Scenario planning: how to run it, step by step, with an example

Who should use it, and when not to

Use scenarios before a long commitment that is hard to undo: a lease, a factory, a hiring plan, an acquisition, a new country. They also help when a team keeps arguing about one forecast and the honest answer is "it depends", or when an annual review needs to test whether the current strategy survives a bad but plausible turn in regulation, technology, or demand.

They are the wrong tool for next quarter's numbers, which need a forecast, and for comparing growth routes in a known market, which the growth options matrix does faster. Scenarios earn their cost when the outside world could change the answer.

The core concepts

Focal question. The decision the scenarios serve, with its time horizon: "Should we sign an eight-year lease for a second site?" A topic such as "the future of retail" produces stories nobody acts on.

Driving forces. The wider social, economic, political, environmental, and technological forces behind the local factors that decide the outcome: customers, prices, costs, competitors, staff.

Predetermined elements and critical uncertainties. Some forces are already under way and will show up in every scenario. Others are both important and genuinely open. Rank the forces on importance and uncertainty, and take the two that score highest on both.

Two axes, four worlds. Give each critical uncertainty two clear end states and cross them. Each quadrant is a scenario; name it so people remember it, and describe it as a world at the horizon.

Implications, signposts, and moves. Go back to the decision and test it in each world. A decision that works in only one scenario is a gamble; name it as one. Signposts are early, observable signals that the world is moving toward one scenario. Robust moves make sense in most scenarios; contingent moves wait for a signpost.

A worked mini example

An illustration with made-up facts. Harbourline English runs one adult English school in Fukuoka with about 700 students; 60% of revenue comes from individuals and 40% from company contracts. The landlord offers a second site near Hakata station on an eight-year lease, to sign or decline within two months. The city's population aged 20 to 39 has been falling for years. Since 2023, 30% of new individual students choose online classes. Two company clients cut training budgets this year and three increased theirs. Nobody knows whether better machine translation will reduce demand for lessons or raise demand for real conversation.

The two critical uncertainties are how adults want to learn (in person or online) and whether company training budgets grow or shrink. The falling young-adult population is treated as predetermined and appears in every scenario.

The lease works well in one world out of four, so signing is a bet on Full Rooms. The robust move is to grow online classes and sell them to company clients, which is useful in both online scenarios and cheap to keep in the other two. The contingent move is short-term or hourly classroom space, triggered if company clients add classroom hours two quarters in a row. Signposts: company renewals each quarter and the online share of new students each month. The answer: decline the eight-year lease or ask for a break clause, and review the scenarios in six months.

How to run it

1. Write the focal question. Name the decision and the horizon in one sentence, and ask the decision owner to confirm it.
2. List the key factors. Write the local factors that decide whether the decision succeeds: demand, prices, costs, competitors, staff, suppliers.
3. List the driving forces. For each key factor, write the social, economic, political, environmental, and technological forces behind it, using facts the team can source.
4. Separate what is settled from what is open. Mark predetermined elements, then rank the remaining forces on importance and uncertainty and choose the top two.
5. Build four scenarios. Give each uncertainty two end states, cross them, name each quadrant, and write a paragraph per world with every predetermined element present.
6. Test the decision in each scenario. Write what goes well, what breaks, and what you would wish you had done, and call a decision that works in only one world a gamble.
7. Choose signposts. Pick one to three observable signals per scenario, with where to watch, how often, and who watches.
8. Sort the moves and decide. List robust moves to make now, contingent moves tied to signposts, and moves to avoid, then answer the focal question and set a review date.

Common mistakes

• Best, worst, and middle cases. That is one forecast with error bars, not different worlds.
• Axes that don't matter to the decision, or that matter but are already settled.
• Stories without implications. The signposts and moves are the point; rich narratives alone change nothing.
• A quiet favourite. Labelling one scenario "most likely" turns the exercise back into a forecast.
• No owner for the signposts. If nobody checks them, the scenarios can't trigger anything.

Where it comes from

Pierre Wack, head of the business environment division in Royal Dutch/Shell's planning department, developed Shell's scenario system with Edward (Ted) Newland and described it in two 1985 Harvard Business Review articles, "Scenarios: Uncharted Waters Ahead" and "Scenarios: Shooting the Rapids". Wack argued that forecast-based planning worked in the steadier 1950s and 1960s but that forecasting errors had grown more frequent since the early 1970s. Art Kleiner's account in strategy+business describes a 1972 report to Shell's managing directors that anticipated an energy crisis, and says Shell's modest preparations left it well placed when the crisis came; it also notes that Shell later ignored some of its own scenario warnings, and that Wack concluded scenarios alone did not do enough to change managers' mental maps. Peter Schwartz, a former member of Shell's planning group, set out a step-by-step method in The Art of the Long View (Doubleday/Currency, 1991). Crossing two uncertainties into a 2×2 grid is a common way to build the scenarios, not the only one; Ramirez and Wilkinson (2014) note that researchers call it the standard approach, and discuss its limits. This page is an independent summary in our own words, not affiliated with or endorsed by the authors, their publishers, or Shell. More frameworks are in the strategy book guides.

Questions

What is the difference between scenarios and forecasts?

A forecast names the one future it expects. Scenarios describe several plausible futures, none called most likely, so a decision can be tested against each and adjusted as signals appear.

How many scenarios should I make?

Four is the common number: cross the two most important and most uncertain forces as axes. Fewer can work; more than four is hard for a team to keep in mind.

Did Shell predict the 1973 oil crisis?

Not in the sense of naming a date. Accounts of Shell's work, such as Art Kleiner's in strategy+business, say a 1972 report to Shell's managing directors anticipated an energy crisis and that Shell began preparing, which left it well placed when prices jumped.