Successful Control of Major Project Budgets

Transcription

1 admistrative sciences Article Successful Control Major Project Budgets Steen Lichtenberg Lichtenberg & Partners, Baneskellet 16, Vedbæk DK-2950, Denmark; Tel.: Academic Editor: Ole Jonny Klakegg Received: 15 March 2016; Accepted: 29 June 2016; Published: 8 July 2016 Abstract: This paper differs from scientific papers describg current research. In le with me this special issue, it challenges conventional risk management practice agast background former research results successfully fished decades ago. It is well-known that conventional practice frequently results budget overruns large projects. International reviews document that. Severe delays schedules are also well-known. This paper describes successful research results from almost three decades ago, which successfully challenges this severe problem and has led to new practices. The research volved is an unusual mix: Scandavian researchers from psychology, statistical ory and engeerg economy. The resultg procedure has been widely used sce around 1990 and challenges conventional procedures. The procedure is documented to be able to yield statistically correct prognoses, when rules game have been correctly followed. After a short summary basic situation, this paper summarizes research, followed by some resultg experiences, focusg on two recent studies each 40 frastructures and or major projects. In both sets, actual fal cost largely equaled expected project cost. This result is a marked change from ternational past and present experience. Fally, need for furr research and progress is discussed. Keywords: budget quality assurance; cost estimation; estimation methodology; major projects; risk management; schedulg; Successive Prciple 1. Introduction Severe budget overruns and delays are still common while usg conventional procedures and prciples, especially among larger projects, despite fact that Project Management and Cost Engeerg have made tremendous advances recent decades. Some reasons behd this are discussed below as a brief overview. This as a background to a presentation Scandavian research results that challenges conventional practice, as it has documented to yield correct statistical prognoses costs and/or duration large projects, clusive IT projects. One most referred sources for describg project cost overruns is Standish Group Chaos report [1]. It presents depressg results from a large sample IT projects. This source is not alone claimg that most projects go wrong. Or sources clude Flyvbjerg, Holm, Buhl [2,3], and ir conclusions from an analysis many frastructure projects was that 90% had cost overruns, generally a significant size. They showed that situation has not improved for decades. Merrow [4] documents that situation is not much better private sector: 65% 75% dustrial megaprojects fail on busess targets. These projects are generally large and complex. However, does size matter? Odeck [5] studied this and his results dicate that small projects can have even worse results than large projects cost overruns were even more frequent and relatively larger small projects. To summarize, historically, we know for a fact that projects still have a strong tendency for cost overruns large and small, private and public, when usg conventional procedures. The reasons behd cost overruns have also been thoroughly examed and discussed. Flyvbjerg et al. [2,3] argued that ma problem is that planners and promoters ten deliberately Adm. Sci. 2016, 6, 8; doi: /admsci

2 Adm. Sci. 2016, 6, Adm. Sci. 2016, 6, The reasons behd cost overruns have also been thoroughly examed and discussed. Flyvbjerg underestimate et al. [2,3] costs argued and that risks andma overestimate problem is that benefits planners and order promoters to crease ten deliberately likelihood that ir project underestimate gets approval costs and and risks fundg. and overestimate This view isbenefits supported order byto or crease studies i.e., likelihood that that re are politicalir or strategic project gets reasons approval forand cost fundg. overrun. This Similar view is reasons supported are by identified or studies i.e., by Merrow that re [4] are busess political or strategic reasons for cost overrun. Similar reasons are identified by Merrow [4] busess projects. These reasons need to be understood and handled order to improve project execution projects. These reasons need to be understood and handled order to improve project execution performance. A key requirement is to have good quality, dependent analysis project budget performance. A key requirement is to have good quality, dependent analysis project budget prior toprior go-ahead. to go ahead. A humoristic A humoristic version version story is illustrated Figure Figure 1 below. 1 below. Figure 1. The figures I have given are wrong. That is because I don t have right figures Figure 1. The figures I have given are wrong. That is because I don t have right figures (unknown source). (unknown source). Anor set reasons for cost overruns are known to stem from human judgment, as recently Anor documented set reasons practical for applications cost overruns by are ternational known tofuturaone stem from group human consultants judgment, [6], and as recently excellently explaed by Kahneman 2011 [7]. In 1985, Lange [8] 1 disclosed reasons for this a documented practical applications by ternational Futuraone group consultants [6], and Master s sis. He identified more than twenty pitfalls when makg subjective quantitative excellently explaed by Kahneman 2011 [7]. In 1985, Lange [8] evaluations. A considerable part total project results, total cost or 1 disclosed reasons for this duration, necessarily stems a master s from such sis. expert He evaluations. identified It is more documented than twenty that se pitfalls when may twist makg evaluations subjective severely, quantitative evaluations. and thus A considerable total result. part This research total was an project eye opener results, Scandavia, total cost or and duration, it itiated necessarily a different stems from such way expert thkg evaluations. about how Itcost is documented estimation should that be se carried pitfalls out. may twist evaluations severely, and thus The total research result. also This revealed research that re was an is a eye-opener tendency to focus Scandavia, on tangible, hard and itaspects, itiated and ato different avoid dealg with more subjective ster aspects. Subjective matters have generally been way thkg about how cost estimation should be carried out. considered only a superficial manner, if cluded at all. Today, problem not dealg with The more research subjective alsoaspects revealed is magnified that re by is ahigh tendency level uncertaty to focus onand tangible, change. hard aspects, and to avoid dealg Traditional with more tools subjective and techniques ster aspects. for project Subjective planng matters and management have generally are based been considered only adetermistic superficiallogic, manner, even if though cluded leadg at scholars all. Today, realized problem early on that not this was dealg not sufficient. with more While subjective aspects is magnified well known bycost items highare level estimated uncertaty and calculated and change. all detail, contgency is typically superficially handled, although its importance is still more important. When ory tried to cope with Traditional tools and techniques for project planng and management are based on determistic this uncertaty via traditional methods, it becomes very complicated. These conventional tools logic, even though leadg scholars realized early on that this was not sufficient. While well-known became creasgly challengg, especially for small organizations. Uncertaty was seen as almost cost items are estimated and calculated all detail, contgency is typically superficially handled, although its importance is still more important. When ory tried to cope with this uncertaty via 1 traditionala methods, summary it English becomes appears very [9]. complicated. These conventional tools became creasgly challengg, especially for small organizations. Uncertaty was seen as almost impossible to handle when usg classical statistical ories planng and estimatg. These procedures were eir too primitive or too difficult to use practice. This realization sparked a development 1 A summary English appears [9].

3 Adm. Sci. 2016, 6, Denmark Adm. Sci. and 2016, Norway, 6, 8 which itiated a completely new approach to estimatg project 3 14 budgets and schedule durations. impossible to handle when usg classical statistical ories planng and estimatg. These procedures were eir too primitive or too difficult to use practice. This realization sparked a 2. Untraditional Early Research Efforts development Denmark and Norway, which itiated a completely new approach to estimatg Successful project budgets resultsand schedule schedulg durations. durg 1950s came from use famous Program Evaluation Review Technique (PERT), one well-known Critical Path Methods. It troduced 2. Untraditional Early Research Efforts triple estimate schedulg as a means quantifyg uncertaty (uncertaty ranges, based on Successful results schedulg durg 1950s came from use famous Program mimum, most likely, and maximum values). The same idea was troduced 1970s for cost Evaluation Review Technique (PERT), one well known Critical Path Methods. It troduced estimates astriple a possible estimate answer schedulg to obtag as a means more quantifyg reliable budget uncertaty forecasts. (uncertaty This ranges, research based effort on was led by author mimum, this most paper likely, atand maximum Technical values). University The same Denmark idea was troduced (DTU). 1970s for cost Theestimates researchas soon a possible encountered answer to obtag practical more difficulties. reliable budget How forecasts. should This uncertaty research effort be was handled? led by author this paper Technical University Denmark (DTU). Projects and, not least, ir situations were unique and complex. The many statistical dependencies The research soon encountered practical difficulties. How should uncertaty be handled? between different parts an estimate were difficult to model. However, this problem was solved by Projects and, not least, ir situations were unique and complex. The many statistical dependencies isolatgbetween and defg different all parts significant estimate dependencies were difficult between to model. However, dividual this problem cost items was solved andby cludg m separately isolatg and a defg statistical all model. significant A dependencies base case between estimate with dividual all cost significant items and terdependencies cludg removedm contaed separately dividual a statistical items which model. were A base largely case statistically estimate with dependent all significant each or. terdependencies removed contaed dividual items which were largely statistically dependent This enabled a much simpler statistical approach to be used to calculate result. The fluence each or. This enabled a much simpler statistical approach to be used to calculate result. The key terdependencies, fluence key risks, terdependencies, and opportunities risks, and were opportunities n added were n afterwards added as afterwards Overall Influences as (see Figure Overall 2). Influences (See Figure 2). Figure 2. There is a wide range fluences world that a project will be executed some Figure 2. There is a wide range fluences world that a project will be executed some positive, some negative. Sce potential effects se overall fluences cannot be calculated by positive, conventional some negative. methods, Sce y are potential ten not sufficiently effects considered se overall and handled. fluences cannot be calculated by conventional methods, y are ten not sufficiently considered and handled. Anor problem was to cope with huge amount data that ten existed. This was solved by usg a top down approach stead workg from bottom up. The top down procedure starts with a set around 8 to 12 ma items that cover whole scope, and a similar number variables which represent key terdependencies (i.e., those aspects that could have a systematic fluence on part or all a budget rar than on a particular budget le item). The estimators evaluate triple estimates all se variables (base estimate and overall fluence factors). The total cost is n calculated usg normal statistical rules. The contributions to total uncertaty by each variable (triple estimate) model are shown as an Uncertaty Prile or a top ten list, with largest contributor uncertaty at top. The most uncerta variable at top list is by nature most critical for total result. This element is refore subject to specifyg it greater detail or clarifyg furr. The calculation is n updated and a new Uncertaty Prile or top ten list appears. The procedure contues with furr successive steps (hence term Successive Prciple or Anor problem was to cope with huge amount data that ten existed. This was solved by usg a top-down approach stead workg from bottom-up. The top-down procedure starts with a set around 8 to 12 ma items that cover whole scope, and a similar number variables which represent key terdependencies (i.e., those aspects that could have a systematic fluence on part or all a budget rar than on a particular budget le item). The estimators evaluate triple estimates all se variables (base estimate and overall fluence factors). The total cost is n calculated usg normal statistical rules. The contributions to total uncertaty by each variable (triple estimate) model are shown as an Uncertaty Prile or a top ten list, with largest contributor uncertaty at top. The most uncerta variable at top list is by nature most critical for total result. This element is refore subject to specifyg it greater detail or clarifyg furr. The calculation is n updated and a new Uncertaty Prile or top ten list appears. The procedure contues with furr successive steps (hence term Successive Prciple or Stepwise Method ). Surprisgly quickly, most significant uncertaties are reduced and clarified as much y can be. The end result due to 20/80 rule will typically hold less than a hundred items and factors, ten fewer than The top-down approach enables best possible result to be generated most efficient manner possible. Figures 3 and 4 below illustrate this procedure.

4 Adm. Adm. Sci. Sci. 2016, 2016, 6, 6, Stepwise Method ). Surprisgly quickly, most significant uncertaties are reduced and clarified as much y can be. The end result due to 20/80 rule will typically hold less than a hundred items and factors, ten fewer than The top down approach enables best possible result to be generated most efficient manner possible. Figures 3 and 4 below illustrate this procedure. Adm. Sci. 2016, 6, Figure Ma Ma Ma items items items (A) (A) to to (A) (D) (D) to are are (D) evaluated evaluated are evaluated while while usg usg while triple triple usg estimates estimates triple under under estimates firm firm under normalized normalized firm normalized calculation calculation assumptions, assumptions, calculation assumptions, while while (X) (X) represents represents while (X) all all represents potential potential all external external potential supplementary supplementary external supplementary cost cost effects socalled effects socalled cost effects so-called Overall Overall Influences. Influences. Overall Influences. From From triple triple Fromestimates, triplemean estimates, and and standard standard mean anddeviation deviation standardare are deviation calculated calculated are calculated accordg accordg to accordg to normal normal to rules. rules. normal The The rules. hatched hatched Thesquares hatchedillustrate squares illustrate standard standard standard deviation deviation deviation second second order second order variance. variance. variance. They They represent represent They represent relative relative relative criticality, criticality, criticality, because because because variance variance variance total total equals equals total equals sum sum sum all all local local all local variances. variances. variances. Item Item Item C, C, beg beg C, beg most most most critical critical critical uncertaty, uncertaty, uncertaty, is is furr furr is furr specified. specified. specified. This This This shown shown is shown Figure Figure Figure Figure 4. As illustrated Figure 3, (C), beg most critical, is now specified to sub items. Figure As As illustrated Figure Figure 3, 3, (C), (C), beg beg most most critical, critical, is now is now specified specified to sub to items. sub items. A triple A triple estimate is made each se. The calculation is updated with se new estimates. The next estimate triple estimate is made is made each each se. se. The calculation The calculation is updated is updated with se with se new estimates. new estimates. The next The next step step step will will be be a specification specification (X), (X), now now beg beg most most critical, critical, to to a set set dependent dependent external will be a specification (X), now beg most critical, to a set dependent external potential external potential potential effects effects Overall Overall Influences. Influences. Refer Refer to to examples examples Figure Figure When When all all domatg effects Overall Influences. Refer to examples Figure 2. When all domatg uncertaties domatg uncertaties uncertaties are are handled handled followg followg steps, steps, calculation calculation procedure procedure is is falized. are handled followg steps, calculation procedure is falized. falized. prototype Successive Prciple procedure was launched locally Denmark durg A prototype Successive Prciple procedure 1970s and published by Lichtenberg [10,11]. Its was obvious launched advances locally soon Denmark made it durg widespread 1970s and published by Lichtenberg [10,11]. Its obvious advances soon made it widespread Denmark, but it suffered from severe failures due to many pitfalls when estimators make subjective evaluations. The results suffered from this and use gradually faded out Denmark. However, durg 1980s, problem was examed and fally solved after pioneerg work

5 Adm. Sci. 2016, 6, Lange [8], when he identified more than twenty psychological pitfalls relatg to estimatg from a study ternational psychological literature, toger with providg prelimary advice on how to overcome m. Furr development effort came about via a collaboration between researchers from statistical ory, psychology and engeerg economics at Technical Universities Denmark and Norway, led by author Denmark and late pressor Reidar Hugsted and his team Norway, collaboration with pressor Psychology at Oslo University K. H. Teigen [12,13]. Today, this subject is represented by pressor Ole Jonny Klakegg at Technical University Norway Trondheim [14]. Norway, for some years, has been leadg partner. Durg 80s, this research largely solved remag problems with method, refer to Aass et al. [15]. 3. A Brief Overview Fal Research Results 3.1. General The research effort between 1970s and 1980s is foundation this article. It constitutes a unique symbiosis between statistical ory, psychology and project economy which Bayesian statistical ory, toger with a consistent drive towards statistical dependency between uncerta data, is combed and supported by sub procedures which keep control over many different psychological pitfalls. Until now, y have distorted results any or models toger with project economy, and its top down calculation prciples. The Successive Prciple was origally a tool for early phase cost estimatg and schedulg construction dustry, and it was known by some users as telligent cost estimatg. It has now developed to an tegrated management strument, which is also used to augment productivity and competition. The Successive Prciple is based on four cornerstones: (1) Accept uncertaty as a significant and exitg issue, and follow its rules; (2) Carry out telligent, unbiased evaluations balanced groups; (3) Work top-down, developg quantitative model successive steps by specifyg most significant uncertaty more detail after each step; (4) Identify and clude Overall Influences On Schedules The results area schedulg seem to be successful as well. Several project schedules large IT projects and or complicated projects have been analysed accordg to se prciples. Surprisgly, many se have met ir deadle or were fished accordg to prognoses. However, no proper documentation exists that results have been an effect analyses or that y had or reasons. Quality assurance schedules is a basic application. The origal PERT procedure for schedulg from 1950s faded out. Most likely, among or reasons, due to fact that it neglected statistical dependency among activities, and thus seriously biased results. However, idea evaluatg activities a network usg triple estimates was already cluded early research leadg to Successive Prciple (refer to Lichtenberg & Møller [16]). The concept safeguardg statistical dependency among all uncerta elements and clusion Merge Event Bias calculations gradually leads to successful results. They, toger with cost applications, were launched around Unfortunately, scientific references from this research are not available today, except for reference Archibald & Lichtenberg [17]. Schedules are basically analysed accordg to same procedure that is described for cost analyses, but stead a cost break down structure, a high level logically-lked critical path network is used as basis for analysis. Typically, it consists ma activities or groups activities which represent base case schedule. Additional activities are n added at end or at relevant pots network to model timely effect from Overall Influences upon base schedule,

6 Adm. Adm. Sci. 2016, Sci. 2016, 6, 8 6, is used as basis for analysis. Typically, it consists ma activities or groups activities usg which qualitative represent Uncertaty base case Analysis schedule. asadditional a guide (see activities Supplementary are n added Materials, step end Cor and at D). Thisrelevant part pots procedure network runs exactly to model as cost timely analyses effect from described Overall below. Influences upon base schedule, Durg usg followg qualitative part Uncertaty procedure, Analysis as duration a guide (see all Supplementary physical activities Materials, are evaluated step C and D). This part procedure runs exactly as cost analyses described below. under base case conditions, while effects from Overall Influences are evaluated accordg to Durg followg part procedure, duration all physical activities are evaluated opportunities and risks each category Overall Influences. under base case conditions, while effects from Overall Influences are evaluated accordg to A schedule uncertaty analysis normally uses Monte Carlo simulation order to identify and opportunities and risks each category Overall Influences. clude A schedule degree uncertaty criticalityanalysis and Merge normally Event uses Biases. Monte This Carlo last simulation concept covers order to identify fact that and when twoclude or more uncerta degree near criticality and paths Merge merge, Event anbiases. additional This last delay concept materialises. covers fact that when two Theor top-down more uncerta procedure near critical is alsopaths used merge, here, but an for additional schedule delay analyses, materialises. any specification usg Successive The Prciple top down takes procedure form is also used a sub-network, here, but for except schedule specifications analyses, any specification Overallusg Issues. Successive The Uncertaty Prciple Prile takes or top form ten list a sub network, are less simple except here, specifications as it is calculated Overall as product Issues. degree The criticality, Uncertaty measured Prile or percent, top ten list andare less local simple variance here, as [16]. it is calculated as product degree criticality, measured percent, and local variance [16] Cost Analyses 3.3. Cost Analyses The followg sections will focus on cost side. The basic procedure normally cludes a ma analysisthe performed followg assections a workshop will focus fromon half acost dayside. to one The or basic a few procedure days. The normally result under cludes a condition ma keepg analysis to performed rules as a workshop game is from half partly a day a to neutral one or a and few correct days. The statistical result under value condition fal keepg to rules game is partly a neutral and correct statistical value fal result cost, duration or pritability, and partly ranked list primary optimization possibilities, result cost, duration or pritability, and partly a ranked list primary optimization possibilities, actually fal Uncertaty Prile or top ten list. actually fal Uncertaty Prile or top ten list. Durg ma workshop project, costs are gradually evaluated and uncertaties Durg ma workshop project, costs are gradually evaluated and uncertaties reduced reduced durg durg some some successive successive steps. steps. In In followg followg period period time, time, identified identified optimization optimization options options are are examed, examed, and and realized realized or or decided decided if if relevant. relevant. The The former analysis result is isn nupdated as a result as a result optimization optimization work. work. The end The result end result beg beg an optimal an optimal and neutral and neutral mean mean value value and and related uncertaty, related uncertaty, ready forready decision for decision makers makers to defe to defe a realistic a realistic workg workg budget and and aa reserve. This is illustrated is illustrated Figure Figure Figure 5. Illustration a typical analysis process. Durg ma analysis, project costs are Figure 5. Illustration a typical analysis process. Durg ma analysis, project costs are gradually evaluated and uncertaties reduced durg some successive steps. The dividual steps gradually evaluated and uncertaties reduced durg some successive steps. The dividual steps may go both up and down. The number successive steps is typically only four to seven before all may go both up and down. The number successive steps is typically only four to seven before all decisive uncertaties are handled. Durg followg period time, identified optimization decisive uncertaties are handled. Durg followg period time, identified optimization options are examed, and realized or decided if relevant. The former analysis result is n updated as a result optimization work. The end result is a neutral and optimal mean value and related mimized uncertaty, ready for decision makers to defe a workg budget and a reserve.

7 Adm. Sci. 2016, 6, The Method volves an Analysis Group a creative, multi-disciplary process which qualitative and quantitative data about future are captured and modelled. All issues are cluded, st factors as well as hard technical ones. The overall idea is to promote a dynamic and stimulatg process, which concentrates thkg on uncertaties (unknowns and unknown unknowns ) rar than certaties (knowns). The process is led by an experienced facilitator. The approach volves distct Qualitative, Quantitative and Action-Planng stages. As noted above, quantitative modellg is top-down nature, detailg and evaluatg most important issues on top-ten list uncertaties successive steps. This allows group to keep an overview, to focus on really important aspects, and to avoid wastg time and resources on many issues which are little or no importance. A basic practical procedure is described Supplementary Materials below, besides beg a congress paper [17], while a full description and discussion is found author s handbook on method [9]. 4. Results and Experiences The decision makers receive normally a defite total dollar value figure from calculators. They n have to more or less guess wher this figure is an optimistically one or it maybe is too high. Obviously, it may go wrong at times. Actually, it happens most frequently as outled Section 1. The decision makers via this research need not to guess anymore. They are now able to decide relevant risk overrun. Acceptg build- uncertaty a budget estimate, decision makers are now able to balance agast a low budget, with related risk overrun and a higher budget, with a related higher possibility to avoid overruns. In latter case, we risk that project discretely uses idle resources nice to have situations, which may be secondary benefit compared to alternative use. In Norway, for years, y have solved this balance problem, while operatg with two budgets, (1) a workg budget, which is typically decided close to calculated mean value, and thus havg ca. 50% chance to keep to budget; and (2) an ultimate budget, clusive a reserve up to 80% likeless for keepg to budget. Major Norwegian projects fanced by state have to be started by government, and accepted by parliament for fal go and fancg. Smaller projects are decided by relevant agency. The difference between workg budget and total budget is kept as a reserve at a higher level decision party. The project may, if necessary, ask for additional funds from this reserve pool accordg to defed rules, but it is not too easy and it requires an unlucky development, such as bankruptcies or like. In case underrun, idle resources go back to fundg agency. Sce year 2000, fal costs all large Norwegian public projects clusive IT projects have been very close to analysed mean values on average, and thus to workg budgets, as furr developed Section 4.2 below. Some organizations use procedure ad hoc, while usg external facilitators from consultants. This was case example Section 4.1 below. Or organizations implement procedure ir organization, managed by a high level admistrative body specialists, who act as facilitators, and mata procedure at a high pressional level. The Norwegian construction company, AF group, and some Scandavian public organizations with rail and road are examples this approach. Durg three decades, up to a thousand projects have been successfully analysed usg this methodology. However, it has been hard for many people to believe that it is actually possible to get full control over budgets (and schedules) large projects. Every successful dividual project may have or reasons behd ir success. Two cases will exemplify typical applications and potential benefit. Oslo Spectrum is a large multiuse arena. The wng project from architect competition estimated cost at $43 million. Three years later, many terested parties had worked with project organization to be

8 Adm. Sci. 2016, 6, Adm. Sci. 2016, 6, Two cases will exemplify typical applications and potential benefit. Oslo Spectrum is a cluded. large multiuse Before arena. go, The a quality wng analysis project after from se architect prciples competition was decided. estimated It showed cost $114 at million $43 million. Three years later, many terested parties had worked with project organization to be as mean-value. The project was n reorganized, part supported by analysis results, down cluded. Before go, a quality analysis after se prciples was decided. It showed $114 to a mean $76 million. This was accepted and became workg budget, while ultimate million as mean value. The project was n reorganized, part supported by analysis budget was $86 million. The ficial fal account was $76 million. results, down to a mean $76 million. This was accepted and became workg budget, while Anor example is Lillehammer Olympic Games The itial vestment budget rose ultimate budget was $86 million. The ficial fal account was $76 million. from $230 Anor million example to $385 is million Lillehammer over Olympic summer, Games more than four The years itial before vestment games. budget rose A risk analysis from $230 showed million an expected to $385 million fal total over cost summer, $1230more million. than This four potential years before overrun games. was A course risk politically analysis unacceptable. showed an expected The vestment fal total cost plans $1230 weremillion. n reorganized This potential overrun part supported was course by analysis politically result and unacceptable. followed The by vestment several updatg plans were analyses. n reorganized The expectedpart vestment supported figure by was eventually analysis some result and months followed later reduced by several to $800 updatg million analyses. as a mean The expected value. This vestment becamefigure workg was budget, eventually while some ficial months committee later reduced was allocated to $800 million a reserve as a mean approx. value. $90This million. became However, workg fal ficial budget, accounts while equaled ficial committee analysis mean was allocated value $800 a reserve million, approx. so reserve $90 million. fundhowever, was saved and wasfal later ficial used to accounts operate equaled facilities analysis after mean games. value $800 million, so reserve fund was saved and Many was later used successful to operate results facilities have after generally games. been explaed as pure luck, perfect project management Many or like. successful However, results thishave capability generally to successfully been explaed predict as pure project luck, perfect outcome project has for management first time been or more like. properly However, documented this capability fromto two successfully dependent predict series project largeoutcome projects. has They for are first time been more properly documented from two dependent series large projects. They summarized below. are summarized below A Set 40 Large Danish Road Projects ( ) 4.1. A Set 40 Large Danish Road Projects ( ) Over period , vestment costs all current large-scale Danish highway projects Over period , vestment costs all current large scale Danish highway were estimated by National Danish Road Authorities usg Successive Prciple. The projects projects were estimated by National Danish Road Authorities usg Successive Prciple. The typically cluded new motor roads, as well as enlargements existg roads, typically from projects typically cluded new motor roads, as well as enlargements existg roads, typically from to 3 3 lanes. They were various phases, a few ir early phases, majority prior to to lanes. They were various phases, a few ir early phases, majority prior to sanction/execute sanction/execute decision, decision, and and a a few few followg followg phases. phases. By By 2013, 2013, a a significant significant number number se se projects projects were were fished, fished, and and it it was was possible possible to to compare 40 prognoses with with related relatedactual actual fal fal costs costs for for se se projects. projects. The The total total vestment vestment was was ca. $2300 ca. $2300 million. million. The The result result is illustrated is illustrated Figure Figure 6 below. 6 Thebelow. vertical The scale vertical dicates scale dicates wher wher actual actual project project cost iscost lower is lower or higher or higher than than expected cost. This cost. is measured This is measured percent percent actual actual cost. cost. Figure 6. Result from series test prognoses fal costs Danish motor road projects. The Figure 6. Result from a series test prognoses fal costs Danish motor road projects. figure illustrates actual project costs proportion to expected project cost ( mean value) The figure illustrates actual project costs proportion to expected project cost ( mean value) from analyses. The downward bars are savgs compared with predicted mean value, while from analyses. The downward bars are savgs compared with predicted mean value, while upward bars are overruns compared with predicted mean value. For example, less than 10% upward bars are overruns compared with predicted mean value. For example, less than 10% means that fal cost was 10% lower than analyzed mean value. The likely reason for majority savgs is explaed text (from Lichtenberg [18]).

9 Adm. Sci. 2016, 6, It can be seen that far more projects were under expected costs compared with those that exceeded it. The average savgs compared with expected costs were approximately 5%. A likely reason to this is fact that project managers were very keen to make use top ten list primary key risks and uncertaties that is generated from quantitative cost analysis. Concentratg ir risk mitigation effort and exploitation potential improvements onto se primary areas reduces cost efficiently, and this may expla why many projects eventually ended up under-runng mean cost that was predicted from analysis. An example may illustrate this. If, for example, cost effect local communities appears as a large uncertaty at top ten list, project manager is urged to itiate talks and best possible cooperation with local decision makers, such as leadg local politicians. This will ten result lesser problems and related additional costs than estimated budget. The estimated uncertaty whole set projects, measured as a standard deviation, corresponded well with spread results. For more details, refer to [18] All Major Norwegian Governmental Projects Approved and Completed over 13 Years This case example is referred from an ficial report, no. 36, from Norwegian research program, concept, Samset et al. [19]. Durg 1990s, Norwegian large public projects too ten lead to serious overruns. Durg same period, Successive Prciple began to manifest itself as able to make realistic budgets, largely among contractors, fshore dustry and or areas private sector. Gradually, some public agencies also implemented procedure and improved ir budgets. Inspired by se results, government itiated a commission. In 1999, it suggested use Successive Prciple towards all major projects fanced by State, Berg [20]. The so-called KS2 or QS2 procedure was made obligatory from year 2000 towards all se projects above $70 million. It became ambition that about 80% projects should stay with approved ficial budget. Any higher ambition was assumed to be too costly. As a result, it became normal se public projects to fer executg body a workg budget close to estimated mean value, which cluded an expected cost for contgency. To compensate for unavoidable uncertaty, a fancial reserve was cluded approved ficial budget. This reserve course was under control a higher level body. Sce n, quality audits based on Successive Prciple have been obligatory for all large-scale Norwegian public-sector vestment projects above $70 million before fal decision to fance and execute. The practical admistration was given to some authorized consultants, which were dependent project owner. The analysis scope was widened to also assess organizational set up and some or aspects. In 2002, a permanent research program, Concept, was launched, with purpose to follow and furr develop this system [21]. By 2013, 40 se projects all categories, cludg IT projects, had been completed. An ficial report [19] by Concept group published results so far. The report shows that actual fal project cost on average was very close to workg budget and thus to expected mean values. This is illustrated Figure 7. The data figure are taken from above mentioned reference [19]. It can be seen that it is possible to predict fal project costs fairly accurately statistical terms. This is true for estimated mean values. However, spread actual costs are somewhat larger than estimated this set. This underles fact that difficulties still exist evaluatg subjective uncertaty an unbiased way. What about ambition that about 80% projects should keep with approved ficial budget? Actually, projects, or 80%, kept to budget! This is visualised Figure 8 with figures from [19]. The results mean that bulk reserve money after 13 years was largely tact. For furr formation, see [19].

10 Adm. Sci. 2016, 6, Adm. Sci. 2016, 6, Adm. Sci. 2016, 6, The result budget quality audits all large Norwegian falized governmental Figure The The result result budget budget quality quality audits audits all large all Norwegian large Norwegian falized falized governmental governmental projects qualified projects qualified while usg Successive Prciple from 2000 to The actual outcomes projects qualified while usg while usg Successive Successive Prciple Prciple from 2000from to The to actual The outcomes actual are outcomes compared with compared agreed with workg agreed budgets workg before budgets execution before execution started (asstarted percent (as percent actual cost). actual These cost). budgets These compared with agreed workg budgets before execution started (as actual cost). These were budgets typically were typically close to close expected to expected cost. Thecost. figures The are figures takenare from taken [19]. from [19]. budgets were typically close to expected cost. The figures are taken from [19]. Figure 8. The same large Norwegian projects as shown Figure illustrate here actual fal total Figure 8. The same large Norwegian projects as shown Figure 7 illustrate here actual fal total project cost compared with approved ficial budget. It is seen that 80% se projects kept to project cost compared with approved ficial budget. It is seen that 80% se projects kept to budget. This is well agreement with ambition for se Norwegian projects, but it is budget. This is is well well agreement with with ambition ambition for for se se Norwegian Norwegian projects, projects, but it but is very it is very different to ternational experiences, where cost overruns are commonplace. The figures are different very different to ternational to ternational experiences, experiences, where cost where overruns cost overruns are commonplace. are commonplace. The figures The arefigures from [19]. are from [19]. from [19]. 5. Practical The results Procedures mean that and Applications bulk reserve money after 13 years was largely tact. For furr The results mean that bulk reserve money after 13 years was largely tact. For furr formation, see [19]. formation, A number see [19]. specific variants this procedure have been developed for a range practical applications: cost estimates, schedule analyses, cost/benefit analyses, as well as for project optimization, 5. Practical Procedures and Applications 5. Practical Procedures obtag and consensus, Applications and team buildg. Examples are given [9,11,14 17,22,23], To number conclude, specific Successive variants Prciple this has procedure been used have as an been tegrated developed approach for range for: practical A number specific variants this procedure have been developed for a range practical applications: cost estimates, schedule analyses, cost/benefit analyses, as well as for project applications: * Quality assurance cost estimates, budgets, schedule schedules, analyses, pritability cost/benefit analyses analyses, and fancial as well analyses. as for project optimization, obtag consensus, and team buildg. Examples are given [9,11,14 17,22,23], optimization, * Identification obtag and rankg consensus, fancial and team risks buildg. and opportunities. Examples are given [9,11,14 17,22,23], To conclude, Successive Prciple has been used as an tegrated approach for: * To Action conclude, planng for Successive improvements Prciple has productivity, been used safety as an and/or tegrated competitiveness approach for: Quality assurance budgets, schedules, pritability analyses and fancial analyses. * Quality Rankgassurance alternative budgets, solutions. schedules, pritability analyses and fancial analyses. Identification and rankg fancial risks and opportunities. * Identification Team buildg andrankg consensus fancial support. risks and opportunities. Action planng for improvements productivity, safety and/or competitiveness * Action planng for improvements productivity, safety and/or competitiveness Rankg alternative solutions. * Rankg alternative solutions. Team buildg and consensus support. * Team buildg and consensus support.

11 Adm. Sci. 2016, 6, Applied towards schedulg tasks, we witness a sharp contrast to many details related conventional estimates and schedules. Surprisgly few issues domate total result. This is accepted by users as a most exitg matter. The Norwegian Gardemoen airport, as an example, was delayed less than a year before deadle. An analysis accordg to Successive Prciple confirmed this. The fal Uncertaty Prile, or top ten list, identified that only three essential aspects or problems responded for 90% total uncertaty, and thus for possibilities for acceleratg. They were settled a few days later at a meetg with relevant top managers and specialists. This accelerated schedule sufficiently and new airport opened at planned day. The procedure has even been successful towards severe uncertaty. A Norwegian gas pipe le project from an fshore construction had a very tough and important deadle. The task cluded buildg a tunnel from shore and out to open sea. This had never been tried before, and uncertaties were course large. A schedule analysis dicated that mean duration equaled time to deadle. Even a mean value as told cludes a fair part potential delayg problems as well as possibilities to expedite task. It proved to be enough, as task successfully met deadle. 6. Critical Discussion on Theory and Practice 6.1. Some Positive Experiences Several managers, project managers, estimators and planners durg decennias have changed ir conventional procedures and implemented se prciples and related procedures. They are typically expressg surprise at how close prognoses were to actual results. This has been most obvious analysis project schedules, because falization large physical projects is ten an event visible media. Even large Swedish IT projects have kept to estimated schedule, when analysed accordg to Successive Prciple. An example illustrates that prciple has even been strumental brgg delayed IT projects back on track. One such strategically important project was delayed half-way through by approx. six months. At this juncture, an analysis process was conducted usg Successive Prciple, which confirmed that, as it stood, project had to be expected to be late by six months. However, as part analysis result, followg three acceleration itiatives were identified: (1) Earlier and closer contact with future users; (2) Allocation a key dividual to be at disposal project; (3) Acceptance by senior management higher priority for this project. A supplementary alternative analysis showed that se steps would accelerate project sufficiently to falize it on time. Senior management authorised se terventions, and y were implemented and project did actually meet scheduled time. The ma reason for successful results most likely is successful attempts to elimate many psychological pitfalls. They tend to severely damage planng and estimatg, which more and more depends on subjective evaluations, regardless model used. Anor reason may be that analysis group is carefully composed to be both competent and broad. This allows everybody to use all ir creativity, telligence and even tuition under control as a supplement to ir concrete knowledge. This also assists compensatg natural drive for engeers and economists to focus too much upon technical and fancial matters, and too little upon more smooth matters. Important benefits have materialized durg recent decades practical applications. The primary strength lies partly fact that unpleasant surprises can be significantly reduced. In fact, largely elimated, and partly new opportunities for project efficiency, competitiveness and creased teamwork that this method helps uncover. Appropriate questions are now asked before a project starts stead hdsight when it may be too late.

12 Adm. Sci. 2016, 6, A large portfolio public and private companies Scandavia and beyond consistently testify to benefits usg Successive Prciple. Many users liken Successive Prciple to a pair future-glasses or to an ultra-sound scan plans, estimates, budgets, etc. This is partly because specific and realistic picture how matters will progress terms budget or schedule, and partly because ranked Uncertaty Prile or top ten list most important external and ternal sources uncertaty. It consists specific threat factors, hidden opportunities or simply key time or cost uncertaties. The value this list lies allowg client to take timely preventive or protective actions agast threats and to make most positive opportunities order to improve competitiveness or to achieve greater efficiency. Control and optimization major project schedules have been a particularly successful sphere application Limitations, Challenges and Need for Furr Development The use Bayesian statistical ory has been criticized from those statistical orists, who did not accept this widely accepted ory. The old and basic statistical ory, which terprets probability as frequency or propensity some phenomenon, is simply unable to handle uncertaty future events. Thus, use Bayesian statistical ory, simply, is necessary to use, and luckily it allows for handlg subjective uncertaty or personal belief. The numerous positive results strongly dicate that this use is relevant. A more essential criticism has focused upon possible disturbg effect from composition analysis group. This, however, does not seem to be a problem. An example dicates that. After an analysis, project owner saw this as a problem and decided to repeat analysis dependently first with anor group participants. They, however, came out with same result. However, it is still essential to follow rules and criteria for composition analysis group, and course this constitutes an uncertaty. A similar methodical uncertaty stems from fact that an analysis group may generally be somewhat too optimistic or too pessimistic. However, it is part procedure to let participants make a fal triple estimate that seeks to compensate such a risk. The fal result thus cludes this methodical uncertaty. The Successive Prciple deals with uncertaties, not proper risks, like catastrophes. However, possibilities to use prciple actual Risk Analyses have been tested with some success. However, this area still needs furr development. The limitations approach largely lie havg to comply properly with rules game some m untraditional and, for some participants, provokg. It can, for example, only be useful organizations with a modern management attitude, and a policy ternal openness. It requires courage and genue will to meet naked truth task, wher it concerns schedules, budget proposals, fancial analyses, etc. This is described [15,24,25]. Because novative character Successive Prciple and weight given to both qualitative and quantitative evaluation processes, a considerable and demandg preparation period is also needed prior to workshop, and not least toger with full, unequivocal backg senior management. Even if this is obtaable, process must be guided by a facilitator who has a good grasp group dynamics, statistics and use Successive Prciple. Such people may be hard to fd. An observed weakness is to focus solely on physical items without sufficient regard to overall general effects from project environment, management, client, current project situation, future developments, etc. (refer to Figure 2). Anor typical weakness lies not handlg triple estimates correctly. Unfortunately, well-documented problem underestimatg range between maximum and mimum values still occur as a problem Norway. This can be compensated for by usg most extreme values whole analysis group, and via use secret triple estimates by each dividual. In [23], Norwegian researchers observe followg furr challenges with today s practice that still need more attention:

13 Adm. Sci. 2016, 6, * expressg real uncertaty very early phases a project; * avoid diggg too deep details; * gettg standard deviation right (expressg realistic uncertaty) all phases project; * handlg human error and team effects (heuristics and group thk) and understandg ir effects; * avoid neglectg opportunities durg risk and uncertaty analysis (we seem to have a bld spot for opportunities). The Successive Prciple addresses to some degree all se challenges, but even more improvements practice are called for. We applaud such itiatives and suggest one specific improvement concludg part. The potential is huge, but it does not come for free. Significant vestment education key dividuals and trag facilitators is needed. So too is development and implementation organizational procedures to support and utilize method. 7. Conclusions Cost Engeerg and Project Management have made serious advances, but projects are neverless ten delivered too late and with severe cost overruns. This paper describes how former research results have successfully challenged traditional way thkg about project planng, cost estimation and risk analysis. The results have been surprisgly promisg sce troduction 90s, but y have ten been explaed away by referrg to luck or or situational factors. Then, fally, 2013 and 2014, has come credible research that confirmed re is more than cocidence behd se results. The results are presented above: by usg Successive Prciple, good control over cost large projects has fally been established Scandavia, first all Norway and to some degree Sweden and Denmark besides sporadically or countries. In two latter countries, prciples have been used more ad hoc dustry and some public agencies, except that Swedish rail agency has implemented prciple. Supplementary Materials: The Supplementary Materials are available onle at 3/8/s1. Acknowledgments: This research was supported by Technical University Denmark, DTU, and Technical University Norway, NTNU. The author thanks colleagues from se two stitutions, who have contributed over years with sight and expertise so that research, fact, is a common team result, although y may not agree with all terpretations/conclusions this paper. The author is grateful to numerous colleagues and partners various consultancies for spiration and assistance connection with practical experiences usg resultg new prciples. First all, dedicated consultants International Futuraone group, today represented by Erlg Hjallen, senior consultant, Kongsberg, Norway. The author is also deeply grateful to Peter Adlgton, senior consultant, Adlgton Associates, Cambridge, partner Futuraone group, for his wise comments and valuable lguistic assistance to this paper. Conflicts Interest: The author declares no conflict terest. References 1. Anonymous. CHAOS Report; Standish Group: West Yarmouth, MA, USA, Flyvbjerg, B.; Holm, M.S.; Buhl, S. Underestimatg Costs Public Works Projects: Error or Lie? J. Am. Plan. Assoc. 2002, 68, [CrossRef] 3. Flyvbjerg, B.; Holm, M.S.; Buhl, S. How common and how large are cost overruns transport frastructure projects. Transp. Rev. 2003, 23, [CrossRef] 4. Merrow, E.W. Industrial Megaprojects: Concepts, Strategies and Practices for Success; John Wiley and Sons: Hoboken, NJ, USA, Odeck, J. Cost overruns road construction What are ir sizes and determants? Transp. Policy 2004, 11, [CrossRef] 6. Klakegg, O.J. (Ed.) Norrn Europe Group Consultants Specialized Application Successive Prciple; Futura One: Kongsberg, Norway, 2002.

14 Adm. Sci. 2016, 6, Kahneman, D. Thkg, Fast and Slow; Pengu Books: London, UK, Lange, N. Sikker Vurderg af Usikkerhed (Accurate Evaluation Uncertaty). Master s Thesis, Technical University Denmark, Lyngby, Denmark, (In Danish, unpublished work) 9. Lichtenberg, S. Successiv Kalkulation (Successive Estimatg); Research Report; Technical University Denmark: Lyngby, Denmark, (In Danish, unpublished work) 10. Lichtenberg, S. Proactive Management Uncertaty Usg Successive Prciple; Lichtenberg & Partners: Vedbæk, Denmark, 2000; p Lichtenberg, S. The Successive Prciple. In Proceedgs PMI International Symposium, Project Management Institute, Washgton, DC, USA, 1974; pp Teigen, K.H. The language uncertaty. Acta Psychol. 1988, 68, [CrossRef] 13. Brun, W.; Teigen, K.H. Verbal probabilities: Ambiguous, context-dependent, or both? Organ. Behav. Hum. Decis. Process. 1988, 41, [CrossRef] 14. Klakegg, O.J. Trnvis-Prosessen ( Handbook on Stepwise Procedure); NTH Technical University Norway: Trondheim, Norway, (In Norwegian) 15. Aass, T.; Jermstad, O.; Aanes Johansen, K.; Klakegg, O.J. Governance Norwegian Government Projects. In Proceedgs International Project Management Association, IPMA World Congress, Istanbul, Turkey, 1 November Lichtenberg, S.; Møller, L.B. Three Types Biases Schedulg and solutions Applicable Practice. In Proceedgs International Project Management Association, IPMA World Congress, Garmisch-Patenkirchen, Germany, September Archibald, R.D.; Lichtenberg, S. Experiences usg Next Generation Management Practices. In Proceedgs International Project Management Association, IPMA World Congress, Florence, Italy, June Lichtenberg, S. Great News: New evidence Accuracy Cost Prognoses. Lichtenberg & Partners, Available onle: (accessed on 1 June 2014). 19. Samset, K.; Volden, G.H. Investg for Impact. Lessons with Norwegian State Project Model and First Investment Projects that Have Been Subject to External Quality Assurance; Norwegian Technical University: Trondheim, Norway, Berg, P. (Ed.) Styrg av Statlige Investerger (Controllg Governmental Investments); Norwegian Mistry Fance: Oslo, Norway, (In Norwegian) 21. Anonymous. Research Program, on Optimizg Large Public Projects; Norwegian University Science and Technology: Trondheim, Norway, Anonymous. Anslagmetoden (Stepwise Method), Handbook R764f ; Statens Vegvesen (National Road Authority): Oslo, Norway, Johansen, A.; Sandv, B.; Torp, O.; Økland, A. Uncertaty analysis 5 Challenges with today s practice. Procedia Soc. Behav. Sci. 2014, 119, [CrossRef] 24. Klakegg, O.J.; Lichtenberg, S. Bedre Kontrol Med Usikkerheden i Projektplanlægng (Better Control Uncertaty Project Management); Projektledelse: Allerød, Denmark, (In Danish) 25. Klakegg, O.J.; Lichtenberg, S. Successive Cost Estimation Successful Budgetg Major Projects. In Proceedgs IPMA 29th World Congress, Panama City, Panama, September by author; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under terms and conditions Creative Commons Attribution (CC-BY) license (