Presentation
Introduction
Predictive Analytics is translating data into information that is necessary for business owners to make informed decisions and investments. It is the difference between running business on a hunch or intuition versus looking at collected data and revealing what’s to come before it actually happens. It leads to better decision making by looking for patterns and trends in the data and by being able to forecast what’s going to happen.
Predictive analytics can serve throughout the whole company and all C-level executives can take an advantage of it. For example Chief Operating Officers (COOs) can get better insight into supply chains and operations and enhance efficiency.
Becoming an analytics-driven organization helps companies to extract insights from their enterprise data and help them to achieve costs reduction, revenues increase and competitiveness improvement.
Predictive analytics plays crucial role in today’s business. The goal of this research is to theoretically describe predictive analytics and demonstrate its power. The research will consist of three parts. First part will briefly describe predictive analytics, how it can be used in association to supply chains and operations, and why does it matter. Second part will then introduce statistical techniques that stand behind. Last part of the research will demonstrate companies that succeeded in doing their business better by using predictive analytics.
Literature Review
In today’s organizations, the most important assets are no longer the big machines and equipment, but the intangibles assets instead. Enterprise data, for example, is a priceless strategic asset a company owns as it represents the aggregate experience and history of its interaction with customers from which to learn. [1]
Organizations generate and collect large amounts of data in many forms and from many sources. Predictive analytics extracts relevant information from this data and uses it in order to predict with confidence what will happen next, so smarter decisions are made and business outcomes improved. [2]
Analytics, also known as “business analytics” or “data analysis”, refers to the software and methods that organizations use to understand data. Organizations use predictive analytics, and other kinds of analytics software, to gain insights into their financial and operational performance and from their customer behaviors, as to make accurate predictions and better-informed decisions about emerging opportunities, competitive threats and shifts in their markets to increase competitive advantage. [3]
Applying predictive analytics across business functions helps companies achieve multiple strategic objectives. Predictive Analytics it is very important in today´s business, the following strategic objectives can be attained to their full potential only by employing it:
- Compete: Secure the Most Powerful and Unique Competitive Stronghold
- Grow: Increase Sales and Retain Customers Competitively
- Enforce: Maintain Business Integrity by Managing Fraud
- Improve: Advance Your Core Business Capacity Competitively
- Satisfy: Meet Today´s Escalating Customer Expectations
- Learn: Employ Today´s Most Advanced Analytics
- Act: Render Business Intelligence and Analytics Truly Actionable
As shown in the next Figure, predictive models generated from enterprise data are integrated with business units across the organization, including marketing, sales, fraud detection, call center and core business capacity. The circle digits indicate where each strategic objective is attained. [1]
Figure 1 Predictive Analytics [4]Studies show that organizations that apply analytics outperform their peers. Further, those with a broad-based, analytics-driven culture perform, on average, three times better. Not only do they drive more top-line growth and control costs, they take timely corrective action to reduce risks that derail their plans. [3]
Statistics Methodology
Forecasting systems offer a variety of techniques, and no one of them is best for all items and situations. The forecaster’s objective is to develop a useful forecasted from the information at hand with the technique that is appropriate for the different patterns of demands. Two general types of forecasting techniques are used: judgments methods and quantitative methods. [5]
Judgments methods
1) Judgment: Include the opinions of managers, expert opinions, consumer surveys, and sales force estimates into quantitative estimates. Some constrain that you can face when using judgment method:
- History file does not exist when a new product is introduced or technology is predictable to change.
- History file exist but if certain events like rollouts or special packages are reflected in the data it would be less useful. However in some cases judgment is the only practical way to make forecast.
The four most effective methods of judgments are:
- Sales force estimates: Are forecast estimated form estimates made periodically by members of a company’s sales force. Nevertheless some people are more optimistic than others when making a forecast causing some biases.
- Executive opinion: Opinions, experience and technical knowledge of one or more managers or customers are summarized to arrive at a single forecast.
- Market research: Is a systematic approach to determine external consumer interest in a service or product by creating and testing hypotheses through data-gathering surveys. But it typically includes numerous qualifications and hedges in the findings.
- Delphi method: Is a process of gaining consensus from a group of experts while maintaining their anonymity. This form of forecasting is useful when no historical data are available and when managers inside the company have no experience.
Quantitative methods
2) Causal: Use historical data on independent variables such as promotional campaigns, economic conditions, and competitor’s actions, to predict demand.
It can be used when data information is available and the relationship between the factor to be forecasted and other external or internal factors. This relationship is expressed in a linear regression, one variable, called independent variable is related to one or more independent variables by linear equation.
Y= a + bX
Y= dependent variable (such a demand for door hinges)
X= independent variable (such as advertising expenditures and new housing arts)
a= Y- intercept of the line
b= slope of the line
The objective of linear regression analysis is to find values of a and b that minimize the sum of the squared deviations of the actual data points from the graph line. Three measures are commonly reported:
- 1-The sample correlation coefficient r: That measures the direction and the strength of the relationship among the independent and the dependent variable this value range from -1 to + 1. A correlation coefficient of +1 means that (increases or decreases) of the independent variable are always accompanied by changes in the same direction by the dependent variable. An r= -1 decreases in the independent variable are always accompanied by increases in the dependent variable and vice versa.
- 2-The sample coefficient of determination measures: Is the square of r. The r2 ranges from 0 t0 1 the r2 close to 1 mean a close fit.
- 3-The standard error of estimate Sxy: Measures how closely the data on the dependent variable cluster around the regression line.
Statistics methods are used for forecasting because unlike the others mention above this kind of method takes into account trends and seasonal patterns so the predictions are more successful.
3) Time series-method: Is a statistical approach that relies heavily on historical demand data to project the future size of demand and recognize trends and seasonal patterns.
The simplest time series method is the Naïve Forecast, whereby the forecast for the next period (Ft+1) equals the demand for the current period (D1). This method might be adapted to take into account a demand trend but it works best when the horizontal, trend, or seasonal patterns are stable and random variation is small
The statistical techniques that do have an adaptive quality in estimating the average in a time series are:
1) Simple moving average: Is the average demand for the n most recent time periods using it as the forecast for future time periods. After demand is known, the oldest demand from the previous average is replaced with the most recent demand and the average is calculated with the following formula:
Ft+1=Sum of last n demands/n=Dt+Dt-1+Dt-2+Dt…+Dt-n+1/n
Dt= actual demand in period t
n= total number of periods in the average
Ft+1= forecast for period t+1
With this methodology large values of n should be used for demand series that are stable, and small values of n should be used for those that are susceptible to changes in the underlying average
2) Weighted moving averages: In the simple moving average each demand has the same weight in the average-namely, 1/n. In this method, each historical demand in the average can have its own weight. The sum of the weight equals to 1.0.
Ft+1= 0.50D1+0.30Dt-1+0.20Dt-2
The forecast with the weighted moving average is more responsive to change in the underlying average of the demand series than the simple moving average.
3) Exponential smoothing
This methodology does not require n periods only requires the next three items of data:
- The last period’s forecast
- The actual demand for this period
- A smoothing parameter alpha α which has a value between 0-1 larger α emphasize recent levels of demand. In practice, various values of α are tried and the one producing the best forecast is chosen.
The equation is the following:
Ft+1= α Dt +(1- α)Ft
Because exponential smoothing is simple and requires minimal data, it is inexpensive an attractive to firms that make thousands of forecasts for each time period. But has a disadvantage when underlying average is changing, as in the case of demand series with a trend.
Trend projection using regression is a hybrid between a time-series technique and the causal method.
A trend in time series is a systematical increase or decrease in the average of the series over time. Where a significant trend is present forecast from naïve, moving average, and exponential smoothing approaches are adaptive, but still lag behind actual demand and tend to be below or above the actual demand. Trend projection with regression is a forecasting model that accounts for a trend with simple regression analysis To develop a regression model for forecasting the trend, let the dependent variable Y, be a period’s demands and the independent variable, t, be the time period. For the first period let t=1; for the second period, let t=2; and so on. The regression equation is:
Ft=a+bt
One advantage of the trend projection with regression model is that it can forecast demand well into the future. The previous models project demand just one period ahead, and assume that demand beyond that will remain at that same level.
Practical Use
In this part we demonstrate how predictive analytics software can help companies to get valuable predictions from their data. We namely focus on IBM SPSS Statistics. The reason why we have chosen this tool is mainly because we think that IBM is undoubtedly the leader in analytics field. Since 2005 IBM has invested $16 billion in business analytics acquisitions, it has the world’s largest math department in private industry, employs more than 10,000 technical professionals and has come up with 500 analytics-related patents every year for the last two years. [6]
IBM SPSS Predictive Analytics Solution
Figure 2 IBM SPSS Predictive Analytics solution [7]IBM SPSS is predictive analytics software that helps organizations to translate data into information that is necessary for smart decisions. It helps predict the outcomes of interactions before they occur and act on their insights by embedding analytics results into business processes. IBM SPSS product portfolio consists of four product families: [8]
- Data Collection family – Set of tools designed for market and survey research and enterprise feedback management. Helps to develop a deeper understanding of people’s attitudes, opinions and preferences.
Figure 3 IBM SPSS Text Analytics for Surveys [9]- Statistics family – Set of tools that’s ability is to analyze information and deliver comprehensive results. This is the most widely used suite of statistical software in the world. It puts the power of advanced statistical analysis in your hands to understand data, identify trends and produce accurate forecasts.
Figure 4 IBM SPSS Advanced Statistics [10]- Modeling Family – With these tools you can discover hidden relationships in your data and anticipate the outcomes of future interactions. This family provides an ability to build reliable models. It allows you to discover patterns and trends in your structured and unstructured data with an intuitive visual interface supported by advanced analytics.
Figure 5 IBM SPSS Modeler [11]- Deployment family – This set of tools allows organization to integrate analytical results into their operations. This bridges the gap between analysis and action.
Case Studies
In this part we mention some areas where IBM SPSS Predictive Analytics solution can be used. We also demonstrate how some organization have achieved tremendous outcomes by using this solution.
Crime Prediction and Prevention
Law enforcement can use predictive analytics to help fight crime. Everyone agrees that there’s no shortage of data available, but making sense of it, turning it into usable information and recommendations you can act on are what’s needed most.
To predict crime we can work with data collected by jurisdictions, data from warrants, arrests, gang data, or incident reports. We can also go further, we can analyze data from non-traditional sources like chat room conversations, emails, blogs, weather forecasts, local event schedules and many other. Then patterns and trends can be uncovered.
By analyzing such large datasets, agencies get real-time decisions-making guidance and are able to position resources in the locations most likely to encounter problems before crimes are committed.
The Memphis Police Department is an example of a law enforcement agency using predictive analytics in predicting crimes. This has led to a 27% reduction in serious crime overall and a 15% drop in violent crimes. Moreover, by using predictive analytics, the Felony Assault Unit now closes four times more cases – their closure rate has increased from 16% to 70%. Also, the department achieved overall improvement in the ability to allocate police resource in a budget-constrained fiscal environment. [12]
Predictive Customer Analytics
Predictive customer analytics uses business information to accurately predict what today’s — and tomorrow’s — customer wants. It can help you improve marketing campaign results, shorten sales cycles, improve customer satisfaction, produce the right products, and align operations more closely with the rest of the business.
It can help you more effectively attract new customers and grow your existing customer base with targeted offers. Business intelligence combined with predictive capabilities help segment current customer demographic attributes to visualize how customers have historically responded to earlier offers. [13]
First Tennessee Bank [14] uses predictive analytics to optimize their marketing resources. For banks today, having more ways to communicate with customers is a good thing. But it has also made it harder to figure out where and how to most profitable use their marketing resources. First Tennessee Bank is combining a granular customer data and optimize its marketing spend, focusing on programs that deliver the highest ROI. The bank has created a predictive model that generates a quantitative measure of the expected profitability of a given product offered to a specific subset of its customers.
First Tennessee Bank has achieved 600% overall return on its investment through more efficiently allocated marketing resources, 3.1% increase in marketing response rate through more accurate targeting offers and 20% reduction in mailing costs and 17% reduction in printing costs due to the ability to target the most attractive segment for specific offers.
Fraud Detection
Organizations face a multitude of threats every day, whether it’s customer fraud, credit risk, debt management, insider trading, cyber attacks or even safety and security issues. These threats can cost organizations millions – even billions – in losses.
Predictive analytics allows you to analyze the billions of bits of data you’re already acquiring – looking through transactions, demographics, surveys – and even unstructured text like email and social media. You can then identify patterns, relationships or anomalies in the data. They then help you to accurately predict the threats your organization was going to face and determine the best actions to take. Imagine that you are able to mitigate those risks before they happen.
One hospital decreased bad debt write-offs by 30 percent and reduced the size of its collections staff by 50 percent. An insurance company generated a 403 percent return on investment by doubling the accuracy of fraudulent claim identification and fast-tracking legitimate claims. A state agency stopped fraud, prior to payment, and saved on average 220 million dollars annually. [15]
Predictive Maintenance for Manufacturing
The challenge for manufacturing has always been to produce high-quality goods while optimizing resources at every step of the production process. Today forward-looking manufacturers rely on predictive maintenance to go beyond preventative and regularly scheduled maintenance, ensure production quality and maximize value at every step of the process. Predictive analytics is helping manufacturers reach new standards of quality, and save money, by minimizing downtime from unscheduled maintenance, practically eliminating unnecessary maintenance and providing superior forecasting.
Predictive analytics software can gather information in real time from a variety of sources, including maintenance logs, performance logs, monitoring data, inspection reports, environmental data and even financial data. It then proactively directs resources towards those areas before risk becomes reality. This early identification of maintenance requirements and operational issues is critical for preventing production interruptions, improving usability and service levels for customers.
A large equipment manufacturer saved $1 million in just two weeks by using preventive maintenance to proactively identify problems and take action before failure occurred. By minimizing downtime and repair costs across all its manufacturing operations, the manufacturer achieved a 1400% return on investment in just four months. [16]
Conclusion
There are many other areas where predictive analytics can increase revenue, reduce costs and improve competitiveness. Becoming an analytics-driven organization helps companies to extract insights from their enterprise data.
Predictive analytics helps you to make better, faster decisions and automate processes. It helps you address the questions and ensure you to stay one step ahead your competition. There is no doubt that business analytics together with predictive analytics is one of the top priorities for CIOs and it’s crucial in today’s business.
Authors
- Matous Havlena
- Maricela Gomez Herrera
- Maria Isabel Hernandez Eguibar
Bibliogprahy
| [1] | Prediction Impact, Inc., “Seven Reasons You Need Predective Analytics Today,” 2010. [Online]. Available: http://public.dhe.ibm.com/common/ssi/ecm/en/ytw03080usen/YTW03080USEN.PDF. [Accessed 20 March 2013]. |
| [2] | IBM, “Predictive analytics,” [Online]. Available: http://www-142.ibm.com/software/products/us/en/subcategory/SWQ50. [Accessed 26 March 2013]. |
| [3] | IBM, “Business Analytics from IBM – What is Analytics?,” [Online]. Available: http://www-01.ibm.com/software/analytics/what-is-analytics/. [Accessed 28 March 2013]. |
| [4] | Data Mining World, “Predictive Analytics,” [Online]. Available: http://dataminingworld.wordpress.com/2010/09/26/predictive-analytics/. [Accessed 20 March 2013]. |
| [5] | L. J. Krajewski, L. P. Ritzman and K. M. Malhotra, Operations Managements: Processes and Supply Chains, 2013. |
| [6] | IBM, “Business Analytics from IBM,” [Online]. Available: http://www-01.ibm.com/software/analytics/what-is-analytics/. [Accessed 18 March 2013]. |
| [7] | [Online]. Available: http://www.spss.com.hk/statistics/images/statistics-18.png. |
| [8] | IBM, “Get to Know the IBM,” 2011. [Online]. Available: http://public.dhe.ibm.com/common/ssi/ecm/en/ytb03015usen/YTB03015USEN.PDF. [Accessed 25 March 2013]. |
| [9] | [Online]. Available: http://www.lewiswire.com/de/uploaded/SPSS/4_stafs_3.0_new.jpg. |
| [10] | [Online]. Available: http://www.ibm.com/developerworks/library/ba-direct-marketing-spss/fig03.jpg. |
| [11] | [Online]. Available: http://www.spss.com.hk/images/software/modeling/1ModelVisualization.jpg. |
| [12] | IBM, “IBM SPSS Solutions for Crime Prediction and Prevention,” [Online]. Available: http://ibmtvdemo.edgesuite.net/software/analytics/spss/demos/crime-prediction/. [Accessed 25 March 2013]. |
| [13] | IBM, “IBM SPSS Predictive Customer Analytics,” [Online]. Available: http://ibmtvdemo.edgesuite.net/software/analytics/spss/videos/predictive-customer-analytics/. [Accessed 28 March 2013]. |
| [14] | IBM, “First Tennessee Bank: Analytics drives higher ROI from marketing programs,” 2011. [Online]. Available: http://www.ibm.com/smarterplanet/us/en/leadership/firsttenbank/assets/pdf/IBM-firstTennBank.pdf. [Accessed 25 March 2013]. |
| [15] | IBM, “IBM SPSS Solutions for Fraud Detection and Prevention,” [Online]. Available: http://ibmtvdemo.edgesuite.net/software/analytics/spss/videos/threat-and-fraud/. [Accessed 28 March 2013]. |
| [16] | IBM, “Predictive Maintenance for Manufacturing,” 2011. [Online]. Available: http://public.dhe.ibm.com/common/ssi/ecm/en/ytw03189usen/YTW03189USEN.PDF. [Accessed 29 March 2013]. |