Back to Glossary

Entry · Economics

Backward Induction

Backward induction is a way of solving a multi-stage decision by starting at the end and reasoning back to the present. You work out the best move at the final stage, assume that move will be taken, use its value to evaluate the stage before it, and keep stepping backwards until you reach the choice you face today.

It is the engine behind decision trees, staged investment appraisal and much of game theory.

From the Money Master HQ dictionary, founded by Shihan Sheriff (FCMA, VP of Finance at Nomod, CFO at Esanjo Ventures). How these definitions are written.

What it means

The intuition is simple: you cannot sensibly decide what to do now until you know what you would do later. By solving the last decision first, you replace an entire branch of possible futures with a single number, which makes the earlier decision tractable.

In business the technique shows up most often in staged investments. A company that can spend a small amount on a pilot before committing to a full build is buying the right to stop, and only backward induction values that right properly, because it recognises that the second-stage decision will be made with better information.

It is equally useful in negotiation and competitive strategy. If you can work out what a rival will rationally do at the last move, you can predict the whole sequence, which is why price war analyses and entry decisions are usually solved from the end backwards.

The mechanics are mechanical rather than mysterious. At a chance node you take the probability-weighted average of the branches, at a decision node you take the highest value available, and you keep folding the tree back until only the present decision remains.

The limits are worth stating. Backward induction assumes you can specify the possible outcomes and their probabilities, and that the other side behaves rationally, so it works well for structured capital decisions and less well where behaviour is erratic or the future is genuinely unmappable.

In practice

Real-world examples.

1

Example

A pharmaceutical company evaluates a compound with three trial phases. It solves the tree from the final phase backwards, and discovers the programme is worth funding only because it can abandon after phase two if efficacy data disappoints.

2

Example

A retailer decides whether to sign a ten-year lease with a break at year five. It first works out what it would do at year five under strong and weak trading, then values the lease today with those responses built in.

3

Example

A software firm plans a negotiation with a large customer over a renewal. By reasoning that the customer would accept a 5% discount rather than face switching costs at the final round, it opens at a smaller concession than it originally planned.

Formula

Calculation

Value at a chance node = sum of (probability of each branch x value of that branch) Value at a decision node = the highest value among the branches available A specialty chemicals firm is considering a new production process. It can spend $2,000,000 on a pilot plant now, and if the pilot works it can spend a further $6,000,000 on a full line that would generate $12,000,000 of value. If the pilot fails, a scaled-down version of the line would generate only $3,000,000 of value. Management assesses a 40% chance of technical success. Solve the last decision first. Following success, building the full line is worth $12,000,000 - $6,000,000 = $6,000,000, so the firm builds rather than taking $0 for stopping. Following failure, building would be worth $3,000,000 - $6,000,000 = -$3,000,000, so the firm abandons and the branch is worth $0. The chance node is therefore worth (40% x $6,000,000) + (60% x $0) = $2,400,000. Subtracting the pilot cost gives $2,400,000 - $2,000,000 = $400,000, so the staged plan is worth doing. Committing $8,000,000 up front instead would be worth (40% x $12,000,000) + (60% x $3,000,000) - $8,000,000 = $4,800,000 + $1,800,000 - $8,000,000 = -$1,400,000, which shows that the value lies in the option to stop.

Case study

Seen in the real world.

Torrance Robotics is a fictional company created to illustrate backward induction in practice. Its board was split over a new welding cell: engineering wanted to commit $8,000,000 immediately to hit a customer deadline, while finance argued the technology risk was too high.

The chief financial officer built a two-stage tree instead. Spending $2,000,000 on a pilot bought the right to walk away, and the analysis showed the staged route was worth about $400,000 while the all-in commitment was worth about -$1,400,000, a swing of $1,800,000 driven purely by the ability to stop after the pilot.

The board approved the pilot. It failed on a materials issue eleven months later, Torrance abandoned the project, and the loss was $2,000,000 rather than the $5,000,000 it would have written off from a full build. The illustrative point is that backward induction does not predict the future; it prices the freedom to change your mind.

Watch out

Common mistakes.

  • Starting the analysis at today's decision. Working forwards forces you to guess at later choices, which is exactly what backward induction is designed to avoid.
  • Assuming the later decision must be taken. The whole value of staging comes from being willing to abandon, so a tree that has no stop branch will overstate the case for investing.
  • Using it with invented probabilities and treating the answer as precise. The output is only as good as the inputs, so the sensible use is comparing options and testing how far the probabilities can move before the answer flips.

Questions

People also ask.

Is backward induction the same as a decision tree?

Not quite, since the tree is the diagram and backward induction is the method used to solve it from the final nodes back to the first.

Where does it come from?

It is a standard technique in dynamic programming and game theory, where it identifies the equilibrium strategies of a sequential game.

How does it relate to real options?

Real options analysis prices managerial flexibility, and backward induction is one of the practical ways to calculate that value when the decision points are discrete.

Was this explanation helpful?

From the founder's library

Accounting Fundamentals: A Non-Finance Manager's Guide to Finance and Accounting, by Shihan Sheriff

Take it further with the book.

Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.

US$2.24US$2.99

25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.

View the book and save 25%
Last updated · October 8, 2026
Browse all terms →

Disclaimer

The information provided in this finance dictionary is for educational and informational purposes only. It should not be construed as financial, investment, legal, or tax advice. Always consult with a qualified professional before making any financial decisions. Money Master HQ makes no representations or warranties about the accuracy, completeness, or suitability of this information. Use of this content is at your own risk.