# Sunk cost vs relevant cost: what to ignore when you decide, with a worked example

> A sunk cost is spent whatever you decide, so it never belongs in a decision. How to find the relevant costs and the opportunity cost, with an example.

- Source: https://productovi.com/blog/sunk-cost-vs-relevant-cost
- By: Sergey Bruh
- Published: 2026-10-01
- Topics: Relevant and sunk costs, Cost allocation

A sunk cost is money already spent that no option on the table can bring back, so it never belongs in a decision, however large it is. What does belong are **relevant costs**: future cash amounts that differ between the options, including the **opportunity cost** of what the same people or capacity could do instead. The test for every number is one question: "Will this amount change depending on what we choose now?" If not, leave it out. Below: a product example where both sides of an argument lean on the wrong number, and a pricing example where the same lesson costs different amounts in May and July.

## Three kinds of cost, one test

- **[Sunk cost](/glossary/fin-sunk-cost):** already spent or committed and unrecoverable. Last quarter's development, a study you paid for, a prepaid annual licence.
- **[Relevant cost](/glossary/fin-relevant-cost):** a future cost (or revenue) that differs between the options. New hires, licences, extra variable costs, the margin a feature is expected to bring.
- **[Opportunity cost](/glossary/fin-opportunity-cost):** the best alternative you give up. It never shows up in the accounts, because nobody sends an invoice for the project you didn't do.

Costs that stay the same in every option (the office, the finance team's salaries, shared overhead) aren't sunk, but they're irrelevant to this choice for the same reason: they can't tip the comparison.

## Worked example: finish the simulator or stop?

Klasna, an online tutoring school, has spent 960,000 UAH over four months on a maths exam simulator. Two more months would finish it. The CEO says: "We can't throw that money away." The head of finance says: "It will never earn back 960,000, stop now."

Both arguments rest on the 960,000, and that number should not be in the room. It's gone whatever Klasna decides today. The right question is not "was it worth it?" but "is finishing worth it from here, compared with the best other use of the same people?"

The two developers stay on salary in every option, so their pay is out. What changes:

| Option for the next two months  | Contribution expected (UAH) | Extra costs (UAH)                        | Relevant result (UAH) |
| ------------------------------- | --------------------------- | ---------------------------------------- | --------------------- |
| Finish the exam simulator       | 380,000                     | 150,000 (editor 90,000 + licence 60,000) | 230,000               |
| Build package-renewal reminders | 260,000                     | 0                                        | 260,000               |

The working: $380{,}000 - 90{,}000 - 60{,}000 = 230{,}000$ for the simulator, against 260,000 for the reminders. The reminders win by $260{,}000 - 230{,}000 = 30{,}000$ UAH. Seen from the simulator's side, 260,000 is its opportunity cost: what finishing it would cost in the reminders the developers didn't build.

Notice what decides it: not the 960,000, not the salaries, not the rent. If Klasna had spent 2 million instead of 960,000, the answer would be exactly the same. If the reminder estimate dropped below 230,000, it would flip. Only new information about the future can change the decision.

![The 960,000 UAH already spent, the salaries and the rent fail the test because they are the same in every option; only the future amounts remain, and the reminders beat finishing the simulator by 30,000 UAH](/blog/sunk-cost-vs-relevant-cost/diagram.png)

## The same lesson costs different amounts in May and July

Opportunity cost matters most when capacity is full. A company asks Klasna for 200 English lessons at 520 UAH each, paid by bank transfer (no card fee), below the usual 620. Each lesson pays the tutor 390 and costs 30 in video, so it leaves $520 - 390 - 30 = 100$ UAH.

- **In July** tutors have free slots, nothing is displaced: $200 \times 100 = +20{,}000$ UAH. Say yes.
- **In May** every slot is booked. Each corporate lesson pushes out a family lesson that would have left 184.50: $200 \times (100 - 184.50) = -16{,}900$ UAH. The same deal now loses money.

The lowest acceptable price in a full month is the lesson's own variable costs plus the margin it displaces: $390 + 30 + 184.50 = 604.50$ UAH. In July the floor is 420.

## Sunk cost examples: where sunk and irrelevant costs sneak back in

1. **"We've already invested so much."** The classic sunk cost fallacy. Spending more to justify past spending is how projects run two quarters too long.
2. **"Use it up."** A prepaid annual plan is sunk. Running extra webinars "because we paid for it" makes sense only if each webinar covers its own extra costs and the tutors' alternative use.
3. **Fully loaded costs.** A report that charges every product a share of rent and salaries can make a profitable line look loss-making. Those shared costs don't disappear if you drop the line; the line's contribution does. That's the [allocation trap](/courses/finance-for-product/allocation-trap/theory).
4. **Free capacity that isn't free.** "The team is on salary anyway" is true for cash, false for opportunity: their time always has a best alternative.

## Honest limits

- **Estimates, not facts.** The 380,000 and 260,000 are forecasts. A 30,000 gap is small against that uncertainty; say so, and name the number that would flip the answer.
- **Sunk isn't always fully sunk.** If part of the spending can be recovered (equipment to resell, a refundable licence), the recoverable part is relevant.
- **Some effects are hard to price.** Credibility with customers who were promised the simulator, or a developer's motivation, are real; list them next to the numbers rather than inside them.
- **Short-term vs long-term.** Selling spare July capacity near variable cost is fine once; as a habit it can reset what customers expect to pay.

## Key takeaways

- A sunk cost is spent in every option, so it can't help you choose between them.
- Relevant costs are future amounts that differ between options; everything else stays out.
- Always price the opportunity cost: the best other use of the same people, money or capacity.
- The cost of capacity depends on how full it is, so the minimum price does too.

## FAQ

### What is the sunk cost fallacy?

Continuing a project because of what you've already spent on it, rather than because of what it will bring from now on. The past spending is the same whether you continue or stop. In the worked example above, "we can't throw away 960,000" is the fallacy in its plainest form; more sunk cost examples are in the section on where sunk costs sneak back in.

### Are salaries a sunk cost?

Future salaries aren't sunk, but if they're paid in every option they're irrelevant to the choice. What matters is the opportunity cost of the people's time.

### Is opportunity cost a real cost?

It isn't in the accounts, but it is in the decision. Ignoring it makes any project look good as long as its own extra costs are covered.

## Learn it hands-on

Module 8 of the free course _Finance for Product People_ works through this on Klasna's numbers: [relevant, sunk and opportunity costs](/courses/finance-for-product/relevant-and-sunk-costs/theory), then [the allocation trap](/courses/finance-for-product/allocation-trap/theory), where a whole subject looks loss-making on a report. If you're new to margins, start with [contribution margin per lesson](/courses/finance-for-product/contribution-margin/theory). For investments that span several years, see [NPV vs IRR vs payback](/blog/npv-vs-irr-vs-payback).

## Related lessons

- [Relevant, sunk and opportunity costs](https://productovi.com/courses/finance-for-product/relevant-and-sunk-costs.md)
- [The allocation trap: should we drop programming?](https://productovi.com/courses/finance-for-product/allocation-trap.md)
- [Contribution margin per unit: what one lesson leaves](https://productovi.com/courses/finance-for-product/contribution-margin.md)