Scenario Planning & What-Ifs

The foundational move of scenario planning is to stop asking the twin to predict the future and start asking it to compare clearly defined options. A scenario is a coherent, fully specified possible future - a described combination of choices and conditions ('build the metro, housing grows at the planned rate, fuel prices hold') - that the twin can simulate end to end. Scenario planning runs several of these side by side, always including a baseline (usually 'do nothing' or 'business as usual'), and reads the differences between them as the decision-relevant information.

This reframing is not a technicality; it is what makes simulation trustworthy enough to act on. Recall the lesson from across this module: a simulation's absolute numbers carry heavy, stacked uncertainty, but its relative comparisons are far more robust, because many of the errors and assumptions that distort an absolute figure apply to all the scenarios alike and partly cancel when you compare them. If every scenario over-estimates future traffic by the same optimistic growth assumption, the ranking between them - which option relieves congestion most - can still be sound even though no single number is. So the twin is on much firmer ground telling a city 'option A floods less than option B and costs less than option C' than telling it 'option A floods to exactly 0.6 metres.' Scenario comparison extracts the reliable signal and discards the unreliable precision.

Good scenarios are an act of design in themselves. They must be genuinely distinct (testing real alternatives, not trivial variations), plausible, and fairly specified - it is easy, and dishonest, to rig a comparison by giving your preferred option generous assumptions and its rivals mean ones. They should cover not just the options the city wants but the futures it fears: the high-growth case, the low-funding case, the extreme-climate case, so the decision is tested against adversity, not just hope. And crucially, the scenarios encode values and priorities - which options are even considered, what counts as a good outcome (traffic flow? equity? emissions? cost?) - so who frames the scenarios shapes what the twin can conclude. A twin that only ever tests car-capacity options against each other can never reveal that the best answer was a tram, because the tram was never a scenario. Framing the right scenarios, fairly, is where scenario planning succeeds or fails.

A scenario comparison gives you a ranking under one set of assumptions. The next, indispensable question is: how much does that ranking depend on the assumptions being right? This is sensitivity analysis, and it is what separates a robust conclusion from a fragile one. The method is to vary the uncertain inputs - growth rates, behaviour, costs, climate, uptake - across their plausible ranges and see whether the answer holds or flips.

The results fall into two very different kinds. A robust conclusion is one that survives: option A comes out best whether growth is high or low, whether people adopt the new service quickly or slowly, whether the climate warms moderately or severely. When a conclusion is robust across the plausible range of what you do not know, you can act on it with real confidence, because being wrong about the uncertain inputs does not change the decision. A fragile conclusion is one that flips: A wins if growth is high but B wins if growth is low, and nobody knows which growth will occur. A fragile conclusion is not useless - but it tells you something completely different: that the decision hinges on a specific uncertainty, and that the honest response is either to reduce that uncertainty (get better data on the thing that matters), to hedge, or to choose the option that does least badly across all cases rather than best in one. This is the core of what is sometimes called robust decision-making: prefer options that perform acceptably across many futures over options that are optimal in one and disastrous in others.

Sensitivity analysis also tells you where to spend effort. If the conclusion is wildly sensitive to one input and indifferent to ten others, you now know exactly which number is worth the cost of pinning down - and which modelling refinements are a waste of time. It is the antidote to the twin's most seductive failure mode: the precise, confident, single-scenario answer that would have been completely different under an equally reasonable assumption nobody tested. A professional never presents a scenario result without having asked what would have to be true for this conclusion to be wrong, and whether that is plausible. A result presented with no sensitivity analysis is a result you do not yet know whether to trust.

A scenario analysis is only as good as the decision it informs, and it informs nothing if it is communicated either dishonestly or uselessly. There are two opposite failures. Overclaiming hides the uncertainty: it presents a single confident number or a clean winner, gives the decision-maker false certainty, and sets them up to be blindsided when reality diverges - and to blame 'the model.' Paralysing drowns the decision in uncertainty: it piles up caveats and ranges until the decision-maker, unable to extract any actionable signal, either ignores the analysis entirely or freezes. The professional path threads between them: be honest about what is uncertain while still surfacing what can be said.

The craft is to communicate the shape of the result, not a point. Show ranges, not bare numbers. Make the comparison and the ranking the headline, since that is the robust part, and state clearly how confident you are in it. Separate what the analysis shows with high confidence ('every scenario that cuts car use improves air quality') from what it cannot resolve ('whether option A or B is cheaper depends on a land cost we cannot yet pin down'). Name the assumptions that matter most and flag the ones the conclusion is sensitive to. And use the visual and narrative power of the twin - which is real - to make uncertainty legible rather than to paper over it: a band of plausible outcomes, a set of scenarios a stakeholder can explore, a clear statement of what would change the recommendation.

This honesty is not just ethics; it is what makes the analysis survive contact with reality and keeps the twin trusted over time. A twin whose confident predictions are repeatedly wrong loses its authority and deserves to. A twin that says 'here are the options, here is what each would likely do, here is how sure we are, and here is what we are assuming' builds the kind of trust that lets it keep informing decisions for years. Honest uncertainty, well communicated, is a feature of a mature decision-support tool, not an admission of weakness - and it is exactly what distinguishes a twin used as an instrument of reasoning from a twin used as a prop for a decision already made.

Everything in this module converges on a single boundary, and it is the most important idea in the course: a scenario analysis informs a decision, but it does not make it - and must not be allowed to. No matter how sophisticated the simulation, the choice between competing urban futures is not a technical problem with a computable right answer. It is a question of values, priorities and trade-offs - more mobility for some against more peace for others, growth against preservation, cost against equity, this generation against the next - and those are political and ethical questions that belong to accountable people and democratic process, not to a model.

The danger, precisely because a twin is so powerful and so authoritative-looking, is that it becomes a way to launder political choices as technical ones - to say 'the model chose option A' when in truth people chose option A and used the model to justify it, or to let the options the model happened to test silently define the choices anyone gets to consider. A twin can narrow the debate to the variables it measures and the scenarios its framers chose, rendering invisible the values it does not encode and the people whose priorities were never made into a scenario. The antidote is to keep the human judgement and the value choices explicit and accountable: to be clear that the twin shows consequences, not preferences; that someone chose which scenarios to run and what counts as a good outcome; and that the decision, with all its trade-offs, is owned by people who can be held responsible for it.

This is also where scenario planning connects to participation (Module 8): the twin is at its best when its scenarios are shaped with the affected public, when what counts as a good outcome is debated openly, and when the tool helps citizens understand and weigh in on real choices rather than being presented with a decision already computed. Used that way - to structure an honest, inclusive, well-informed debate about options and their consequences, with uncertainty visible and values explicit - scenario planning is the twin at its most valuable. Used the other way - to deliver a single computed answer that ends the debate - it is the twin at its most dangerous. And through all of it, the binding technical results (the traffic capacity, the flood depth, the structural and energy engineering) stay with the qualified engineers, the statutory decisions with the authorities and the law, and the political choice with the people. The twin lays the options on the table, honestly and vividly; the city, through its accountable institutions, decides.