Classroom Dilemmas — Operations Research Education
What Industry Wants From O.R. Grads
Preliminary results from analysis of employment ads offer insight on job market for students, instructors, university program directors and employers.
By ManMohan S. Sodhi and Byung-Gak Son
As we go "back to school" with this issue of OR/MS Today, it may be useful to reflect on what we are going to school for. For students, school usually means gainful employment and for their instructors it means, or at least it should mean, getting their students employed. But what do employers want? Alumni, meetings with employers and attending industry conferences provide some answers. Students themselves may learn something from job interviews, whether successful or not. But such information is neither structured nor is it disseminated very well. We take the view that employers state what they want in their recruitment advertisements so we have analyzed 647 ads for O.R. jobs. [For the purposes of this article, the term "O.R." encompasses both operations research and management science (OR/MS)]. Here we share the preliminary results of our analysis.
Our goal is to provide a way for undergraduate and master's O.R. program directors and instructors to have a detailed understanding of the job market. They can use this understanding to evaluate the extent to which they are providing their students with the necessary training to enter the workforce. Students can use our analysis to supplement their formal education at school and to be prepared for job interviews and jobs. Finally, hiring managers can look at what skills, etc. other companies are seeking and use this information to refine their own ads. Although all ads that we analyzed are from the United States, a quick check of ads from United Kingdom-based and Indian companies indicates that these companies are looking for the same skills and backgrounds. So our results may be applicable for other countries as well.
It is possible to read too much into these preliminary results especially when it comes to percentages. We have tried to make broad qualitative inferences from our analysis. After all, we are analyzing text not numbers, and the text could have a different meaning than what we are inferring. Still, we share these results because we believe that they are useful.
Our Sample
Our sample comprises job ads from U.S.-based industry employers except where a university department is seeking non-faculty hires. Government jobs, in particular from the defense forces and intelligence services, are also included. We collected ads from Monster.com and OR/MS Today for U.S. placements only that were posted between the beginning of April 2005 and the end of June 2005. At Monster.com, we searched for ads using the keywords, "operations research" or "operational research" (i.e., the entire phrase with the "or" clause). This search returned more than 650 ads. We stored these in a database.
Some employers had posted the same ads multiple times so we manually deleted duplicates from our database. There were also quite a few ads that were identical except for location, and we deleted these as well. Thus, for California-based health insurer Kaiser Permanente, we were left with 41 ads from the original 140. One quirk we discovered of the Monster's search engine was that our phrase "operations research" (within quotes) also returned ads in which the words "operations" and "research" were separated by a punctuation mark. Thus we removed dozens of ads containing, for instance, "operations, research." Eventually we ended up with 434 ads from Monster.com.
To these 434 ads in our database we added 213 industry ads from OR/MS Today posted between January 2004 and July 2005, resulting in 647 job ads that we analyzed. We cannot claim that this sample is typical or random, as ads from the health sector and from the U.S. defense and intelligence sector seem over-represented. But even though it is really a convenient sample, we hope that the large number of ads provides indicative results. Also, a sector-by-sector analysis that we do not report here resulted in similar patterns across sectors, so the impact of sector(s) being over-represented may not be significant at the level of our analysis.
Our Approach
We analyzed the text of these ads, tallying up relevant phrases and keywords in the ads. For instance, if the phrase "bachelor's degree" (or variants like B.S. degree) appears in an ad, we tally up that ad as a relevant case in the "Bachelor's" category. The same ad may specify "C++" so we tally it up in the "Programming" category as well. We have 29 such categories, and our approach is to draw out inferences from how many ads belong to each category and, for a relevant pair of categories, to both categories. Each category is comprised of many keywords or phrases, and an ad containing any of these keywords or phrases is said to belong to this category.
We have tried to ensure that we do not look only for things we want to find at the outset. Rather than start with our own keywords and phrases, we used Wordstat content analysis software from Provalis Research to list out the frequency of all words and phrases (up to five words long) in all the ads. We studied the words and phrases occurring in at least 2 percent of the ads and, from these, discarded words and phrases that are not meaningful. Then, we categorized the remaining 300-plus words and phrases into 29 categories within four master categories: (1) type of degree, (2) discipline background, (3) skills required, and (4) the nature of the advertised work. There are three degree categories — Bachelor's, Master's and Ph.D. — corresponding to the employers' requirements for a position. Likewise there are seven categories for the disciplinary background needed, nine skills-based ones and 10 categories describing the different types of work that the advertised jobs would entail.
The categories within a master category are neither mutually exclusive nor exhaustive. An ad can fall in more than one category within a master category. For instance, an ad seeking someone with either a master's or a Ph.D. degree would fall in both the Master's and the Ph.D. categories even though the lists of keywords corresponding to the two categories are mutually exclusive. Also, an ad for an experienced person may not mention any degree at all and would not fall in any of the three degree categories. We do not list all the 300-plus key words or phrases that comprise these categories, but examples of keywords and phrases for some categories appear in Tables 1 and 2.
| Table 1: Some of the keywords and phrases associated with three types of skills. | ||
| Analytical skills | Managerial skills | Project Management skills |
| • Ability to conduct and interpret • Analytical • Analytical results • Analytical skills • Analytical support • Analytical techniques • Analyzing information • Business analysis • Business analyst • Problem solver • Problem solving • Problem solving skills |
• Action plans from multi-disciplinary perspectives • Acts as liaison • Attention to detail • Change management • Leadership • Management skills • Manager/supervisor of staff • Organizational skills • Staff leadership • Strategic • Strategies |
• Assesses project impact • Develops project plans • Facilitate sharing of project outcomes and best practices • Multiple projects • Program management • Project documentation • Project management • Project managers • Project support specialists • Strategic direction of projects |
Table 1: Some of the keywords and phrases associated with three types of skills. An ad that has any of these words or phrases (among others) is categorized in the corresponding category. While the lists of keywords and phrases for the different categories do not overlap, an ad may have phrases from multiple lists and may therefore belong to more than one category.
| Table 2: Some of the keywords and phrases associated with three types of work. | ||
| OR Modeling type of work | Data Analysis type of work | Operations Management type of work |
| • Algorithms • Capacity planning • Conduct and interpret • quantitative and qualitative • Decision support • Forecasting • LP • Mathematical modeling • Mathematical models • Modeling • Modelling • Model development • Network modeling • Optimization • Simulation |
• SAS • SPSS • Analyzing data • Data analysis • Data collection • Data gathering • Data integrity • Data mining • Logistic regression • Predictive models • Regression • Statistical analysis • Statistical modeling • Statistical models • Statistical techniques |
• Business processes • Business transformation • Manufacturing and production • Manufacturing methods and procedures • Performance evaluation of staff • Performance measurement • Performance metrics • Performance metrics and resources • Process improvement • Quality control |
The reasoning behind the master categories is that any employer wants a certain background in terms of a degree level (B.S., M.S, Ph.D., etc.) in a particular discipline — O.R., computer science, business, etc. Along with this background, the employer usually specifies skills of certain kind. The ad will likely specify the nature of the work that needs to be done (Figure 1).
Figure 1: Four master categories — Discipline, Degree, Skills and Nature of Work — subdivided into 29 different categories as shown. Each category has a subset of more than 300 keywords and phrases. An ad belongs to a category if any of the associated keywords or phrases occurs in the ad.
Degree Requirements
The majority of the companies that placed job ads require a bachelor's or a master's degree. The number of ads in which a Ph.D.was sought was only about one in nine — small but not insignificant. For two in five jobs, a bachelor's degree was sought or was acceptable, and the same applies for a master's degree. Recall that these categories overlap in the sense an ad asking for a master's or a Ph.D. degree would be tallied up under both (Figure 2).
Figure 2: The percentage number of ads that mentioned a particular degree requirement. An ad may belong to more than one category or to no category at all.
Background Discipline
In general, companies are looking for a variety of backgrounds including operations research. In our sample, almost all the ads (95 percent) require an O.R. background. This should not be surprising given that we selected Monster.com ads using "operations research" as the keyword. (The phrase "operations research" appears less frequently in the OR/MS Today ads than in the Monster.com ads.) One in five (19 percent) of the ads we evaluated mentions computer science and one in nine (11 percent) mentions business (Figure 3).
Figure 3: The percentage number of ads that require a particular disciplinary background. Almost all ads mention O.R. because of the way the sample was obtained.
One way to interpret these results is that these other disciplines provide skills that are also desired by companies. This could mean that an O.R. graduate with some training in computer science or business studies would be well regarded in the job market.
Skills
Companies require a variety of skills and analyzing the ads has been extremely useful to the authors in building an inventory. We would advise educators, students and employers to pay special attention to these categories. Indeed, the bulk of our analysis effort has gone into selecting more than 100 skills-related keywords and phrases from which we constructed 10 categories (Table 1).
Of these categories, some are what might be considered "hard" skills such as "programming" and "database" skills that might already be a part of many O.R. programs. Having such skills would benefit O.R. graduates given that almost half (45 percent) of the ads require programming skills (with C++ a clear favorite) and 33 percent require database skills. About a quarter (24 percent) of all ads require basic IT skills — spreadsheet, word processing, basic data manipulation skills — that most O.R. graduates, at least in the United States, are expected to have (Figure 4). But most of the categories are "soft" skills that may not be easily gained in most O.R. programs that we are aware of.
Figure 4: The percentage number of ads that require a particular set of skills. Each "skill" comprises a number of different keywords and phrases.
At the top of list of soft skills is the "Analytical" skills category that a majority (56 percent) of ads requires. Following these are "Communication" skills (55 percent of all ads), "Managerial" skills (53 percent) and "Project Management" (31 percent). The requirement of communication skills is typically qualified by the adjective "excellent" in an ad. This could suggest that the employers find these skills weak in O.R. graduates. About one in five ads requires "strong Presentation" skills (18 percent) to the extent these ads specifically list these skills beyond the usual "excellent Communication skills." One in nine (11 percent) ads emphasize "Teamwork," while 10 percent mention the ability to "Work Independently."
The School of Hard Knocks may be a better provider of such skills. Still, O.R. programs could do more to teach these skills just as MBA programs help their students gain soft skills through team exercises, competitive exercises, group discussions, presentations and case analysis.
Nature of Work
We also tried to infer from the ads what the job itself would entail. From the data, we grouped together keywords and phrases into 10 different types of work (Figure 5). Again, an ad could belong to more than one category and may belong to none. It should not be surprising that "O.R. Modeling" (60 percent) and "Data Analysis" (46 percent) came out top (Figure 5). More than 45 percent of the ads indicated "Consulting," whether internal or external. More than a third of the ads (34 percent) indicated "Operations Management" type of work. Interestingly, "Service" jobs (13 percent) compete with "Supply Chain Management" (16 percent) jobs. Table 2 shows the some of the keywords and phrases associated with some of these categories.
Figure 5: The percentage number of ads that describe the nature of the work in any of 10 categories.
Let's see how different categories within the nature of work correlate with each other. "Correlation" here means how many ads belong to both work categories in any pair. Our analysis shows that there are four key types of work to which half or more of the ads in any work category also belong: "Consulting," "Data Analysis," "Operations Management" and "O.R. Modeling." For instance, of all the jobs belonging to the category Supply Chain Management, 46 percent also belong to Consulting, 34 percent to Data Analysis, 29 percent to Operations Management, and 85 percent belong to O.R. Modeling (Table 2). This could suggest that O.R. graduates desired in the supply chain field are attractive for their "hard" O.R. backgrounds.
We could interpret these results as follows: There are four corresponding dimensions of O.R.-related work: consulting, data analysis, operations management and O.R. modeling. Thus, an O.R. program could choose to devote part of the curriculum preparing its graduates for these four types of work.
Skills and the Nature of the Work
Next let us see what skills profile these different types of jobs require. Our ads indicate that all four main types of O.R. work have a similar needs profile across different types of skills. We infer this from the number of ads that belong to both the work category and the skills category in any pair. However, O.R. modeling work appears to require significantly more programming skills (Figure 6) than other main types of work. The other types of work show the same pattern overall as the main types of work (Figure 7).
Figure 6: Number of cases that belong to of any of the four main types of work and to the different skills category.
Figure 7: Number of ads that belong to some of types of work other than the main ones and to the different skills category.
The analysis indicates regardless of the type of work, there are six main types of skills that employers seek: (1) managerial skills, (2) analytical skills, (3) communication skills, (4) programming skills, (5) project management and (6) programming skills in no particular order. These are then the skills that an O.R. program should aim for its graduates to have.
A similar analysis by industry sector yielded the same pattern of needed skills, allaying some of our fears of the sample being over-represented by health or military and intelligence jobs.
| Table 3 | ||||
| Consulting | Data Analysis | Operations Management |
OR Modeling | |
| Consulting | 100% | 52% | 38% | 60% |
| OR Modeling | 46% | 48% | 27% | 100% |
| Data Analysis | 51% | 100% | 40% | 61% |
| Operations Management | 50% | 54% | 100% | 47% |
| General Information Technology | 60% | 51% | 44% | 49% |
| Systems Implementation | 66% | 54% | 46% | 48% |
| Service | 63% | 52% | 49% | 45% |
| Supply Chain Management | 46% | 34% | 29% | 85% |
| Marketing | 43% | 52% | 30% | 52% |
| Audit | 18% | 88% | 100% | 0% |
Table 3: The overlap of all the cases belonging to a particular type of work in terms of other job categories. Half or more ads in any type of work (rows) also occur in four types of work (columns).
Conclusion
We have shared preliminary results of text analysis of more than 600 OR/MS job ads in the belief that these results would be useful for O.R. program directors, students, instructors and employers. Our analysis indicates the following:
1. The overall market for O.R. is strong. The number of ads on Monster.com that came up for U.S. locations was 658 and were in a variety of industries including defense, logistics, pharmaceuticals, banking, insurance, airlines and consulting. Many were at junior levels, but quite a few were at senior and middle levels. We believe this means the market for O.R. skills is robust.
2. Demand for Ph.D.s is smaller than that for graduates of master's or undergraduate programs. This may mean that, at least in the United States, native students may be less likely to pursue doctoral degrees than foreign ones unless they are specifically interested in academia. Foreign students in Ph.D. programs must try to develop the skills we outlined earlier if they want jobs in industry. O.R. departments should consider strengthening their master's-level programs and undergraduate courses relative to the effort on Ph.D. programs. On the other hand, our analysis cannot show whether the need for Ph.D.s has grown or shrunk over time.
3. Consulting is a key area. Even where the employer is not a consulting company, travel requirements specified in job ads suggest that the position is of a consulting nature. This in turn means that O.R. graduates must have consulting skills — presentation ability, the ability to work and lead in small (and sometimes large) teams, the ability to interact with senior executives and manage projects. This is an important issue, and many O.R. programs, especially at top-tier universities, may find they are short of such skills within their faculty (the same applies to business schools in general). Some may have to get outside help, such as visiting lecturers from industry.
4. Manipulation and analysis of data is a common requirement. Many traditional O.R. programs, based on my own experience, emphasize mathematical programming and algorithms and either do not deal with statistical analysis at all or do so in passing. But the information economy is inundating us with data rather than with information; consequently, many jobs in data mining and statistical analysis are available. Many job ads specifically asked for SAS or SPSS skills, and many other job ads asked for database skills like SQL programming or ability to use Oracle or Microsoft SQL Server.
5. Programming skills are still quite important. Mathematical programming skills will be handy especially if the students have used C++ and/or Visual Basic. Java and C# come up occasionally, but C++ is the most common requirement among job ads that require programming.
6. Core O.R. methods are in demand, but the emphasis is on "analysis," not on using algorithms. Some of the core O.R. methods in demand include forecasting, scheduling algorithms, statistical analysis, simulation and network optimization. But mostly employers want the ability to analyze. So the fact that a student knows how to use 10 different algorithms for quadratic programming may be less relevant than the student being able to create a prototype model and to interpret and present its output.
Going back to our goal, for students applying for jobs, especially online, at the very least the above should help them figure out what to mention and emphasize in their resume. Students starting their programs could use this analysis to figure out how they should supplement their program-related learning with additional skills gained on their own. For educators, our analysis may provide a discussion point in terms of not just redesigning their programs, but also revisiting the age-old question of what O.R. is. They could also use the above to redesign homework (or "coursework" in the United Kingdom) by requiring students to work in groups and to use spreadsheets and/or database software. Presentations (not the same as Powerpoint slide shows) should be a norm for a class rather than something done once a year in a conference. Employers could also see that other employers are asking for similar things and use their links with universities to lean on O.R. programs.
For INFORMS and other O.R. societies, it may be fruitful to recognize the needs of the world of O.R. outside these societies and figure out why the people and employers from this world are not beating a path to the O.R. profession's door. Likewise, the "Science of Better" campaign should also include "normal" jobs rather than just mathematical programming successes.
We hope to provide a more detailed analysis after consultation and peer review. But, stemming from these preliminary results, we believe plenty of opportunity exists for growing the O.R. field and for placing our graduates in good jobs in industry and government. But we also need to actively consider the needs of employers in optimally allocating our resources to what we teach and how we teach it. Resource allocation is something we are good at, so here's an opportunity for us to optimize O.R.
ManMohan S. Sodhi (m.sodhi@city.ac.uk) heads the Operations Management and Quantitative Methods area at Cass Business School, City University London. Byung-Gak Son (b.g.son@city.ac.uk) is a Research Fellow at Cass Business School who recently finished his Ph.D. in supply chain management.
|
- list item
OR/MS Today copyright © 2005 by the Institute for Operations Research and the Management Sciences. All rights reserved.
Lionheart Publishing, Inc.
506 Roswell Rd., Suite 220, Marietta, GA 30060 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com
Web Site © Copyright 2005 by Lionheart Publishing, Inc. All rights reserved.