June 23-26, 2013
INFORMS Healthcare 2013
October 6–9, 2013
2013 INFORMS Annual Meeting
June 10-14, 2013
Predictive Analytics World
September 8-14, 2013
2013 ASE/IEEE International Conference on Big Data
“Data science begins with data. Nothing gets built without data. Data science continues with science. Accurate, persuasive and effective prediction requires patterns. The process of discovering that pattern is science. Any product worth building requires a reliable pattern to exist in the data.”
– Christopher Berry, co-founder and chief science officer of Authintic, in his article on recommendation engines in the current issue of Analytics.
Special ArticlesBig data paying off for big companies
A new research report, “Big Data in Big Companies,” describes how 20 large firms benefit from big data projects. Report co-authors Tom Davenport of the International Institute for Analytics (IIA) and Jill Dyché of SAS, the leader in business analytics, explore how these companies have deployed analytics to generate value from their big data assets.Read More
Special ArticlesContinuing education courses for analytics professionals
The Institute for Operations Research and the Management Sciences’ (INFORMS) Continuing Education program will offer its first two courses this fall. These intensive, two-day, in-person courses will provide analytical professionals with key skills, tools and methods that can be implemented immediately in a work environment.Read More
Industry NewsBanks speed lending with FICO’s cloud-based loan solution
FICO, a leading predictive analytics and decision management software company, recently announced that 1st United Bank, First Utah Bank, Lowell Five Cent Savings Bank, and MetroBank are implementing the FICO LiquidCredit service, a cloud-based scoring solution that speeds consumer and small business lending decisions. The FICO solution will enable these lenders to approve small business loan applications in hours, not days, and help improve customer satisfaction and attract more small business customers.Read More
Serving the Services Industry
The emerging science of service management opens opportunities for business analytics.
By Brenda Dietrich and Terry Harrison
The services industry continues to be a rapidly growing segment of many developed economies, including the U.S. economy [1, 2]. Although a significant portion of the services industry is focused on providing services to individuals (medical, insurance, legal, financial), the business services sector, in which one company provides service to another company, is also a rapidly growing segment [3, 4]. Examples include traditional consulting, design, technical support (typically for products), call center operations, IT implementation and IT outsourcing. New business models, based on improving efficiency through automation, aggregation of risk, economies of scale or reduction of capital assets, lead companies to outsource and in some cases off-shore business processes that do not provide differentiation in the marketplace. Transportation and warehousing, procurement, manufacturing, benefits management and back-office processes such as accounting are all now being provided as services. Business services are complex, and are typically purchased and managed by separate organizations within an enterprise.
Over the past several decades mathematical models of traditional manufacturing and logistics systems have been developed and used for strategic planning. More recently similar models have been used to support operational decision-making. Significant gains in efficiency within the manufacturing and logistics industries have been attributed to the use of such models, together with a supporting information technology infrastructure [see 5, 6, 7, 8 for examples]. Manufacturing Resource Planning (MRP), which automated the calculations of material requirements within manufacturing, evolved in to Enterprise Resource Planning (ERP), which monitors all manufacturing enterprise processes, and formed the information base for advanced planning and e-commerce.
Information technology support, and in particular models and analytic tools, for the business services industry lags behind manufacturing and financial services counterparts. ERP, Customer Relationship Management (CRM) and other classes of enterprise software developed for other segments can address some of the data management issues in business services, but it is difficult, and in some cases impossible, to directly apply the analytic tools developed for manufacturing and supply chain to business services.
In a supply chain, one key control lever is inventory.Managers can determine how much of what part to hold in each location to buffer against variation in either supply, demand or process yield. In services, resources typically cannot be held: a technician’s time, if not used today, is not available for use tomorrow. Furthermore, resources in a services industry, largely people, are more complex to model than machines and parts. A programmer may acquire new skills by working on a project; a machine rarely acquires new capabilities just by being used. Attrition is also an issue, in that as a person becomes more skilled, he or she may have an increased probability of changing employers. New models are required to represent and understand these complexities, and to further increase the efficiency of the services industry. This article describes some of the characteristics of business services and discusses opportunities for business process support that could be addressed through the use of management science methods.
Characteristics of Business Services
The Service Management Interest Group at the Harvard Business School differentiates services from manufacturing according to the following four features: (1) services are often marketed and performed by the same people; (2) services rarely can be inventoried in conventional ways; (3) services present special problems of quality control in a real-time delivery environment; and (4) services involve the majority of jobs in any developed economy and thus benefit from creative human resource management . For business services, the marketing and delivery are likely to be performed by different organizations within the same enterprise; similarly, business services are likely to be purchased and consumed by different individuals within an enterprise.
Creating and delivering a service requires the use of some collection of assets, whether capital assets such as information technology infrastructure, consumable assets such as service parts and materials, labor assets such as skilled employees, or intangible assets such as an individual’s skills or an organization’s proprietary data or processes. Many commodity manufacturers, such as computer card assemblers and chemical processors, are repositioning themselves in the market as service providers, by offering value-added services wrapped around the core product. Their physical processes remain essentially the same, but their relationships with customers change, and their management and planning processes must evolve to support this change. In other business services, such as professional consulting services, the bill of resources associated with an offering (e.g., consulting engagement) may have significant variability depending of the specific features of the transaction. Further, the ability to substitute one resource for another, as in one consultant for another, or one computer for another, is generally higher in business services than in traditional manufacturing.
Since in most cases, service cannot be inventoried, consumption takes place at or near the time and location of production. Additionally, the value from services is often “co-produced” jointly by the service provider and service buyer. Prior to the availability of worldwide communication networks, consumption also took place at or near the point of production. Now many services can be produced in one location and consumed in another. Technical help desks, many of which are located in India but serve clients in North America, provide a quintessential example. Location-independent delivery, especially when coupled with standardization, gives the purchaser of services more ability to drive price.
In many cases an organization not only purchases a service, but also contracts for flexibility on the use of the service. Since service is perishable, either the rate of production must be flexible so that production can be matched with consumption, or the rate of consumption must be controlled through mechanisms like commitment management, multiple service classes or pricing. It is important to note that in business services, there are typically multiple users of a service, most who are not involved in the purchasing decision. This makes issues of “quality” and customer satisfaction more complex and difficult to define.
The concept of maturity of products is different from maturity for services, as a service generally includes not only the function that is being provided, but also the means by which the function is sold and priced. Electricity is a mature commodity, but it is being sold and priced in new ways. Telephone service is being offered under a huge range of pricing plans and being bundled with other services such as caller-ID, phone mailboxes and Internet connections. As with physical products, services can be differentiated by tailoring to meet the needs of specific customer segments; further, physical products can be differentiated by adding service components.
The sales process for many business services, particularly for professional services, is largely relationship-based, with a long “pre-production” phase that might involve a number of small transactions leading up to larger, longer duration engagements. However, it is typical for business services contracts to be renegotiated long before the contract term is completed.
Although one typically thinks of services as being delivered directly from the provider to the receiver, there is, in fact, a significant level of subcontracting in the services industry, and aggregators and distributors of service capability are emerging. As with physical goods distribution networks, there are issues regarding the degree to which information on the end-consumer is shared with the participating providers. It is obvious that information regarding the end-customer can lead to improved service, through, for example, differentiated offerings, however there are economic forces and in some cases governmental or societal forces, that work against the aggregation and analysis of end-customer behavior.
Business Services Processes
A buyer of business services considers five process steps: analyze, plan, source, buy and monitor . A provider (or seller) of service typically considers five complementary processes: plan, sell, execute, charge and evaluate. Notably absent from this representation is a “build” process that is the center point of many models of manufacturing. In business services it may be helpful to think of processes from the perspective of both the buyer and the seller. Planning the production and/or consumption of services takes place in both buyer and seller organizations, although for the buyer the planning may be driven by cost issues coupled with considerations of core competencies, while in the seller it is likely to be driven by revenue considerations. Both buyers and sellers may have relationships with subcontractors who provide or consume some portion of the service. Hence the supply or demand and other participants in the services chain may be considered in the planning process.
Selling, which may involve defining a specific offering or specifying the terms and conditions of a contract, includes sub-processes such as customer identification, advertising, pricing and contract negotiation. The complementary buyer process is, obviously, buying, which may include releasing a request for proposals, comparison of bids, selection of providers and negotiation of contract terms. Business service contracts typically include payment terms, and if the payment is based on usage, may specify the means by which the usage is measured. Execution corresponds to production and consumption of the service, by the seller (or its subcontractors) and the buyer (or its end customers), respectively. Billing, and the corresponding buyer process of paying, have significant impact on an enterprises financial health, and must be executed according to terms specified in the contract. The on-going monitor/evaluate processes focus on contract compliance and resulting business performance, and often drive contract renewals and renegotiation.
Opportunities for O.R. Applications in Business Services
Key questions that planners and operators of business services need to answer include:
- forecasting demand for services,
- forecasting demand for resources to produce services,
- strategic and tactical planning of acquisition, training, and termination of resources,
- allocating resources to specific activities, and
- pricing service contracts.
Furthermore, because of the simultaneity of production and consumption, these issues are very tightly linked, and it is difficult to address any single issue without knowledge of the manner in which the others will be addressed. In most manufacturing models, the “unit of work” is reasonably well defined. Many models assume that for each specified work unit, resource requirements (bill of materials, machine-processing time) are known, often with certainty. Because of the dynamic, time sensitive nature of service production and consumption and the significant labor component of service production, variability, both in the number of “service units” and in the composition of individual service units, must be a consideration in any approach to supporting services decision processes. This is a critical aspect for successfully managing a service enterprise and a significant opportunity to apply quantitative modeling to gain competitive advantage.
For conventional goods, demand forecasts are used to set safety stock and replenishment levels; to drive procurement, especially of long lead-time items; to establish production plans; and as input to capacity and financial planning processes. In services, demand forecasts are used primarily for financial planning, capacity and workforce planning, and to drive procurement or creation of the assets used to deliver services. Because services cannot be inventoried in conventional ways, for accurate capacity planning it is necessary to forecast peak demand, or at least peak usage, rather than mean and range. Services assets often have large fixed costs and low variable costs, or, more generally, a step-function cost structure, where one unit of capacity can serve up to D units of demand at cost C, but servicing the D+1st unit of demand costs an additional C units. This means that demand management, through contract terms and pricing that specify for example time of service, location of service, and timeliness of service, play a critical role in determining the profitability of a portfolio of contracts. In some business services segments, demand forecasting is further complicated by the existence of a small number of very large deals and the small number of market participants bidding against one another for these deals. Additionally, services demand is influenced by corporate and government budget cycles. The stochastic nature of demand likely has an even larger impact in service industries planning than in manufacturing environments.
Research areas that could support improved demand forecasting for business services include statistical methods and models for truncated data and censorship, statistics of peak estimation, statistics of heavy-tail distributions, improved models linking demand with pricing and other promotional actions, and other models for representing and estimating the demand for business services. Methods for connecting forecasts of demand to forecasts of requirements for resources and forecasts of revenue are required. For some business services, particularly where demand occurs as a small number of large deals and a small set of providers compete for most of the deals, forecasting is especially problematic, as one has to deal with the individual probabilities of winning each deal. Methods from game theory may lend insight in the planning process.
In some cases it may be possible to adjust peaks in demand by anticipating and providing service before the customer realizes it is required. Equipment maintenance is one example. If the usage and performance of the equipment can be remotely monitored, some problems can be detected and resolved before a failure occurs. Statistical methods to predict failures, and models to understand the relative benefit in cost, customer satisfaction and peak capacity needs or preemptive action, could be used to support both planning and operational processes.
Not all business services are time-critical, so demand management, that is moving demand in time, in space or from one set of resources to another through the use of subcontracting, for example, can be extremely beneficial in smoothing demand. However, it can require complex pricing structures and increased transactional complexity (for example, communicating the time urgency of the service request, or negotiating a time-based price for each request), so the relative benefit must be carefully evaluated. In some cases,where a buyer wants to reserve specific service capacity there may be an opportunity to develop pricing and/or planning models based on real options.
Capacity planning for business services can be a complex process. The need to provision for peak demand, coupled with the flexibility of many of the assets used to produce a service and the variability of the demand, can make even answering simple questions like “Do I have enough capacity to do X” quite difficult. Just as in manufacturing systems that produce a range of products with flexible resources, the capacity to deliver a business service can be a function of the service mix (percentage of each type of services) and the way in which each resource is deployed. A reasonable analogy may be found by considering capacity planning for semiconductor manufacturing . Throughout the 1980s, semiconductor manufacturers made extensive use of detailed discrete-event simulation models to understand the interactions between line throughput, product mix, tool utilization and classes of production (e.g., standard vs. expedited jobs). These simulation studies were used for, among other things, setting tool utilization and buffer targets and understanding the relative value of dedicated tools and reserve capacity.With the exception of call centers (and related high-volume, single-step processes), there are few quantitative studies exploring the relationships between factors such as utilization, variability, service levels, dedicated resources, reserve capacity and cost of delivery for services.
In business services, particularly those with a significant labor component, there exists significant variability in the amount of resource required to complete a step in service production.We also expect to see non-linear effects in team formation (both super-additive and sub-additive), and changes in the capability and efficiency of individuals over time. Additionally, human workers have the capacity to learn, to improve their efficiency and to acquire new skills through training or work experience. Hence the choice of which workers to use for a specific task may be influenced not only by current needs, but by a view of expected future requirements.
Methods for capacity planning and assessment that can effectively deal with the variability of work content and delivery rates are required. Further, in some services businesses, the provider has some ability to influence the work content of tasks and/or the arrival rate of tasks through pricing, through response rate, and through the allocation of resources to individual tasks. Models which capture these interactions would be valuable to both support execution decisions and to make longer term capacity planning decisions.
Many business services involve a significant degree of automation, which can be a source for valuable data that can be analyzed to find predictive relationships. The data and the relationships can be used as input to both planning processes and to support operational decisionmaking. Data analysis may also support efforts to define a “bill of services” that would play an analogous role to a bill of materials in manufacturing, by specifying the relative quantities of service production resources that must be applied in combination and over time to fulfill a unit of service consumption. It is likely that to be truly useful, a bill of services would have to capture some representation of variability, especially across individuals with varying experiences and skill sets.
Traditionally consulting and other professional services are priced on a time and materials basis, although other pricing mechanisms are possible.At two extremes are fixed-price, unlimited service contracts and per-transaction contracts. Fixed-price, unlimited service contracts are common in business to consumer services; examples include annual parking permits, transportation passes and many health club memberships. Business services examples of fixed-price contracts include product service contracts and business recovery services that provide access to back-up data centers.
Other pricing models are possible in business services including value pricing, based on an estimate of buyer savings, gain-sharing based on establishing a base line on a performance measure and sharing in any improvement, and bundling of services with a product. When setting prices for such bundles, one must consider the price of the “products” and the price of the related “services” in combination, especially when the products and services can also be acquired separately. Service contracts often specify specific timetables and quality of service levels and may include significant penalties to the seller for missing due dates or denying requested service. In other cases the cost of not delivering is less clear and may be reflected only in reduced future business.
Data indicates that simple flat-rate pricing tends to be most effective in business-to-consumer services, perhaps in part due to the difficulty of making comparisons between offerings with more complex pricing structures . This may not be the case for business services; however the ability to monitor usage of the service and to provide status and incremental cost information to the buyer (or user) is a prerequisite for the acceptance and use of most complex pricing structures. Factors that might be considered for business services pricing include flexibility, quality, time of delivery, risk responsibility, variability in cost to purchaser, and parameters related to the time, location and volume of consumption. It is important to note that most business services prices known only to the buyer and the seller.
Service providers could benefit from models that help determine a minimal acceptable price, help evaluate the range of possible outcomes of a given pricing structure, and evaluate the profitability and/or risk of an individual contract, either with or without consideration of other existing or possible contracts. Service buyers could immediately benefit from models that help them evaluate and select from multiple pricing structures. Experienced service buyers may also benefit from methods that help them consider additional service attributes, such as integration and switching costs, availability of information on details of the consumption of the service by end-users, and the ability to manage and control that consumption. The existence of subcontractors, consolidators and resellers in many of the business services markets further complicates the both the pricing and the selection process.
Methods for understanding the impact of pricing structures on participant behavior, especially identification of pricing structures that, like Vickery-Groves auctions, help align the incentives of the buyers and the sellers and encourage information disclosure in the negotiation process would be useful. Further, as most service contracts are renegotiated periodically, models for representing the buyer usage pattern and understanding the value associated with the consumption of the service may prove useful. Finally, pricing models, for both buyers and sellers, that capture anticipated renegotiation, perhaps drawing from theory of repeated games, might be valuable.
Once business service contracts are in place, and the buyer and seller move from planning and negotiation to operations, there are additional opportunities for the use of O.R. methods. From the provider’s perspective, one of the most critical decisions involves determining how resources are to be allocated when there is a temporary surge in demand or shortage of resources. This allocation will determine the provider’s performance relative to the performance criteria in service contracts, and hence will impact the contract’s revenue and profitability. Therefore, to really understand the economic profile of a contract or set of contracts, one must model the demand variability, the variability in resources requirements, and the allocation process used to resolve temporary mismatches in supply and demand. Advanced planning systems, finite scheduling, capacitated planning and other techniques from supply chain may have a role here, provided the underlying models can be augmented to represent the more complex cost structures of service contracts and the limited ability to use inventory as a damping mechanism.
In the area of resource allocation, particularly for professional services such as consulting and software application development, there is a role for project management capability. But most currently used tools fall short, in that they tell the user when there is a problem but do not provide recommendations for fixing the problem or even evaluation of actions proposed by the user. Further, they do not provide support for managing a portfolio of projects that are to be executed using shared resources, or provide adequate means for dealing with variability in either task duration or resource requirements. Project scheduling and managements, project risk assessment and project portfolio management are areas that could be addressed by methodologies from O.R. However, to be broadly used, these methodologies would have to be encapsulated in tools that can access data from existing project and personnel management and monitoring processes, and include user interfaces that effectively isolate the mathematical models from the end-users. Professional Services Automation (PSA) tools, which are in the early stages of deployment in the services industries, may provide much of the data needed to drive human resource planning and allocation. These tools currently make only limited use of the capability available in quantitative models. The O.R. community has an opportunity to influence the direction of these tools and to impact the services industry through the inclusion of O.R. methods in the next generation of PSA tools.
Most business services, and in fact most services, involve a significant labor component. Business services that involve process outsourcing often result in the transfer of employees from a company to its supplier. Thus, the planning and management of human capital is an important consideration for both buyers and sellers of business services. The field of Human Capital Management (HCM) encompasses activities such as attracting, hiring and retaining human resources. Software tools and consulting firms offer tools and strategies that help firms better plan and manage workforces. HCM includes functions such as job classification, leadership assessment and training, organization design, analysis of labor markets and career paths, studies of organizational culture, and tools and strategies for knowledge management. HCM functions are usually applied at the enterprise level, not at the individual level. The task of translating a strategy or policy in to local actions typically rests with local decision-makers, perhaps supported by some reporting or monitoring functions. There is a significant opportunity to extend tools and consulting methodologies through the use of O.R. capability. Initial activity toward augmenting HCM with O.R. could include defining the attributes used to categorize human capital, modeling the role of social capital and analyzing the value of flexibility within organizations and workforces. For example, one could use simulation and/or queueing models to compare the efficiency and adaptability of systems involving highly skilled but inflexible workers with those involving a mix of experts and less skilled but flexible employees, much as we previously compared rigid optimized assembly lines to flexible manufacturing systems.
Just as supply chain simulation has been valuable in understanding the effects of variability, the value of information and the potential impact of improved decision-making, simulation models of the engagement process might be useful from the receipt of the request for proposal through the delivery of service and collection of payment, and may provide significant insight into the value of information and process transformation in the business services industry.
There are notable successful examples of the use of O.R. methodology in business services. For example, in the transportation industry, carrier bid optimization has been successful from both the carrier and shipper perspective . Recent work with routes in airline yield management has been applied in the healthcare industry, supporting hospitals in their negotiations with insurance companies . However, the use of O.R. in services remains relatively small, especially when one considers the importance of business services in developed nations. Although there is some applicability of existing O.R. models to business services, broader applicability and acceptance of these methods will need to leverage results from other domains, including marketing, behavioral psychology and economics. To capitalize on these opportunities the O.R. community must take steps to forge connections with a number of other professional communities who have the knowledge, experience and relationships necessary to influence the direction of business services management.
Brenda Dietrich is vice president of business analytics and mathematical sciences at IBM. Terry P. Harrison is a professor of Supply Chain and Information Systems at Smeal College of Business, Penn State University.
Many of the ideas in this article were formulated during a small workshop attended by Cynthia Barnhart, Dirk Beyer, Gabriel Bitran, Morris Cohen, Amr A. Farahat, Terry L. Friesz, Noah Gans, Steven Graves, Karla Hoffman, Donald Rosenfield, Michael Rothkopf, Eric Lesser, Paulo Rocha e Oliveira.
1. Bryson, John, Peter Daniels, and Barney Warf, 2005, “Service Worlds: People, Organisations and Technology,” Routledge Publishers, Andover, U.K.
3. Tanninen-Ahonen, Tiina, 2003, “Professional Business Services: The Key to Innovation,” Institute for the Future 10-Year Forecast, Menlo Park, Calif.
4. Ravi, R. et al, 2005, “Worldwide and U.S. Business Process Outsourcing (BPO) 2005-2209 Forecast: Market Opportunities by Horizontal Business Function, IDC report 33815, August 2005.
5. Ton de Kok, Fred Janssen, Jan van Doremalen, Erik van Wachem, Mathieu Clerkx, Winfried Peeters, 2005, “Philips Electronics Synchronizes Its Supply Chain to End the Bullwhip Effect,” Interfaces, Vol.35, No. 1, pp. 37-48.
6. Roman Kapuscinski, Rachel Q. Zhang, Paul Carbonneau, Robert Moore, Bill Reeves, 2004, “Inventory Decisions in Dell’s Supply Chain,” Interfaces, Vol. 34, No. 3, pp. 191-205.
7. Moritz Fleischmann; Jo A. E. E. van Nunen; Ben Gräve, 2003, “Integrating Closed-Loop Supply Chains and Spare-Parts Management at IBM,” Interfaces, Vol. 33, No. 6, pp. 44-56.
8. Gail Hohner, John Rich, Ed Ng, 2003, “Combinatorial and Quantity-Discount Procurement Auctions Benefit Mars, Incorporated and Its Suppliers,” Interfaces, Vol. 33, No. 1, pp. 23-35.
10. Christa Degnan, 2003, “The Services Supply Chain Automation Benchmark Report, Strategies for a Buckshot Market,” Abredeen Group, June 2003.
11. Stuart Bermon, Sarah Jean Hood, 1999, “Capacity Optimization Planning System (CAPS),” Interfaces, Vol. 29, No. 5, pp. 31-50.
13. Yossi Sheffi, 2004,”Combinatorial Auctions in the Procurement of Transportation Services,” Interfaces, Vol. 34, No. 4, pp. 245-252.
14. Chris Born, Monica Carbajal, Pat Smith, Mark Wallace, Kirk Abbott, Surain Adyanthaya, E. Andrew Boyd, Curtis Keller, Jin Liu, Wayne New, Tom Rieger, Bert Winemiller, Ron Woestemeyer, 2004, “Contract Optimization at Texas Children’s Hospital,” Interfaces, Vol. 34 No. 1, pp. 51-58.