , , , , , , , , ,
Leave a comment

Analytics at it’s best: Top guided, bottoms up

Posted By Piyanka Jain

Dec
20

In a recent article on searchCIO, John Lucker, a principal at Deloitte Consulting LLP, lays out the fundamentals behind a fact-driven organization vs. instinct driven organization, and strategies to break down the barrier towards successfully utilizing analytics in decision making.

In my experience, analytics is best utilized at organizations where the leadership is asking questions of their data, leading to a fact-driven culture. So analytics is fundamentally driven from the top. Data has a lot of answers hidden inside it, but the microscope that magnifies the answers is guided by the strategic priorities and business questions coming from the top.

So before the analysts in your organization go scuba diving, deep into the data ocean again or start work on yet another dashboard, I recommend, thinking through

  • What is the real business question behind that dashboard request?
  • What are the fundamental drivers of the business?
  • What information does the organization need today to make better decision for the future

And once a well formulated business question is identified, then proceed with looking into relevant data to answer the business question. I recently talked more about this “Data to Decisions”™ framework with example at the Google Analytics User Conference. See the keynote video” $120k incremental revenue from 2 hours of analysis”

Happy Holidays! And see you next week.

Posted in Analytics Methodology, Article, Business Analytics, Stategies | Tags , , , , ,
Leave a comment