A. Introduction
Dean Felix E. Asprer, outgoing PCDEB president, Dean Nelson S. Abeleda, incoming PCDEB president, CHED Representative Dr. Conrad Inigo, the PCDEB staff, fellow educators and industry practitioners, ladies and gentlemen, good morning.
I can think of two reasons why I am here as your keynote speaker. First, as the 2010 President of the Philippine Marketing Association, a non-profit non-stock organization. PMA fully supports PCDEB and the CHED-industry linkage.
Second, because of my involvement in predictive analytics not the least of which is my association with IBM-SPSS.
I got into this advocacy mainly because of my experience in performance management in the private sector. Having been through this initiative for some time and having been in an educator myself, I have always maintained that there must be a stronger and deliberate effort to promote the principles of governance in the academic sector.
Similar initiatives in several sectors yielded very positive results. These include the banking and healthcare sectors (hospitals in particular).
As a result, the banking sector has demonstrated resiliency and strength even in the face of the global crisis that affected many foreign banks.
Hospitals that have institutionalized the principles of governance are now very profitable as compared to their situation some ten years back and are offering healthcare services that are comparable to the best in the world.
B. Discussion Coverage
I will focus on two elements of research and try to give examples or approaches that are being used by researchers and decision makers around the world. These are data collection, and analyses for insighting including deployment of research findings.
Let me start with a case that presently dominates our free time, air space, advertising space and much of the time and resources of Filipinos. The presidential election. On the screen is a statistical estimate of the percentage of voters who remain undecided as of our latest survey.
In October 2009, about 40% of voters were yet undecided as to who among the presidential candidates they will vote for. Undecided, as a concept, includes those who did not have a choice, those who had a choice but were not definite, and those who claim they were definite but were open to switching depending on whether the endorsement remains the same or the platform of other candidates is more attractive.
In December 2009, the percentage of undecided went down to 35%. In our February 2010 survey, the percentage of undecided went down to 27% which was expected. I anticipate that this ratio will go down to around 10% towards the end of April 2010. What is the significance of these figures?
There are three things to consider. First, election is often decided by the undecided particularly in a tight race. My gut tells me that in this year’s election, for the President, that will be the case. Limiting election expenditures to communicate to the undecided will keep election spending to a level that is more realistic and optimal. Overspending stems from reckless targeting.
Second, with data on the undecided, it is possible to create a predictive model that will tell us who those undecided voters are –their profiles including age, gender, income, and concerns – and where they are located. This model may then be deployed in an existing database. Once tagged, a subsidiary database may then be created. This subsidiary list will show where these undecided voters are. Limited resources can then be allocated with maximum effect.
Third, it becomes a guide for a cost-effective campaign management. While it is not a guarantee for success, it improves that chance and sets a limit to election spending. Why spend money to cover those who have already decided? Why not spend that money on those who are still open to persuasion?
This video clip shows how this approach helped the Obama team reach out to those who mattered most during the U.S. elections – the undecided voters.
C. Data Collection
I recently attended a seminar organized by ESOMAR (of which I am a member) on the social media and how they are applied in research. The transition is rather spectacular – from a scenario where the researcher is in full control of all elements of the research to a scenario where the respondents become co-creators of the research output.
What is remarkable in this new approach, called Web 2.0, is how qualitative research has transitioned from a live discussion to an online panel. What is even remarkable is how it is able to replicate a social community online. The result is a very rich collection of insights that could never be generated in a traditional panel or FGD.
Imagine a branded online social community composed of around 30 to 50 diet pill users and would be diet pill users. Imagine further that these community members share their concerns, fears, experiences, successes, and new information online, not just one time but as desired or as many times over a period of three to six months. Imagine some more that in this group of 50 diet pill users and would-be users, anyone may initiate a discussion or present an issue for discussion.
I tried to promote this new approach to my clients. It is an uncharted territory and there is hesitation among the local clients to venture into this methodology.
Data collection is certainly evolving. It is moving towards a multi-mode or the use of multiple data collection channels for one survey. However, users and researchers are quite skeptical about the inclusion of the non-traditional channels for data collection and the combination of various channels for one survey.
In the 2008 SPSS Directions that was held in Las Vegas, there were several breakout sessions and presentations on the multi-modal research. Multi-modal is a “survey that is administered in multiple research modes, for example, web-based and phone-based, or web-based and paper-based.”
Proponents of the multi-modal approach believe that “A low cooperation or response rate does more damage in rendering a survey’s results questionable than a small sample because there may be no valid way scientifically of inferring the characteristics of the population represented by the non-respondents.”
There are three types of multi-modal approaches, each one with a different level of difficulty. One type is the multi-mode where one channel is used for one sample segment in a survey. A second type is the sequential mixed mode where different modes are used to collect data from different types of respondent. The third type is the parallel mixed mode where the respondent is able to choose the preferred mode and is able to switch from one mode to another.
For those who have been using it, results have been spectacular: Increased response rates, reduced non-response error, improved sample coverage, respondent centric, and reduced cost.
We have tried using this multi-mode – CAPI, CATI and FTF. There are indeed challenges. But those challenges are surmountable. The challenges are related more to factors such as availability of a sample frame for channels that require it, and data merging.
I have presented this to industry and clients remain skeptical about it. My sense is that for this to become an acceptable approach for data collection, it has to start in the classroom. Future decision makers (the students today) will have to be exposed to these new systems and be given the opportunity to experience and personally evaluate this approach.
D. Analytics
When I did my MBA and took the subject of marketing research, I learned that the objective of research is to provide information, not to make a decision. It was the belief that decision making was a function of many other variables. Similarly, the objective then of advertising was to communicate a message, not to sell. It was also the belief then that selling was a function of many other variables.
Arguably, that scenario has changed. In advertising, they talk about success fee, no longer just commission and production cost. In research, they are now talking about empowering frontline personnel to make decisions based on the output of research.
This video clip (crime watch) shows how investigators or police officers have been empowered to make decisions using research output.
This type of a research falls under the umbrella of data mining or predictive analytics. Predictive analysis helps connect data to effective action by drawing reliable conclusions about current conditions and future events.
Data mining is the process of discovering meaningful new correlation, patterns and trends by sifting through large amounts of data stored in repositories, and by using pattern recognition technologies, as well as statistical and mathematical techniques.
There are two types of data mining – supervised and unsupervised.
Data mining or predictive analytics is very useful for managing risk and for institutional research. In industry, there are applications of data mining in marketing, risk management, production management and supply chain management.
Popular applications of this methodology are the cross-sell and the upsell. You will notice that bank branches are now more aggressive in selling multiple products such as car loans, housing loans, pension funds, and other insurance products. What they are doing is actually cross selling. The starter product is the deposit account. The cross-selling is done on other bank-owned products.
But cross-selling to whoever walks into branch or to every client in the database is usually not very productive. It is like finding a needle in a haystack. By using data mining and the appropriate analytic technique, we can create a model that will tell us who among the clients in the database are more likely to respond to a cross-sell initiative. Each prospect is tagged and becomes a target for cross-sell campaign.
The same principle applies for up-selling. Other popular applications are credit scoring, customer retention or churn management, fraud detection, and campaign management.
In the academic sector, there are many applications. In Westpoint, they conducted a study to determine who among the cadets are likely to continue through graduation and those who are more likely to stay in the military service. By predicting who are less likely to continue, interventions may be implemented before the cadet decides to drop out.
Some of the questions that are ideal platforms or subjects for data mining are shown in this chart. You will notice that there are equivalent corporate questions. Who are the students likely to take most credit hours? Who are the ones likely to return for more classes? Which alumni are likely to donate more? What courses can we offer to attract more students?
Case # 1 on the screen is about the use of data mining to create meaningful learning outcome typologies. By applying analytics, the school was able to improve understanding of student types and helped educators and administrators better meet the needs of varied students.
Case # 2 is about transfer prediction. The objective is to prepare the University for the number of transferees from the community college. The approach was to define a model that will identify who among the community college students are likely to transfer. The school also created a model that predicted the community college students who are more likely to succeed and complete the program.
Case # 3 is about predicting alumni pledges. Why do we have to attempt predicting this outcome? This approach is not necessary if there are only 100 graduates. We can send mailers to all 100 graduates and make personal calls to all 100 graduates without incurring huge costs. But if there are say over 5,000 graduates and growing, it would take a huge investment to reach out to all of them just to find who among them are going to pledge.
The application of advanced research techniques has provided institutions in many countries with the instrument for managing risk and improving performance. However, in the Philippines, decision makers have not fully embraced the value of using such research techniques to improve performance and to manage risk.
Again, my sense is that we need to first bring this to the classroom, make faculty more comfortable in using them, then equip the students who will soon become the decision makers.
(Keynote address of Dr. Nick Fontanilla during the 9th Annual Philippine Council of Deans and Educators in Business on March 12, 2010 at the Auditorium of Lyceum Philippines University)
No comments:
Post a Comment