Supply Chain Wide Optimization at TNT Express...Fleuren et al.: Supply Chain–Wide Optimization at TNT Express Interfaces 43(1), pp. 5–20, ©2013 INFORMS 7 management understood

This article was downloaded by: [139.179.113.31] On: 18 October 2014, At: 21:26Publisher: Institute for Operations Research and the Management Sciences (INFORMS)INFORMS is located in Maryland, USA

Interfaces

Publication details, including instructions for authors and subscription information:http://pubsonline.informs.org

Supply Chain–Wide Optimization at TNT ExpressHein Fleuren, Chris Goossens, Marco Hendriks, Marie-Christine Lombard, Ineke Meuffels,John Poppelaars,

To cite this article:Hein Fleuren, Chris Goossens, Marco Hendriks, Marie-Christine Lombard, Ineke Meuffels, John Poppelaars, (2013) SupplyChain–Wide Optimization at TNT Express. Interfaces 43(1):5-20. http://dx.doi.org/10.1287/inte.1120.0655

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial useor systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisherapproval, unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitnessfor a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, orinclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, orsupport of claims made of that product, publication, or service.

Copyright © 2013, INFORMS

Please scroll down for article—it is on subsequent pages

INFORMS is the largest professional society in the world for professionals in the fields of operations research, managementscience, and analytics.For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org

http://pubsonline.informs.org

http://dx.doi.org/10.1287/inte.1120.0655

http://pubsonline.informs.org/page/terms-and-conditions

http://www.informs.org

Vol. 43, No. 1, January–February 2013, pp. 5–20ISSN 0092-2102 (print) � ISSN 1526-551X (online) http://dx.doi.org/10.1287/inte.1120.0655

© 2013 INFORMS

THE FRANZ EDELMAN AWARDAchievement in Operations Research

Supply Chain–Wide Optimization at TNT Express

Hein FleurenDepartment of Econometrics and Operations Research, Tilburg University, 5037 AB Tilburg, The Netherlands, [email protected]

Chris Goossens, Marco Hendriks, Marie-Christine LombardTNT Express, 2132 LS Hoofddorp, The Netherlands{[email protected], [email protected]}

Ineke MeuffelsDepartment of Econometrics and Operations Research, Tilburg University, 5037 AB Tilburg, The Netherlands; and

ORTEC, 2803 PV Gouda, The Netherlands, [email protected]

John PoppelaarsORTEC, 2803 PV Gouda, The Netherlands, [email protected]

The application of operations research (OR) at TNT Express during the past seven years has significantlyimproved decision-making quality and resulted in cost savings of 207 million euros. The global optimization(GO) program initiative has led to the development of a suite of optimization solutions to assist the operatingunits of TNT Express to improve their package delivery in road and air networks. To create and deploy thesesolutions, we established communities of practice (CoPs), at which internal and external subject matter expertsmeet three times annually at an internal conference. We also created a unique two-year learning environment,the GO academy, where employees of TNT Express are taught the principles, use, and deployment of optimiza-tion techniques. As a result of these combined initiatives, OR is now an effective part of the core values atTNT Express.

Key words : express service providers; transportation; supply chain optimization; network design problem;pickup and delivery problem; change management; OR deployment.

TNT Express N.V., one of the world’s leadingbusiness-to-business express delivery companies,

operates the largest express road and air networkin Europe, and air and road transportation net-works in China, South America, Asia-Pacific, and theMiddle East.

In this line of business, express delivery companiesmove packages (i.e., parcels, documents, or pieces offreight) from a sender to a receiver under variousand guaranteed service-level agreements that specifydelivery dates and times. Each service offering con-sists of collecting packages at a customer site, trans-porting them via a road and (or) air network, anddelivering them to a recipient.

Each week, TNT Express delivers 4.7 million pack-ages to recipients in over 200 countries, using anetwork of more than 2,600 facilities, a fleet of about30,000 road vehicles and 50 aircraft, and a work-force of 77,000. Because of the highly volatile andcompetitive nature of the express delivery market,the company must ensure that its network is robust,agile, and able to effectively absorb demand fluctu-ations. Express delivery companies are focused onachieving both cost efficiency and high levels of cus-tomer service, two goals that are often contradictory.The challenge is to design a supply chain that caneffectively meet both criteria and manage this criti-cal balance between them. Given that point-to-point

5

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.

Fleuren et al.: Supply Chain–Wide Optimization at TNT Express6 Interfaces 43(1), pp. 5–20, © 2013 INFORMS

Figure 1: The TNT Express supply chain consists of road and air operations. The pickup and delivery (PUD)process concerns the collection and delivery of packages at the customers; the network process addresses thetransportation of packages between depots.

volumes are typically too low to justify a single trans-port over large distances, we prefer to maximize con-solidation to reduce costs. Conversely, our ability totransport packages efficiently is restricted by the typesof services we offer, which may vary considerably. Forexample, we offer time-definite express services witha guaranteed next-day delivery before 9 am, 10 am,12 noon, or the end of the day, but we also offerday-definite services for less-urgent shipments that donot require next-day delivery. As a result, consolidat-ing packages from similar collection origins can bedifficult because the delivery time frames may varyconsiderably. Figure 1 illustrates our air and roadsupply chains.

After collection, packages are transported to depots,which are local sorting centers that manage the col-lection and delivery of customer packages. Hubs arelarge sorting facilities used to consolidate the trans-port of packages between the depots. TNT Expressrefers to the collection and delivery of packages atthe depots as the pickup and delivery (PUD) process,whereas the transport between the depots is the net-work process. Cutoff times (i.e., due times) separatethe PUD process from the network process: at thepickup cutoff time, the packages must be availableat the origin depot for the network process; at the

delivery cutoff time, the packages must be avail-able at the destination depot for the delivery process.The PUD process encompasses the processing timeat the depots, whereas the network process includesprocessing times at the hubs. Although we separatethe PUD and network processes, implementing eachprocess presents a number of challenges. For example,the PUD process includes assigning customer pick-ups to a particular depot, deciding on the number ofvehicles required, and determining when the vehicleswill visit the customer for collection or delivery. Inthe network process, a number of important decisionsare made; these include sorting centers to be visited,sequence and time of departure, and the schedulesneeded for the vehicles connecting these locations. Atthe supply chain level, decisions about cutoff times(e.g., allocating more time to one process in favor ofanother) and the number and location of hubs anddepots required must be made.

Operations Research at TNT ExpressIn 2005, TNT Express embarked on its first oper-ations research (OR) project. The initial strategywas to expand business activities rather than justfocus on cost reductions and asset utilization. Senior

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.

Fleuren et al.: Supply Chain–Wide Optimization at TNT ExpressInterfaces 43(1), pp. 5–20, © 2013 INFORMS 7

management understood that focusing on growthwould not guarantee a competitive advantage in thelong run. Triggered by a story on optimization byTilburg University professor Hein Fleuren, MarcoHendriks, Director of Strategic Operations and Infras-tructure at TNT Express, sensed that quantitativemethods should become the key enabler to increasethe company’s competitiveness. This awareness ledto TNT Express’ first OR project, which was aimedat optimizing Italy’s domestic road network. Theresults were promising: by rescheduling vehicles andreassigning packages, asset utilization increased andtransportation costs decreased by 6.4 percent. Thisinitial success paved the way for the global optimiza-tion (GO) program and the close working relation-ship between TNT Express, Tilburg University, andORTEC, an OR consulting and optimization softwareprovider that partners with TNT Express on optimiza-tion activities.

The idea of incorporating OR into our decisionmaking grew steadily. We were convinced that opti-mization activities should be at the core of our busi-ness and that we should not organize these activitieswithin a separate, large, centralized OR department.To achieve these goals, we established communities ofpractice (CoPs) and the GO academy. A CoP is a com-munity of TNT Express business experts worldwide,which a central GO team organizes. CoP meetings, atwhich participants share best practices and optimiza-tion knowledge and discuss these ideas with sup-pliers and academia members, are held three timeseach year. These meetings typically last two to threedays and have 15 to 20 attendees. The GO academyis a two-year program designed to teach managementand staff the principles of optimization, which willenable them to recognize optimization opportunitiesand develop a common global optimization language.As the added value of OR gained visibility, the TNTExpress board adopted OR to develop strategies torespond to the consequences of the economic crisis in2008. Senior management understood that by apply-ing OR across the business, it would be able to man-age unit costs more effectively, while continuing tofulfill all service obligations. Unit costs had becomeimportant because they were rising steadily as a resultof decreasing demand, increasing fuel prices, andmore stringent environmental regulations.

The GO ProgramThe goal of the GO program is to improve decisionmaking throughout the TNT Express organization andin each part of its supply chain. To address the chal-lenges in networks, the PUD process, and the entiresupply chain, we set up separate subprograms foreach. These subprograms led to the creation of a port-folio of models, methodologies, and tools designedto solve the optimization challenges we encountered.Given that each operating unit is at a different levelof operational maturity, the solutions differ in opti-mization complexity. At the lower end of the matu-rity scale, dashboards and guards help the operatingunits to analyze their actual performance and identifynew optimization opportunities. For the more matureunits, we deploy advanced optimization solutions.

These solutions are important in identifying andrealizing each GO subprogram cost saving; thesesavings were 207 million euros over the period2008–2011: 132 million euros from the supply chainsubprogram, 48 million euros from the networks sub-program, and 27 million euros from the PUD sub-program. The GO program also enabled us to reduceCO2 emissions by 283 million kilograms—the CO2

equivalent of 1,000 trucks traveling around the worldseven times.

Although we expect more savings in the future,this paper focuses on the solutions that have been inuse for some time and therefore contributed the mostto the reported savings. These solutions, which wedescribe in the sections below, are: TRANS in the net-works subprogram, SHORTREC in the PUD subpro-gram, and DELTA supply chain in the supply chainsubprogram.

We also describe how our use of OR has evolvedto become a key component in decision making viathe GO program. We illustrate this by describing theOR methods we applied and the benefits accrued andchallenges encountered to date. We then highlight theGO academy—a game changer for successfully apply-ing OR within our company. We conclude by describ-ing our findings on applying OR at TNT Express.

Subprogram 1: TNT Express Routingand Network Scheduling (TRANS)Network optimization is concerned with optimizingroutes for the transportation of packages and vehicle

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


tours. Within TRANS, our optimization software, thenetwork infrastructure (i.e., depots and hubs), the vol-ume of packages to be transported, and the cutofftimes are considered to be fixed. A transportationroute defines the sequence of hubs, from the depotof origin through to the destination depot, includingscheduled times of arrival and departure at the hubs,that a package will visit. A path is the simplificationof a route, denoting only the sequence of hubs andexcluding time information. Thus, multiple routes canoperate along the same path. A tour describes thesequence of locations visited by a vehicle (and driver),including the times at which each location is visited.In general, tours start and end at the same location forthe convenience of the driver. A movement connectstwo successive locations of a tour, with no intermedi-ary stops. Characteristics of a movement are its depar-ture and arrival times and the corresponding vehicletype. An empty movement is a repositioning of a vehi-cle that is not carrying packages. Figure 2 illustratesthese definitions.

Because of the size of the networks that TNTExpress operates, using a combined problem to deter-mine the routes and tours would be too complex.Path generation is exponential relative to the numberof locations; Italy’s domestic network has about 100

Figure 2: The path and route illustrate the movements of packages from their origin (e.g., Turin) to their finaldestination (e.g., Florence). The path is the subset of a route that excludes drop-off times. A movement connectstwo successive locations and is a subset of a tour.

depots and 10 hubs, resulting in over 35 billion pos-sible paths for packages. Therefore, we separate theproblem into several subproblems, each supported bya specific module in TRANS:

• The service capability analyzer determines thefastest feasible routes based on the prespecifiedmovements in the network. The resulting fastest pos-sible service offerings are visualized on a map. Ser-vice implications of modifications to the movementscheme (i.e., the total set of movements operated inthe network) are recalculated within a few seconds tosupport what-if analyses.

• The routing module is an extension of the ser-vice capability analyzer. It generates a set of routes(not only the fastest) and assigns the packages tothe movements of these routes. If packages cannotbe assigned to movements, this is usually because ofinsufficient capacity, an issue that must be resolved.Conversely, if movements are underutilized, oppor-tunities for improvements may exist. The routingmodule visualizes the overutilization and (or) under-utilization of movements; however, it does not auto-matically resolve these issues.

• The movement heuristic module constructs anew movement scheme based on the packages andtheir corresponding paths and assigns them to the

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


resulting movements. The movement heuristic mod-ule can be used in combination with the optimal pathsmodule to generate a new movement scheme, and canalso be used to evaluate existing movement schemes.

• The optimal paths module determines theoptimal paths for each package, given the currentinfrastructure of depots and hubs; it also considersservice restrictions. In combination with the move-ment heuristic, this module supports the redesign ofa network’s arcs in its entirety.

• The tour generation module generates toursbased on a movement scheme, including any emptymovements. In general, empty movements are mini-mized to ensure that the tours generated are efficient.

All the above modules were designed and tested inclose collaboration with the CoPs, who provided thebusiness knowledge and requirements. Each modulecan be used as a standalone module or in combina-tion with other TRANS modules; the results can beanalyzed via a range of graphical visualizations andkey performance indicators (KPIs).

Operations Research Techniques Used in theTRANS ModulesThe problem solved by the service capability ana-lyzer and routing modules can be formulated as amulticommodity flow problem (Ahuja and Magnanti1993) in a time-space network, because of the requiredconnectivity of movements at all locations. Root andCohn (2008) describe the setup of this time-spacenetwork; a node (fi1 ti5 corresponds to a facility(i.e., a depot or hub) fi at a point in time ti, andan arc between (f11 t15 and (f21 t25 represents the flowof packages from facility f1 to facility f2, departingat facility f1, at time t1, and arriving at facility f2 attime t2. Each arc represents a movement. To solvethe problem within a reasonable run time, irrespec-tive of network size, we use a heuristic approach. Forroute generation, we apply a branch-and-bound algo-rithm to generate a user-defined number of routesthat will meet the service requirements (note thatthe service capability analyzer needs only one routeper origin-destination pair). The bounding rules area combination of the number of hub touches (as lowas possible), the arrival time at the destination depot(as early as possible), and the departure time atthe origin depot (as late as possible). The branches

result from the movement scheme; the occurrenceof these branches is restricted to hub locations only,because the transfer of packages from one movementto another is permitted only at hub locations.

In addition to generating the routes, the route mod-ule assigns the packages to the routes that are gener-ated. The assignment of packages to routes is basedon route preferences and available capacity. In partic-ular, the preferred route is taken for each set of pack-ages (where the preference follows the same rules asthe bounding rules), and the packages are allocatedto route movements only if available capacity exists;if no capacity remains in any of the route move-ments, the second most-preferred route is selected;this assignment process is repeated until all packagesare assigned to movements or until all routes havebeen evaluated.

The optimal paths module solves what the litera-ture refers to as the tactical service network designproblem. Crainic (2000) provides an overview of solu-tion methods to solve this problem. At TNT Express,we implemented a mixed-integer programming for-mulation as proposed in Meuffels et al. (2010b), wherethe number of paths to be generated is restricted. Themovement heuristic module generates movementsbased on the paths of the packages using the heuristicthat Meuffels et al. (2010b) suggest. In the heuristic,any of the following three rules are used to schedule amovement: (1) when all packages are available at thedeparture location, (2) when a full vehicle movementcan be created, or (3) because of the time restrictions.To generate tours based on a movement schedule,the tour generation module uses a set-partitioningapproach, as described in Van Krieken (2006).

Implementation of TRANSPrior to implementing the TRANS solution, TNTExpress analysts were using spreadsheets to conductnetwork optimization analyses. However, because ofthe large size and complexity of the networks, theycould analyze only small parts of the networks, whichinevitably led to suboptimal solutions. Given that ouranalysts had been using spreadsheets for many years,we were reluctant to completely change their waysof working. Instead, we decided to develop a solu-tion that was close to their spreadsheet environment,extended with several decision support modules to

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


enable the analysts to work faster and more effec-tively on improving the performance of the networks.

The requirements and business logic for TRANSwere developed and discussed in the CoPs and builtby ORTEC and Tilburg University. The first modulesdeveloped were the service capability analyzer andthe routing module, which our analysts used to iden-tify, visualize, and apply step-by-step improvementsto our networks. This step-by-step approach led tostrong user acceptance for the tool. The GO academyand the central GO team were important enablersin creating awareness of and implementing this newway of working to improve network performance inour business units worldwide. TRANS has about twodozen users (members of the central network analysisteam of a business unit or country) worldwide. As ouranalysts became more familiar with using quantita-tive models, we gradually increased the level of opti-mization complexity deployed. Working with TRANShas become a best practice at TNT Express.

Benefits of TRANSBy using KPIs of the whole network, we gained manyinsights that we could not gain by using spread-sheets. By carefully considering factors such as emptykilometers and movement utilization, the analysts

Figure 3: The service capability analyzer can be used to improve service capabilities, as this figure illustratesfor the depot in Barcelona, Spain. On the left, the figure shows the old situation (i.e., prior to using this analyzer)with the locations for which a one-day service from Barcelona is available; on the right, the figure shows the newsituation (i.e., after using the analyzer) in which more locations and regions have a one-day delivery serviceavailable from Barcelona.

were able to search for the most cost-effective meansof transporting the packages.

The service capability analyzer was frequently usedto compare our commercial service offering and thecapabilities of a network. Figure 3 shows an exam-ple of service improvements for the Barcelona, Spaindepot. Our analysts discovered that some serviceoffering deadlines were too tight, resulting in low lev-els of customer service. By analyzing these types ofscenarios in advance, we were able to avoid sellingunachievable services to customers.

With the routing module, analysts can quickly eval-uate changes to the movement schedule and eas-ily apply any adjustments (e.g., alter a movement’stime frame, adjust the vehicle type of a movement,change a movement’s start or end location, or removea movement from the schedule). Because it can gen-erate more efficient tours, the tour generation modulesubstantially reduced subcontracting costs and CO2

emissions. We recently introduced the optimal pathsand movement heuristic modules, which we are cur-rently piloting in Italy; the preliminary results areencouraging.

Thus far, we have used TRANS to analyze approx-imately 15 road networks. From 2008–2011, weaccrued cost savings of 48 million euros, reduced

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


kilometers driven by 69 million, reduced CO2

emissions by 44 million kilograms. Some countriescontributing to these savings used only the servicecapability analyzer and routing modules—not thefull suite of modules described above. Based on theachievements to date, we anticipate additional sav-ings in the near future, particularly as the use of themore advanced optimization modules becomes morewidespread.

In addition to these quantitative benefits, weachieved a number of qualitative improvements, par-ticularly in the area of service provision. This typeof solution reduces the analysts’ workload and pro-vides them with opportunities to improve their anal-ysis quality and their way of working. An exampleis the creation of annual schedules. Prior to usingTRANS, analysts created a single annual schedulethat included exceptions during peak periods. WithTRANS, analysts can now create multiple schedules,including ones that can cope with the volume dif-ferences between workdays and weekends. Finally,TRANS changed the mind-set of both analysts andmanagers: each group now focuses on searching foradditional opportunities to optimize the network.

Subprogram 2: Tactical Planning inPickup and Delivery (SHORTREC)At TNT Express, PUD, which is planned at the depotlevel, impacts the first and last mile in the supplychain. Given that PUD accounts for more than 30 per-cent of operational costs, it is an important focus areaof the GO program. At TNT Express, a round corre-sponds to a single vehicle starting at the depot, vis-iting customers in a certain sequence for collectionor delivery of packages, and returning to the depot.Customer PUD rounds are determined during tacticalround planning. Effectively organizing the PUD pro-cess is challenging because millions of packages mustbe picked up and delivered each week.

The optimization problem in the PUD process is tominimize the total pickup and delivery costs whilemeeting all service-level requirements. This impliesminimizing the number of rounds (fixed cost) and thekilometers and hours driven for each round (variablecost). In PUD optimization, the depot locations andtheir cutoff times are fixed. Constraints that must be

considered are vehicle capacity, service levels, driverregulations, and some softer constraints to ensurerepetitiveness in the rounds and workload balanc-ing. From an operational point of view, it is impor-tant to ensure that the daily pickup and deliveryrounds remain consistent to prevent (1) disruptionsto the sorting and loading processes at the depots,(2) increased workload, and (3) potential errors. Thecreation of consistent rounds increases customer sat-isfaction because the same driver visits the customereach time and can establish a positive working rela-tionship with that customer. However, generating andexecuting similar rounds for each day of the week isdifficult because of volume fluctuations. To deal withthese challenges and support our analysts in PUDoptimization, we implemented a modified version ofORTEC’s advanced vehicle-routing and optimizationsoftware, SHORTREC.

Operations Research Techniques in SHORTRECIn logistics, the problem of generating rounds at min-imum cost is known as the vehicle-routing problem(VRP), which Dantzig and Ramser (1959) studied first.Golden et al. (2008) provide a more recent overview.Given that the VRP problem is NP-hard, only smallinstances can be solved to optimality. Because ofour problem sizes (e.g., the Rome depot handles90,000 stops per week), combined with the additionalnonstandard constraints to enhance productivity atthe sorting facilities, we selected a heuristic optimiza-tion approach in SHORTREC.

To ensure that the rounds are sufficiently robust tohandle volume fluctuations and to ensure a balancedworkload across rounds, we introduced the conceptof �-zones. A �-zone is a geographical area compris-ing a set of customer visits and a total average work-ing time (i.e., drive time plus stoppage time) withina specific time bucket (e.g., an hour). Our preferenceis to establish visually attractive (i.e., nonoverlappingand convex) �-zones. In the PUD process, a roundtraverses a series of neighboring �-zones; each cus-tomer location in the �-zone is visited, while driverregulations are satisfied. Volume fluctuations lead tochanges in working time within a �-zone (and, as aresult, the working time of the round) and the uti-lization of the vehicles. By reassignment of �-zonesto rounds, the change in working time of the round

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


Figure 4: The figure shows �-zone construction in SHORTREC for a part of Slovenia. On the left side, the oldsituation (i.e., prior to using SHORTREC) shows the stops to be clustered; on the right side, the new situation(i.e., after using SHORTREC) shows the resulting clusters—each with a different shading.

can be absorbed with minimal change to the over-all round structure, preventing large changes in thedepot sorting process. In cases of large fluctuationsin volume, new rounds are added or removed. The�-zones are created using the k-means clustering algo-rithm (Kärkkäinen 2006) and can be clustered intoseparate, geographic territories, each one served by asingle PUD round (see Figure 4).

To ensure that a PUD round is both feasible andcost effective, the round is evaluated in detail andoptimized using the local search improvement algo-rithms in SHORTREC. The round might be improvedby changing the sequence of customer visits withinor between the �-zones. If required, �-zones canbe exchanged between rounds. For an extensivedescription of the improvement algorithms used inSHORTREC, refer to Kant et al. (2008).

Because of the size of the problem instances,calculating distances and driving times was anotherchallenge. The normal procedure is to calculate thedistance and driving times, store the results in mem-ory, and construct the rounds. However, this processrequires a huge amount of memory (recall that theRome depot has 90,000 stops per week). By usingthe concept of highway-node routing (Schultes 2008),

a method that exploits the layered structure ofdigital road networks, we were able to reducecomputation times and memory consumption, allow-ing for on-demand calculation of distance and drivingtimes. As a result, depot managers and analysts canoptimize large instances on a normal desktop com-puter using SHORTREC.

Implementation of SHORTRECGiven that TNT Express has more than 2,000 depots,it was infeasible to implement SHORTREC in alllocations at short notice. Therefore, we set a goal ofoptimizing all the rounds in the depots of our mainbusiness at least once a year. Local staff and the cen-tral GO team implemented the optimization projects.To build trust and overcome resistance, the first stepin the standard SHORTEC implementation procedurewas to model the existing round structure of a depot.Supported by the graphical capabilities of SHORT-REC, a member of the central GO team demon-strated to the depot manager that the SHORTRECresults using the current rounds were comparable tothe actual costs, kilometers driven, and round struc-ture. Schedule improvements were then generatedusing a standard set of optimization scenarios. This

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


standardized approach was developed in conjunc-tion with the CoP members, carefully documented,and tested in various countries. The set of scenar-ios describes different ways of working in PUD andcontains various scenarios, including evaluating com-bined PUD rounds, combining the pickups and deliv-eries of different types of packages (with regard toservice and volume), or analyzing the cutoff timesat the depot. Each depot must apply this standardset of scenarios, allowing each to reach a high levelof optimization. This standardized implementationapproach enabled us to rapidly deploy SHORTRECand disseminate PUD optimization knowledge withinthe organization.

Benefits of SHORTRECSHORTREC has been deployed successfully in manyEuropean countries, including the UK, Germany,The Netherlands, Belgium, Italy, France, Spain,Austria, Denmark, Norway, Sweden, Portugal, andGreece, and has been used in various optimiza-tion projects worldwide. This deployment is ongoingbecause our ultimate goal is to optimize each depotat least once a year.

During 2008–2011, six percent of the depots inEurope (260,000 rounds) have been optimized, result-ing in 25 million euros in cost savings and anestimated accumulated CO2 reduction of 11 millionkilograms. With SHORTREC, we are now able toweigh the additional cost of creating visually attrac-tive PUD rounds against the advantages that can beachieved in the sorting process at the depots. More-over, because of the improved quality of the rounds,customer service has improved. Furthermore, dailyimprovements in terms of our ability to cope withvolume fluctuations have been achieved as a resultof the generated �-zones. By using OR techniques,we are able to easily adapt decisions on the structureof rounds to absorb volume fluctuations by simplyexchanging �-zones between rounds.

Subprogram 3: Supply ChainOptimization (DELTA Supply Chain)Because of the worldwide financial crisis at the endof 2008, TNT Express faced a strong decline in vol-ume that continued until mid-2009. This drop in

volume and associated revenue brought about anabrupt decline in air network performance, a prob-lem that required an immediate solution. Because theair network forms a crucial part of our global serviceoffering, we were impelled to start an end-to-end sup-ply chain optimization project that would reduce air-craft use, preserve future growth capabilities, but notworsen service. Based on the achievements of the GOprogram up to that time, we realized that we neededa tool that could support us in making strategic deci-sions and bring fact-based decision making to theboard room. We decided to build the DELTA sup-ply chain model, which would include every relevantdetail of our supply chain and focus on reducing air-craft use as its first priority. Using the results of thismodel gave us the insight that we could decommis-sion 12 of 59 airports and open one new airport, thussignificantly reducing air transportation costs withminimal impact on customer service. More impor-tantly, the results enabled us to survive the financialcrisis. Stimulated by this success, the DELTA supplychain model became an important instrument in thedevelopment of our board’s Vision 2015 strategy.

The DELTA model enables us to optimize ourcomplete supply chain for a fixed depot and hubinfrastructure under varying volumes and ways ofworking (e.g., cutoff times, road and air transport).To the best of our knowledge, this is the only model inthe express delivery industry that covers a completeair and road supply chain. We decided not to buildone integrated model, but to design specific submod-ules to separately optimize the key components of thesupply chain. By choosing to model the supply chainin this way, we could more easily understand and relyon the model’s capabilities, which led to increasedsupport for the decision-making process.

Operations Research Techniques in the DELTASupply Chain ModelA typical DELTA model run begins with a volume-demand scenario. Cutoff times are imposed to securethe times required for both the PUD and network pro-cesses. The model aims to use road transport ratherthan air transport because the former generally resultsin lower costs and CO2 emissions. The road networkmodel determines the shortest paths for the pack-ages by using the locations that may be visited as

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


input. To incorporate network timing effects, we usehub time windows for the sorting activities, settingthe latest arrival and earliest departure times to andfrom the hubs. Based on these time windows, we areable to determine if the service requirements for thespecified packages can be met via road transporta-tion. For packages that can be shipped by road, thenumber of required movements is calculated basedon the routings of the packages. Any empty move-ments are determined using a classical transportationmodel that calculates the number of empty reposition-ing movements to estimate the cost of repositioning.

For the packages that are unable to meet the ser-vice requirements via the road network, we constructan air network using a separate model to create aminimum-cost air schedule between the airports inthe network. The model starts by assigning depots tothe airports in the air network based on one of twocriteria: (1) best service (i.e., latest departure from, andearliest arrival at, the depots, based on the predefinedearliest departures or latest arrivals at the airports),or (2) lowest cost (i.e., shortest distance between theairport and depot). Based on these assignments, themodel determines the packages to be transportedfrom the airport to the air hub and vice versa. Next,a mixed-integer programming problem is solved todetermine the minimum-cost air schedule. The modelensures that sufficient aircraft capacity is available tocarry all the packages, and it balances the numberof incoming and outgoing aircraft per aircraft typeat each location. The aircraft that can be used arerestricted by a minimum and maximum number peraircraft type and the aircraft operating characteristics,such as maximum flying range, effective speed, cargocapacity, and landing restrictions. For airports, themodel includes the following: the earliest permittedarrival or departure times, airport closing times, andthe consideration that some airports do not permitmultiple stops by TNT Express airplanes. Finally, forthe air hub, the sorting time window is included, set-ting the latest arrival and earliest departure time forthe aircraft, and the runway capacity at the air hub.

This model is based on the work of Armacost et al.(2002), with some additions to capture the specificsof the TNT Express operation. One main differenceis the restriction of the number of stops at an air-port. At some airports, we strongly prefer that all

packages arrive and depart via one aircraft becausethis simplifies the handling process. A second differ-ence is the inclusion of more detailed modeling of therunway capacity at the air hub. Instead of restrict-ing the total number of arrivals and departures atthe air hub, we would like to position them acrosstime. Our approach is similar to the work of Barnhartand Schneur (1996). We also include the functionalityto use a minimum or maximum number of aircraftper type, which may be used in the European AirNetwork, to cope with restrictions on fleet availabil-ity. For some situations, we even would have func-tionality to fix specific aircraft operations betweentwo airports in the network. This request originatedbecause the European Air Network is sometimes usedin a combined setting for domestic network opera-tions (in which TNT Express-owned aircraft are used).

With the road and air network complete, a binaryinteger programming model estimates the impact ofthe network movement arrival and departure timeson the PUD cost. This model determines the opti-mal wave structure of every depot. We define a waveas a set of rounds that start and end at the sametime at the depot. In the case of multiple arrivals ata depot, starting part of the rounds before all pack-ages have arrived might be beneficial because thisensures that the rounds have a longer working day.This is particularly useful when packages destined fornearby customers might arrive at the depot at dif-ferent times. If so, multiple rounds will have to beassigned to the same location area, resulting in largeraverage distances between stops. Similar advantagesand disadvantages apply to collections whereby mul-tiple departures occur in the network. The binary inte-ger model determines the optimal number of wavesrequired to balance both the length of rounds in awave and the extra kilometers to be driven. In a finalstep, the model calculates the total cost and serviceKPIs of the complete supply chain to support man-agement decision making.

DELTA Supply Chain ModelImplementation ChallengesBecause of the board’s tight schedule to rationalizeair operations in 2008, substantial work was requiredwithin a very short time frame. Some team mem-bers were requested to completely clear their agendas

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


of other activities for a number of weeks. A cross-functional strategic operations team, including mod-eling and optimization specialists from both ORTECand Tilburg University, was set up. The team mon-itored the progress of the project and became theplatform for discussion and agreement on the manyissues encountered during the project. The develop-ment of the DELTA supply chain model was a diffi-cult task because the model covers all of Europe andconsists of about 650 depots, 90 hubs, and 150,000origin-destination combinations. The challenge was toprovide the right level of detail to support the boardin its decision making, but not so much detail asto render the results useless and distract the boardfrom crucial insights and decision factors. Fortunately,work done in the previous years in each GO subpro-gram provided us with the experience we needed toagree on the relevant details of the model. However,because many team members were not yet familiarwith strategic modeling, some members wanted toincorporate much more detail than was needed. Asa result, a great deal of discussion and salesmanshipensued so that the team members would agree on theappropriate (strategic) level of detail that would beacceptable to each stakeholder.

Data gathering was another major challenge. Ourexperience led us to believe that the data gatheringand verification exercise for this type of analysis couldtake months, even for one country. We were taskedwith acquiring data for all of Europe within onlyeight weeks.

Convincing people to execute the model was not anissue, because its major users were people in the strat-egy department of GO and consultants of ORTEC;however, building the decision makers’ trust in usingthe model results was a challenge. They often did notimmediately accept the initial results, mainly becausecertain details had been omitted from the calculations;however, more importantly, the new insights gainedfrom the DELTA supply chain model violated theirprior beliefs about operating the TNT Express supplychain. To show that the model’s results were relevantand consistent, we formulated a number of scenar-ios, evaluated them, and explained them in detail.For example, to determine which airports to close,we generated and optimized more than 20 scenar-ios. Although the model showed consistent results

for these scenarios, the air network analysts recalcu-lated the results with very similar outcomes. As trustin the DELTA supply chain model grew, the TNTExpress board became more confident in the resultsand decided to move forward based on the insightsgained. It implemented the model results and usedthem to build its Vision 2015 strategy.

Benefits of the Strategic AnalysesThe DELTA supply chain model has become a vitalinstrument for generating and analyzing air networkoptimization scenarios for various airport composi-tions. In 2008 and 2009, we opened one new airportand closed 12 of our 59 European airports (see Fig-ure 5); the impact on service levels was minor—lessthan 0.5 percent of the volume arrived more than onehour later. Furthermore, we also eliminated six air-craft, three of which were expensive A300 aircraft.Naturally, this incurred some additional costs becauseof the longer distances driven between airports anddepots. However, total net accumulated savings were132 million euros and the CO2 emissions reductionwas 228 million kilograms. Achieving these reduc-tions within such a short period improved our agilityand ability to create value, even with volatile demand.

To develop the TNT Express Vision 2015 strategy,a number of operating modes and European net-work designs were evaluated using the DELTA sup-ply chain model. Various scenarios were analyzed(e.g., separating parcel and freight volumes in Europe,reducing stop-time in PUD, altering the available timebetween PUD and network operations by varyingthe cutoff times, and investigating the robustness ofour road and air networks). The insights gained fromthe DELTA supply chain analysis strongly contributedto the development of our strategic vision. Combin-ing volumes (i.e., parcel and freight) resulted in acost avoidance of four percent on supply chain costs.Although this insight was initially counterintuitive,analyzing the cost details resulting from the modelconvinced us of the correctness of the outcomes.

The DELTA supply chain model allows us to testour supply chain operation improvement ideas with-out having to experiment in practice. For example,when we used this model to calculate the impact of

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


Figure 5: The figure shows the final result after management decisions; management decided to close 12 airportsand begin TNT Express operations at one new airport.

changing the cutoff times, it provided insights on thetrade-offs between network costs and PUD costs thatwe would have to make. As a result, we initiatedfollow-up projects to analyze and change the cutofftimes for potential new locations in our network. Themodel also indicated that substantial benefits could beachieved by increasing the capacity of certain depotsand hubs. The reported benefits did not include thebenefits of the Vision 2015 projects.

Finally, after observing the potential improvementsthat could be achieved by optimizing the balancebetween the PUD and network processes, we startedto develop models to optimize the depot and hubinfrastructure. Initial results show that the savingspotential is enormous; in the coming years, these sav-ings will be achieved gradually because changing theinfrastructure quickly is impossible. We will continueto use the DELTA supply chain solution regularly tosearch for large improvements.

General Deployment ChallengesImplementing tools such as TRANS and SHORTRECtakes time and effort. First, it takes time to convinceanalysts, network managers, and depot managers,most of whom are unfamiliar with optimization, toadopt new tools that they might initially see as reduc-ing their control of the analysis. Second, the availabledata, although numerous, were spread across manylocal information technology (IT) systems, thus reduc-ing the quality and quick availability of the datarequired. Some data (e.g., the delivery or pickupaddress) must be detailed, and the data received areoften either incomplete or incorrect. This problem waspartially solved by introducing GO data management,a data cleansing and conversion tool that makes thedata retrieval from our IT systems repeatable. In par-ticular, this tool describes a set of business rules tomap the source data onto the GO data structures; thisis an evolving process because business rules or datastructures sometimes change. Furthermore, we real-

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


ized that significant effort is required to create a user-friendly and fast model that supports the analystsin evaluating various planning scenarios and under-standing the differences between them.

When developing decision support solutions, weknew that we would be setting the standard for allcountries in which TNT Express operates. This provedchallenging because these countries have many waysof working, variances in volume profiles, and localregulations, any of which could impact our creationof a standard model. For example, TNT Express usesloose-loaded trucks in Italy, pallets in France, andcages in the European road network. TRANS had tosupport all these requirements; in addition, we had tofind the right balance between generic and country-specific requirements in each instance.

Another challenge we faced was objectively track-ing results. Managers who successfully meet theiroptimization targets often receive a budget decreasein the subsequent year. This could affect the qual-ity of the submitted results, especially because themanagerial bonus system is based on the budget tar-gets. Even a slight deviation from the optimal solu-tion in practice may increase cost, making it difficultto attain the bonus targets. Therefore, to measure theresults objectively, we introduced a benefits trackingsystem that uses three levels of savings to monitor thebenefits: identified expected savings, agreed savings,and implemented (realized) savings. The savings pre-sented in this article are the implemented (realized)savings; cost avoidance is not included.

GO AcademyMost new users were resistant to change, even whenwe could demonstrate successes in nearby operatingunits, because they felt that their business was notcomparable. Furthermore, the task of explaining thegeneral optimization principles of the tools was timeconsuming. These two challenges led us to imple-ment the GO academy, a unique learning concept inoptimization.

The main objective of the GO academy is to teachoptimization principles to TNT Express employeesand to acquaint them (at a high level) with theavailable optimization tools, without turning theminto mathematicians. We discovered that one of the

Module Name Topics

1 Introduction module Customers and their supply chains2 Strategic optimization Infrastructure design (DELTA)3 Networks and PUD Planning in networks and PUD

(TRANS and SHORTREC)4 Hubs and depots Bottleneck theory, mechanization

principles5 Implementation Change management techniques6 Graduation Presenting for impact, elevator pitches

Table 1: The six GO academy modules train participants in optimizationprinciples and practices used by TNT Express. The table lists the modulesand the topics included in each module.

most important lessons we can teach is that invest-ing a little in one part of the supply chain can resultin large benefits in other parts. Another principlewe taught is that to strategically optimize a supplychain, considering all details is unnecessary. Third, weteach that the combinatorial explosion forms the basisof many frequently encountered planning problems.We use various methods to teach these principles;these range from conceptual explanations and prac-tical assignments to simple but powerful computergames. An example of the latter is the GO game, inwhich a tactical and strategic solution for an expressnetwork must be constructed; Meuffels et al. (2010a)contain a description of this game. To date, over 500managers and staff, including the TNT Express boardof directors, have successfully completed this game.

The GO academy training program consists of sixthree-day modules conducted over a period of twoyears, interspersed with small group assignments(see Table 1). After a group completes each mod-ule, the group composition is changed to promotenetworking—a key benefit of the academy. After com-pleting the fifth module, students are given two daysper week for six months to complete their final assign-ment, a master case study. The case study, spon-sored by one or more senior managers and guided byan academic supervisor, is based on an actual chal-lenge at TNT Express. Students present the mastercase results on graduation day. Within the operationsarena of TNT Express, this day has become a big net-working event that most of our senior managers andour CEO, at times, attend.

In addition to optimization skills, the GO academytraining program focuses on the development of

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


personal and interpersonal skills: presentation skills,debating ability, working on camera, and elevatorpitches.

Graduating employees are designated as supplychain masters, an internal title that remains in effect aslong as the employee completes at least one optimiza-tion project per year. Project results, suggestions, andideas are published on an internal website, Collabo-rate. These projects are judged as part of an annualsupply chain master competition; on graduation day,an award is given for each category (i.e., networks,PUD, and supply chain).

Since its inception, the GO academy has success-fully met its main objective—to teach the principlesof optimization. We frequently find that it is unneces-sary to convince people of the benefits of optimizationbecause they are coming to us for support and advice.This is a definite turn of events and one that provesthe effectiveness of the training academy concept.

Furthermore, the supply chain master case stud-ies have delivered a number of significant benefits inseveral areas. In addition to cost savings of approx-imately 5.7 million euros, networking within TNTExpress has improved dramatically. After collaborat-ing on assignments during their two years in theGO academy program, employees build strong rela-tionships with each other; they also feel more empow-ered to ask for support from colleagues in otheroperating units or even in other parts of the world.Employees now use the same business language andhave the same understanding of definitions and ter-minology (e.g., cutoff time, �-zone). At TNT Express,it has become apparent that platforms such as the GOacademy are ideal arenas for discussing and explain-ing the operational implication of strategy changes.The GO academy has exceeded our expectations.

TransportabilityThe above lessons are applicable and transportableto any organization that wants to apply and embedOR on a large scale. To illustrate the transportabil-ity of our approach, we outline our contribution tothe World Food Program (WFP), the world’s largesthumanitarian aid organization, which feeds morethan 90 million of the poorest people on earth. In con-junction with the WFP, we developed a simple hub-and-spoke network for food distribution for Ethiopia.

The feeding of children in more than 2,000 schoolsin Liberia has been optimized with SHORTREC andhas yielded 10 percent savings on transport costs. Weare proud that the WFP is actively involved in ourCoPs and GO academy. We currently are sharing ourcore ideas of supply chain optimization with the WFPstrategic logistic team in Rome.

ChallengesThe introduction of optimization to TNT Expresshas been and will continue to be an intensive pro-cess with many challenges. Of course, data avail-ability and quality are always an issue in an ORproject. In addition, more decision makers need tobecome familiar with optimization principles, espe-cially in cross-functional areas such as marketing,sales, and finance. Applying OR principles at theboard level brings another set of challenges. Given theinvolvement of people from varying backgrounds, werequired a great deal of discussion to determine theright level of detail for strategic modeling. Gainingacceptance of our approach required a high level ofdidactical skills and salesmanship. The most signifi-cant challenge related to the time pressure because ofstrict time lines (e.g., shareholder meetings, executiveboard meetings) and short decision time frames.

Concluding RemarksThe best way to apply OR in a business that is unfa-miliar with the concept of optimization is to start sim-ple and follow the maturity level of the business inapplying more advanced methods. Forcing the use ofOR—if it is not well understood—will only increaseresistance and decrease user acceptance. The abilityto visualize the business challenge contributes signifi-cantly to lowering resistance because users are able tomore easily recognize the challenges. Using basic ORtechniques from the start makes possible the build-ing of trust and understanding, and improves dataquality. Most of the savings and CO2 reductions dis-cussed in this paper can be attributed to these basicOR solutions.

As soon as people become more familiar with ORmodeling, advanced techniques (e.g., scenario analy-sis, simulation, and mixed-integer programming) canbe introduced. The development of OR models andtools in close collaboration with the business leads to

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


increased trust and acceptance by end users. We facil-itated model and tool development by introducingCoPs, in which our subject matter experts and exter-nal OR experts derived and honed the requirementsof the solutions to be developed.

Parallel to developing tools, we realized theimportance of teaching employees the fundamentalsof optimization to encourage the dissemination ofknowledge and allow for fast implementations. As aresult of the GO academy, a huge network of peoplewho recognize optimization possibilities and whoseknowledge the company can easily tap into existswithin TNT Express. We initially set up our centralGO team with five people and currently have about30 people whose full-time jobs involve applying ORat TNT Express; senior management and over 200supply chain masters support them. Optimization hasbecome part of the core values at TNT Express.

AcknowledgmentsWe thank all the people who contributed to the success ofthe GO program and the GO academy. Moreover, we arevery grateful to the people who gathered the data presentedin this paper. We give special thanks to Yoshiro Ikura andAndres Weintraub for their active guidance as our coachesin the Franz Edelman Award Competition.

ReferencesAhuja R, Magnanti T (1993) Network Flows: Theory, Algorithms, and

Applications (Prentice Hall, Englewood Cliffs, NJ).Armacost AP, Barnhart C, Ware KA (2002) Composite variable for-

mulations for express shipment service network design. Trans-portation Sci. 36(1):1–20.

Barnhart C, Schneur RR (1996) Air network design for express ship-ment service. Oper. Res. 44(6):852–863.

Crainic T (2000) Service network design in freight transportation.Eur. J. Oper. Res. 122(2):272–288.

Dantzig GB, Ramser JH (1959) The truck dispatching problem. Man-agement Sci. 6(1):80–91.

Golden B, Raghavan S, Wasil E, eds. (2008) The Vehicle RoutingProblem: Latest Advances and New Challenges, 1st ed. (Springer,New York).

Kant G, Jacks M, Aantjes C (2008) Coca-Cola Enterprises opti-mizes vehicle routes for efficient product delivery. Interfaces38(1):40–50.

Kärkkäinen I (2006) Methods for fast and reliable clustering.Accessed April 16, 2012, http://epublications.uef.fi/pub/urn_isbn_952-458-816-1/urn_isbn_952-458-816-1.pdf.

Meuffels I, Fleuren H, Cruijssen F, Van Dam E (2010b) Enrich-ing the tactical network design of express carriers with fleetscheduling characteristics. Flexible Services Manufacturing J.22(1–2):3–35.

Meuffels I, Fleuren H, Poppelaars J, Hoornenborg H, De Rooij F(2010a) The design of express networks in a nutshell—Playingthe global optimisation game (GO game). OR News 39(2):6–8.

Root S, Cohn A (2008) A novel modeling approach for expresspackage carrier planning. Accessed April 16, 2012, http://deepblue.lib.umich.edu/bitstream/2027.42/60965/1/20310_ftp.pdf.

Schultes D (2008) Route planning in road networks. AccessedApril 16, 2012, http://algo2.iti.kit.edu/schultes/hwy/schultes_diss.pdf.

Van Krieken M (2006) Solving set partitioning problems usingLagrangian relaxation. Accessed January 4, 2011, http://arno.uvt.nl/show.cgi?fid=47051.

Hein Fleuren is a full professor in the application ofOR at Tilburg University; he has his own consultancy, ORCoach, and works at the intersection of science and prac-tice. He directed the CentER for Applied Research at TilburgUniversity and was a consultant and partner at the Centrefor Quantitative Methods (CQM BV) in the Netherlands. Heworks part-time, on behalf of the Tilburg Sustainability Cen-ter (TSC), for the United Nations World Food Programmein Rome.

Chris Goossens is managing director of global networksand operations at TNT Express. She is the first woman totake on this role within the company and the express indus-try. She joined the company in 1988 as area sales managerin Belgium. In 2004 Chris led the differentiation strategy forcustomer service, which has achieved the desired results,not only internally but more importantly externally, withTNT recognized as the most customer-focused companyby the European Business Awards 2008. Her educationalbackground is in political and social science. She is a non-executive on the board of Bodycoat and sponsors an inter-nal female leadership network. She is an expert in customerexperience, leading TNT to be recognized in two books forbest practice.

Marco Hendriks joined TNT in 1983 after completinghis education in logistics and transport management. Hedeveloped and became global manager of network strat-egy, followed by a key role in the partnership between TNTand the World Food Programme of the UN, “Moving theWorld.” He is responsible for supply chain optimizationwithin operations globally, including optimization of theentire supply chain and in specific areas of road and airnetworks, pick-up and delivery, and hubs and depots.

Marie-Christine Lombard was Chief Executive Officer ofTNT Express until October 2012. She holds an MBA fromthe ESSEC Business School. Before joining the express ser-vices sector, she was employed in American retail businessand then worked for Chemical Bank and Paribas Bank.

Ineke Meuffels earned a master’s degree in economet-rics and OR from Tilburg University in the Netherlands Shebegan a career as a consultant with the ORTEC Consult-ing Group, a division of ORTEC, one of the world’s largestindependent providers of advanced planning solutions and

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.


OR consulting services. She has worked on many tacticaland strategic projects, particularly in the field of expressnetwork, the subject area of her PhD research at TilburgUniversity.

John Poppelaars is director of consulting for the ORTECConsulting Group. Throughout his 22-year career, he has

applied the business maxim of “improving decision-makingquality” to numerous projects across a myriad of industriesin order to support clients in optimizing their businesseswith OR techniques. He is a frequent invited lecturer at uni-versities and has created a blog, “OR at Work,” where hewrites about the practical application of OR in business.

Dow

nloa

ded

from

info

rms.

org

by [

139.

179.

113.

31]

on 1

8 O

ctob

er 2

014,

at 2

1:26

. Fo

r pe

rson

al u

se o

nly,

all

righ

ts r

eser

ved.

Documents

Supply Chain Wide Optimization at TNT Express...Fleuren et al.: Supply Chain–Wide Optimization at TNT Express Interfaces 43(1), pp. 5–20, ©2013 INFORMS 7 management understood