Every model you’ve built on the path runs the same direction: inputs in, answer out. Set the price, see the margin. Set the contribution, see the retirement year. But the question in your head usually runs the other way: what price gives a 30% margin? What contribution retires me at 60? The forwards model answers it only by trial and error — nudge, look, nudge again.

Goal Seek is the trial and error, automated. You name three things: the formula cell, the value you want it to reach, and the one input Excel may turn. It nudges at machine speed until the answer matches.

Data → What-If Analysis → Goal Seek:

Set cell: margin · To value: 0.30 · By changing cell: price

Two seconds later: the price that produces exactly 30%. That’s the entire tool — three boxes, one solved question.

the input, turned… price = ? the model (untouched) margin → 30.0% ✓ target hit
You name the target and hand Excel one dial. It turns the dial until the model reads what you asked — trial and error at machine speed.

What it’s actually for

The pattern is threshold questions — anywhere a target is fixed and one input is negotiable:

And one habit makes all of it work: Goal Seek can only turn a dial that exists. It changes one hard-valued cell — if the rate is buried inside a formula as *1.2, there’s nothing to turn. Models built the way this series builds them — every assumption a named input cell — are Goal-Seek-ready by construction. If the tool feels unusable on your sheet, that’s the sheet reviewing your architecture.

The fine print

Three things worth knowing before you trust it. It overwrites the input — the found value is written into the cell, so note the original (or Ctrl+Z after reading the answer; better, run what-ifs on a copy of the inputs block). One dial, one target — it won’t split the effort across two inputs or satisfy two goals; that’s Solver, its heavyweight sibling (Excel’s free optimisation add-in — worth meeting the day a real constrained-optimisation problem walks in, and not before). It finds a solution, not the solution — on curvy models (percent changes, compounding) start the input near a plausible value; and some targets are simply unreachable, where it will politely report failure rather than inventing arithmetic.

If a question is recurring rather than one-off — “show me the required saving at every target date” — stop seeking repeatedly and lay the scenarios out as a Data Table, the next lesson. Goal Seek answers one backwards question crisply; Data Tables answer a hundred at once.

Set cell, to value, by changing. The model you already built knew the answer all along — Goal Seek just asks it the question from the direction you actually care about.