Means-Ends Analysis

Thinking Tools6.2 Creating an Action Plan

Developed by Allen Newell, J. C. Shaw, and Herbert A. Simon for the General Problem Solver (1959), and set out in full by Newell and Simon (Human Problem Solving, 1972).

A solving loop that compares where things stand with where they should be, picks the biggest difference, applies an action that shrinks it, and repeats.

Closing the gap to a self-serve launch, one biggest difference at a time

At each of the product lead's turns, pick the difference to work on next; the product's turn shows the new state, including any smaller gap the action opened. Each round compares where the product is with the goal again. Illustrative.

Product lead

The product, as it stands

Product lead

The product, as it stands

Select a turn to read it. At each product lead turn, pick the difference to work on.Where it stands. The goal: a small business can find the product, sign up, pay by card, and use the free plan, with no salesperson involved. Today there is no card billing, signup needs a salesperson, the free plan has no limits, there are no help pages, and the login page shows the old logo.First difference. Compare the state with the goal and pick the difference that matters most.Pick a move to see the reply.Two engineers build card payments in two weeks.Right: without payment there is no self-serve business at all.A designer swaps it in an afternoon.The smallest difference: done, and the launch is no nearer.A writer starts on forty articles.Needed later, but nobody can sign up to read them yet.After billing. Card payments work, but charging customers in other states needs sales tax collected, and the product does not do that. The action closed one difference and revealed a smaller one.The new difference. The billing action has a condition of its own.Pick a move to see the reply.Finance turns on the payment provider's tax service and checks three states.Right: an action's unmet condition becomes the next goal.The team moves on to signup.Missed: the billing gap is not closed until the product can charge legally.Two that block launch. Billing is complete. Two differences still block a launch: a new customer cannot create an account alone, and a free account can use everything the paid plans offer.Next difference. Both block launch. Which is larger?Pick a move to see the reply.The team builds signup with email confirmation in nine days.Right: no customer reaches the limits until they can sign up.Limits are built, for accounts nobody can open yet.Smaller for now: it matters only once signup works.After signup. In the first test week, forty accounts signed up on their own. All of them used paid features for free.Last blocking gap. Compare again: one difference still blocks launch.Pick a move to see the reply.Free accounts are capped at three users and one project.Right: the last difference that blocked launch is closed.The logo is refreshed.Polish: the product still gives paid features away.Launch-ready. The product meets the goal. Help pages and the logo are scheduled for the first month after launch, as improvements, not blockers.

Right calls: 0 of 4Right calls: 1 of 4Right calls: 2 of 4Right calls: 3 of 4Right calls: 4 of 4

  1. Where it stands (The product, as it stands). The goal: a small business can find the product, sign up, pay by card, and use the free plan, with no salesperson involved. Today there is no card billing, signup needs a salesperson, the free plan has no limits, there are no help pages, and the login page shows the old logo.
  2. First difference (Product lead). Compare the state with the goal and pick the difference that matters most.
    • Card billing: today nobody can pay without an invoice Two engineers build card payments in two weeks. Right: without payment there is no self-serve business at all.
    • The old logo on the login page A designer swaps it in an afternoon. The smallest difference: done, and the launch is no nearer.
    • Help pages A writer starts on forty articles. Needed later, but nobody can sign up to read them yet.
  3. After billing (The product, as it stands). Card payments work, but charging customers in other states needs sales tax collected, and the product does not do that. The action closed one difference and revealed a smaller one.
  4. The new difference (Product lead). The billing action has a condition of its own.
    • Set up sales tax now, before the rest of the list Finance turns on the payment provider's tax service and checks three states. Right: an action's unmet condition becomes the next goal.
    • Park it and return to the original list The team moves on to signup. Missed: the billing gap is not closed until the product can charge legally.
  5. Two that block launch (The product, as it stands). Billing is complete. Two differences still block a launch: a new customer cannot create an account alone, and a free account can use everything the paid plans offer.
  6. Next difference (Product lead). Both block launch. Which is larger?
    • Self-serve signup: nobody can get in without sales The team builds signup with email confirmation in nine days. Right: no customer reaches the limits until they can sign up.
    • Free-plan limits first Limits are built, for accounts nobody can open yet. Smaller for now: it matters only once signup works.
  7. After signup (The product, as it stands). In the first test week, forty accounts signed up on their own. All of them used paid features for free.
  8. Last blocking gap (Product lead). Compare again: one difference still blocks launch.
    • Limits on the free plan, with an upgrade prompt Free accounts are capped at three users and one project. Right: the last difference that blocked launch is closed.
    • Now redo the logo properly The logo is refreshed. Polish: the product still gives paid features away.
  9. Launch-ready (The product, as it stands). The product meets the goal. Help pages and the logo are scheduled for the first month after launch, as improvements, not blockers.

An example

A software company that had always sold through a sales team wanted a self-serve plan: a small business should be able to find the product, sign up, pay by card, and use a free tier without speaking to anyone. The product lead had six weeks and a backlog of forty tickets.

She set the goal beside the current state and listed what separated them: no card billing, signup only through sales, no limits on a free account, no help pages, and an old logo on the login page. She took the largest first. Card billing took two weeks and exposed a smaller gap of its own, since charging customers in other states required collecting sales tax, so tax became the next goal before anything else. Then she compared again. Signup came next, because nobody could reach a limit without an account, and free-plan limits last. Help pages and the logo moved to the month after launch, and the plan shipped in week five.

Why it works

When a goal is clear and the route is not, the number of possible next moves is too large to search. The method cuts it down by asking one question each round: what is the most important difference between where things are and where they need to be? Picking an action because it reduces that difference keeps every step tied to the goal, and comparing again after each step catches progress and new gaps alike.

The second half is what makes it more than prioritizing. An action often cannot be taken until some condition is met, and that condition becomes a goal of its own, worked the same way. The analysis recurses until each action can be done, so the plan grows only where the problem demands it.

The thinking behind it

Allen Newell Allen Newell, J. C. Shaw, and Herbert A. Simon built the General Problem Solver, described in their Report on a General Problem-Solving Program (1959), the program that first implemented means-ends analysis. It compared the current state with the goal, found the largest difference, and chose an operator that reduced it, setting up subgoals when an operator could not yet be applied. Newell and Simon set out the method in full, as a model of how people solve problems, in Human Problem Solving (1972).

Herbert A. Simon Herbert A. Simon, a Nobel laureate in economics and a Turing Award winner, is known for bounded rationality and satisficing: the view that people and organizations, unable to search every option, use simple rules to find good enough answers. Means-ends analysis fits that view. It is a rule for limited searchers, which is why it works as well for a product team as for a program.

How to run it

  1. Describe the goal precisely. Write the finished state as something checkable, such as a customer can pay by card with no salesperson involved.
  2. Describe the current state the same way. Use the same terms as the goal, so the two can be compared line by line.
  3. List the differences. Write every gap between the two, without solutions attached yet.
  4. Pick the most important difference. Choose the one whose closing matters most to the goal, usually the one other gaps depend on.
  5. Choose an action that reduces it. Name a concrete action, and check whether it needs anything that is not yet in place.
  6. Make unmet conditions their own goals. If the action needs something missing, work that gap first with the same method.
  7. Apply, compare again, and repeat. After each action, rewrite the current state and the list of differences before choosing the next.

Where it goes wrong

The easiest difference gets picked. Small gaps close quickly and feel like progress, so teams clear the cosmetic items while the blocker waits. Ask which difference, if closed, would matter most to the goal.

The goal is vague. A goal such as better onboarding cannot be compared with anything, so the differences never come into focus. Rewrite the goal until each part can be checked.

The route needs a step backward. Some problems require making the state look worse before it gets better, such as taking a system offline to migrate it, and a strict rule of always reducing the gap rejects that step. When progress stalls, look for a move that undoes something on purpose.

When to use something else

  • Issue Tree. When the problem is not yet understood, and the job is breaking it into parts before any action is chosen.
  • Backcasting. When the goal is far off and the route is best planned backward from the end state.
  • Theory of Constraints. When the system already runs and the question is which single step limits its output.

Where it sits in SOLVE eX

Go to the source

  • Report on a General Problem-Solving Program, RAND paper P-1584 (1958); UNESCO International Conference on Information Processing, Paris, 1959, 1959
  • Human Problem Solving, Prentice-Hall, 1972

Back to the catalog