By Peter J. Kennedy and Robert J. Avila

Decision making under extreme uncertainty: Blending quantitative modeling and scenario planning

Sometimes the best view of the future re- now, but have an impact tail reaching out be- quires an odd pair of bifocals: quantitative yond five years and as far out as 50 years. modeling and scenario planning. Quantita- Scenario planning is essentially qualitative. It tive modeling is designed to support rela- is not predictive. Scenarios don’t “come true.” tively near-term, tactical decision making, They are meant to provide realistic alterna- while scenario planning is a planning tool to tive frameworks for business analysis, strate- develop insights about the longer term. gy and decision making. Occasionally, scenar- These two analytical approaches to decision ios are uncannily prescient about an event or making are typically not used together. But a trend, though their real purpose is to accu- some planning challenges justify using both rately describe the range of operating envi- to get a better read on possible discontinu- ronments an organization must be prepared ity. For example, this case study of a sub- to cope with. Thus, if researched and con- sidiary of a multinational automotive firm structed rigorously, they will effectively an- operating in extremely uncertain market ticipate risks, threats, challenges and endur- conditions shows how the two approaches ing areas of uncertainty facing an organiza- were successfully blended – producing some tion. important lessons during the process. BACKGROUND TO THE CASE: MANAGING AMID If done rigorously, quantitative modeling is EXTREME TURBULENCE an unquestionably useful and necessary de- “BrasilAuto, Incorporado”2 is a subsidiary of a cision-support tool for short- and medi- large, global automotive enterprise, with many um-term market and operational planning. years of experience in Latin America. In years This is espe- immediately prior, BrasilAu- cially true in to was struggling under chal- the automotive SCENARIO PLANNING: A DEFINITION lenging conditions in Brazil, industry, and circumstances were where many Scenario planning is a strategic planning growing increasingly uncer- executives process that explicitly considers a range of tain, as political actors de- come from en- plausible alternative business environments bated the merits of economic gineering or fi- and the issues critical to the organization’s liberalization at a time of nance back- success as a way to manage longer-term chronically high inflation grounds. risks, identify hidden opportunities and pri- oritize alternative courses of action. Sce- and stagnant purchasing In contrast, nario planning should not be confused with power. On one end of the po- scenario plan- “spreadsheet scenarios” in which numeric litical spectrum were re- ning is normal- baseline forecasts are tweaked with alterna- formers who wanted the na- ly used for tive assumptions. Unlike strategic planning tion to open up its economy longer-term scenarios, spreadsheet scenarios by them- to the world as many other decision mak- selves do not yield strategies, though they emerging market nations ing – actions can deliver strategic insights.1 had done, and to remove sti- that need to be fling regulations that tended considered to favor large domestic pro-

Copyright © 2009 Futures Strategy Group, LLC. All rights reserved. ducers and labor elites. On the other end dustry. Though it was rare for clients to re- were nationalistic and populist interests urg- quest the two tools be applied in an integrat- ing more gradual change that ensured labor ed process, this is precisely what Brazil-Auto rights and consumer protections. Neat “left” wanted. The core analytical need articulated versus “right” divisions didn’t exist here, and by the company was for the project to answer many businesspeople, government officials two questions: and politicians found themselves agreeing . Where is Brazil headed? with elements of both perspectives. Given . How many cars can we sell? this dynamic situation, it was impossible to see where the nation was headed. Against this confusing, unpredictable politi- co-economic backdrop BrasilAuto’s management team, both in country and at headquarters, faced a range of complex decisions re- lated to capacity, vehicle mix, pricing, distri- bution, dealer relationships, exports, labor and government relations. There was no consensus around either long- or short-run prospects for the nation or the vehicle market. BrasilAuto expatriate executives tended to be pessimistic, possibly because they felt overwhelmed by an unfamiliar business cul- ture and arcane financial practices that had taken root in the inflation-ravaged nation. Their thinking could be summed up by the aphorism, “Brazil is the country of the future – and it will always be.” The company’s Brazilian executives tended to be somewhat more optimistic, insisting that the nation’s large market, human talent, and determina- tion in the face of adversity assured an eventual upside. But even the optimists were hard pressed to predict when circumstances would improve. Meanwhile, the pressure to resolve key business decisions was growing.

EMBRACING A SCENARIO PLANNING SOLUTION A team of consultants3 from the U.S. was called in to provide an independent evaluation. They had experience working with BrasilAuto and had several ongoing projects in the U.S. and Europe with the firm’s par- ent. The consultants proposed a short-term scenario-planning approach because of the extreme nature of the uncertainty facing the country and pressures facing management to quickly come to terms with the threats and opportunities. The consultants had experience with scenario planning as a qualitative tool to build sturdy strategies in complex and highly uncertain business environments and also with econometric modeling in the automotive in-

Copyright © 2009 Futures Strategy Group, LLC. All rights reserved. Most importantly, the company wanted results delivered in less than 12 weeks, a diffi- BRAZIL SCENARIOS AND THE CHALLENGES THEY cult goal given the need for interviews on HIGHLIGHTED three continents, extensive market analysis Four scenarios were chosen (see Exhibit) and model building and the integration of based on three decision criteria: outputs from processes. 1. Events and conditions that could plausibly Besides needing results delivered quickly, materialize within the company’s five-year there were several other key considerations: planning horizon. 1. The company sought a balanced, indepen- 2. Scenario conditions that could have the dent evaluation of the prospects of country biggest impact on vehicle sales and competi- and the market. To ensure a representative tive conditions in Brazil. set of perspectives on the serious, structural 3. A set of at least four5 scenarios that would problems that clouded the short- and medi- together capture the range of threats, chal- um-term outlooks, the consultants conduct- lenges and surprise that the company could ed nearly 50 double-blind4 interviews in plausibly face. country, as well as in the U.S. and Europe, with government officials, business leaders, EXHIBIT: BRAZIL SCENARIOS think tank analysts, Brazilian labor leaders and opposition politicians. Each of the four scenarios contained both 2. The company wanted to hear both good qualitative characteristics and quantitative and bad news about what might be ahead indicators. The latter would serve as inputs for the nation, politically, economically and to a vehicle forecasting model. The qualita- socially. This gave the consultants license to tive elements were expressed as an integrat- create a realistic set of scenarios that re- ed narrative, looking back in time from five flected the company’s concerns, plus issues years out, and relating a logical story of how and strategic elements and that arose in the critical national events turned out. A sample course of the interviews. of the strategic issues presented by the sce- 3. The scenario analysis needed to be tied to narios follows. automotive demand, so the company would Downward Spiral combined bad, populist have reliable insight into capacity needs and politics and extremely inept economic man- financial implications. agement with widespread loss of business CUSTOMIZING A NON-TRADITIONAL SCENARIO and public confidence in the direction of the SOLUTION country. These conditions pointed to pro- longed depression in consumer demand for The project scope and time constraints dic- durables, like cars. tated a comparatively simple scenario struc- ture. That is, the scenario “uncertainty Stumbling Along looked at piecemeal mar- space” was built around a rudimentary 2x2 ket reforms, including privatizations, but of- matrix, with the axes being “Political Lead- fered no consensus on future direction. This ership” and “Level of Economic Reform.” scenario represented an impossible environ- Note that this matrix did not include an ex- ment for automotive production planning. ternal dimension, such as global economic Orthodox Shock imagined that Brazil swal- growth or level of global trade. This would lows the harsh medicine prescribed by its have been a grave omission if the company creditors after a foreign exchange crisis. As a was making a long-term strategic assess- result, demand plummets and unemployment ment. surges. But with stabilization comes a robust recovery and the formation of a labor-left coalition that provides political cover to re- formist elements.

Copyright © 2009 Futures Strategy Group, LLC. All rights reserved. Potential Realized represented success on fundamentals of motor vehicle demand were both economic and political front, with not in any way unique in Brazil. Instead, the Brazil embracing gradualist reforms and ex- seemingly chaotic swings in motor vehicle port-led growth. This scenario offered the sales were largely due to the fact that Brazil best possibilities for sustained automotive was a high growth market with a high degree market growth, but also the eventual chal- of economic instability. As a result both re- lenges of competing in an open, global trad- placement and growth demand could under- ing system. go significant and complex cyclical fluctua- tion. Thus, our scenarios, driven by the un- BUILDING A SCENARIO-CONTINGENT FORECASTING certainty around economic policy and the na- MODEL ture of the political leadership that would im- The first step in developing any econometric plement and carry them out, could be used to market analysis is to establish an under- produce alternative scenario contingent eco- standing of the fundamental drivers of the nomic forecasts, which in turn could drive market in question. For the executives of a credible motor vehicle forecasts. global car company with extensive experi- Annual econometric models for each of the ence in “normal” markets, Brazil with its hy- major categories of motor vehicle sales based perinflation, wildly swinging economy and on about 30 years of data were built; factors political instability seemed to be a place included urban adult population, non-agricul- where none of the rules held true. For the tural GDP, investment, the total stock of vehi- local management team it was a given that cles, and the inflation rate. Critically, one key Brazil was a special case. An initial objective factor for which we could get no historical of our econometric modeling would have to data was vehicle price. Essentially, we had no be a test of these assumptions. time series on vehicle prices to work with at Motor vehicles are durable goods that can all. Thus, there was no way to estimate the provide many years of service. Thus, new price elasticity of demand. We were told, vehicle sales are driven by both customers however, that in Brazil’s high inflation envi- seeking to replace existing vehicles and new ronment, a major concern of management buyers. In countries with growing popula- was to keep prices in line with costs to main- tions and rising income levels, growth tends tain profitability and that assuming relatively to be a significant driver of new car sales. In constant real prices would be a safe bet. Safe lower growth, mature economies replace- or not, there was no data available to us, so ment demand tends to dominate. Over the that assumption was made. business cycle the decision to replace a vehi- Each of the scenarios laid out detailed politi- cle may be delayed or accelerated, a fact that cal and policy timelines that noted the most can produce large swings in car sales. likely economic impact based on an analysis Our first step was to build auto replacement of the Brazilian economy and the track models using basic economic and demo- record of such policies in Brazil and other graphic data from over a dozen countries. countries. The idea was not to come up with These models clearly demonstrated that the a precise economic forecast – almost always

Copyright © 2009 Futures Strategy Group, LLC. All rights reserved. a fool’s errand – but rather one that was each event, assessing the probability of the credible and representative of what could event taking place in each year of the fore- reasonably be expected in the environment cast as well as the nature of its impact on mo- described by each of the scenarios. These tor vehicle sales. Their impact could be posi- scenario- contingent baseline forecasts were tive or negative, sudden or delayed, perma- then used to drive motor vehicle sales fore- nent or transitory, major or minor. The esti- casts within each of the scenarios. Each of mates for all of these impacting aspects of the four strategic scenarios addressed both these events were developed through re- the broad nature of the business and eco- search, third-party-expert interviews, discus- nomic climate, and specific insights as to sions with client teams and our own assess- how the auto market should be expected to ment of these findings and opinions. trend within that scenario. We used a Monte Carlo type analysis package, The problem we faced was that the Brazilian which simulated the random occurrence of economy at that time did not really trend. the complete set of events on each of the sce- Instead it lurched, stumbled, ran and hov- nario-specific motor vehicle forecasts. This ered. Even if the general direction could be enabled us to produce estimates of the ex- discerned, its actual path could be erratic. pected range of variation in motor vehicle We needed a way of indicating within each demand. scenario the expected range of variation in motor vehicle sales given the high year- WHAT WE LEARNED to-year variability in the economy. After considering all the major uncertainties facing the Brazilian motor vehicle market – In addition to the uncertain economy, our political, economic, market – we were able to research and interviews had uncovered convincingly demonstrate the range of more than a dozen possible one-off scenari- volatility in demand that the client needed to o-independent events (“wildcards”), ranging be prepared for. from a major increase in the global price of oil, another military takeover in Brazil and a From the start of the project, the company’s slashing of the motor vehicle sales taxes. Brazilian executives understood that it was Any one of these events could potentially not operating in an orderly market in which a disrupt our baseline forecast of the motor straight-line growth forecast was all that was vehicle market in any one of the scenarios. needed in order to plan their operation. They had already accepted the idea that their suc- ESTIMATING WILDCARD IMPACTS cess would depend on their ability to re- We used a form of Monte Carlo analysis to spond to a rapidly fluctuating range of mar- address the GDP driven variability in sales. ket conditions. What was needed, in order to Over the previous 20 years the Brazilian plan for that type of dynamic operation, was economy had averaged a respectable 4.8 a better understanding of the range and na- percent per annum in real GDP growth. Un- ture of the volatility that they would have to fortunately the year-to-year variance in that manage. As long as the risks and uncertain- growth had been +/-5 percent. Five hundred ties were no more than a continuously re- iterations were run with this normal vari- volving list of hypothetical “what ifs,” deci- ance about the GDP baseline forecast for sion making would remain paralyzed. But each of the scenarios, which in turn was once the major uncertainties were all clearly used to drive the scenario specific motor ve- identified, defined, researched, quantified hicle forecasts. The mean, standard devia- and integrated – and corporate headquarters tion, maximum and minimum for each year’s bought into the analysis – only then could forecast was calculated to provide the ex- management focus on developing a strategy pected range of variation for the motor vehi- and building the required organization to cle forecasts in each of the scenarios due to deal with the anticipated challenge. the expected variability in Brazilian GDP. They could clearly see that there was a base- Quantifying the impact of the various wild- line of operations and that there would be card events required a detailed analysis of peak demands for which they had to have the

Copyright © 2009 Futures Strategy Group, LLC. All rights reserved. ability to quickly ramp up to. They under- have helped the company learn the scenario stood that the range and nature of this process from the bottom up, and better volatility varied across the major segments equipped its executives to both autonomous- of the motor vehicle market and their expe- ly run the model and continue to work the rience during past swings enabled them to scenarios after the consultant team complet- prepare for future market gyrations. Finally, ed the engagement. This kind of knowledge they had a better idea of what to watch for transfer to the client is an extremely impor- in the political sphere and how to evaluate tant, though under-appreciated, scenario- the impact of changing policy options. Hav- planning deliverable. ing looked at the business environment hard 4. Resist relying on conventional wisdom and carefully, the uncertainty was signifi- from the client, especially when it concerns cantly less unsettling or paralyzing than it market fundamentals such as prices. The had been. econometric models were built with no rela- PROCESS IMPROVEMENTS tive price information and the wildcard impact of a government cut in the sizable new From the consultants’ perspective, the expe- vehicle tax was based solely on expert opin- rience taught some important lessons, which ion. We should have introduced ad hoc esti- we’ve since incorporated into our scenario mates of price elasticity from other compara- consulting practice: ble markets. As it turned out the government 1. Attempting to integrate the qualitative did remove the tax, and this policy move re- and quantitative dimensions of the scenar- sulted in a vehicle sales surge that exceeded ios was overly ambitious. We tried to cram our most aggressive both quantitative indicators and strategic insights into five-year scenario stories. In the end, executives’ narrow focus on vehicle demand estimates diverted their attention from the qualitative insights. Better practice would incorporate into the scenario planning process at least one executive-strategy workshop to qualitatively explore the scenarios before any vehicle forecasting num- bers were displayed. Such workshops ex- pose executives to a broad range of strategic insights and implications that otherwise tend to get buried when quantitative values are presented. 2. The scenarios should have been longer term – an end-state ten years into the future, instead of five. Emerging markets re- quire that kind of longer-term perspective. In this case, Brazil passed through three of the four scenario worlds in the ten years after this project was completed. Companies today considering scenario work focused on China and India, for example, should similar- ly embrace at least 10-year scenario hori- zons, if not longer. It takes about a decade for real structural changes in an economy or society to play out. 3. Closer consultant/company working relations would have benefited the project. The formation of a client “Core Team” would

Copyright © 2009 Futures Strategy Group, LLC. All rights reserved. upside forecast. A few years later at the client’s headquarters senior executives on this project confided in us: “You guys under- estimated the upside when the government withdrew the taxes on vehicle sales. But so did we, and you were a lot less wrong than we were!”

Endnotes 1 For a longer discussion on different types of scenarios, see Charles Thomas’s Types of Scenario Planning at http://www.fu- turesstrategygroup.com/outlook- may08.htm 2 BrasilAuto, Incorporado is a fictional name for an actual auto company’s Brazilian subsidiary. 3 The Futures Group was the predecessor firm to the Futures Strategy Group. 4 Interviewees did not know the identity of BrasilAuto. In turn, BrasilAuto received a summary version of the interviews; no com- ments were attributed to any specific indi- vidual. 5 Four is generally considered the minimal number of scenarios for most organizations. A set of three sets up a “Good,” “Bad” and “Most Likely” mental model that is detri- mental to rigorous scenario thinking about the possibilities of futures that are truly different from expectations. * * * Originally published in Strategy & Leader - ship , Volume 41, Issue 4 (2013). Peter J. Kennedy is a founding principal of the Futures Strategy Group, a strategy consulting firm based in Glastonbury, Connecticut. [email protected] Robert J. Avila is managing director of Fu- tureCrunch, a New York City-based economic consulting firm, and an associate of the Fu- tures Strategy Group. [email protected]