Cost-Benefit Analysis When the Costs and Benefits Aren't in the Same Units

A cost-benefit analysis compares money that arrives at different times by converting every future amount into today’s terms — a process called discounting — and then netting the totals against each other. It is a clean, well-established method, and the cost-benefit analyzer on this site does the discounting arithmetic correctly for whatever numbers you give it. The trouble is not the arithmetic. The trouble is that the method only works on numbers expressed in a single, shared unit — typically a currency — and a great many real decisions have costs and benefits that were never naturally denominated in dollars to begin with. This piece works through a real example to show exactly where that gap opens up and what to do about it honestly, rather than papering over it.

How the discounting actually works

The core idea is that a dollar of benefit next year is worth slightly less than a dollar of benefit today, because you could otherwise have that dollar now and put it to use in the meantime. The tool discounts each period’s net flow by 1 ÷ (1 + r)^t, where r is your chosen discount rate and t is how many periods out that flow lands, then sums the discounted benefits and costs separately. The results are the present value of the benefits, the present value of the costs, their difference (the net present value, or NPV — positive means worthwhile), the benefit-cost ratio (benefits divided by costs, where above 1 means worthwhile), and the payback period, the first point at which the cumulative undiscounted flows turn positive.

A worked example: insulating a home

Suppose upgrading your home’s insulation costs $4,000 up front and is expected to save roughly $650 a year on heating and cooling for the next six years. Feeding a $4,000 cost in period zero and a $650 benefit in each of periods one through six into the analyzer, at a 5% discount rate, returns a net present value of −$700.80, a benefit-cost ratio of 0.82, and no payback period at all — the tool reports it as never recovering the outlay within the six years modeled. Even setting discounting aside entirely (running the same flows at a 0% rate), the picture barely improves: total undiscounted benefit is $3,900 against a $4,000 cost, for a plain NPV of −$100 and a benefit-cost ratio of 0.98. On the numbers as entered, this upgrade does not pay for itself in dollar terms within the window modeled, discounting or no discounting.

Choosing a discount rate, and why it matters less than it looks

The discount rate is meant to reflect what you could otherwise do with the money — a rough opportunity cost of capital. For a household decision like this one, a common approach is to use a rate close to what you would otherwise earn on savings or pay on debt; 5% is a reasonable, unremarkable middle-of-the-road figure for this kind of multi-year household calculation, which is why it was used above. People sometimes agonize over the exact discount rate far more than the decision warrants. Compare the insulation example at 5% (NPV −$700.80) against the same flows at 0% (NPV −$100): the discount rate moved the answer, but the underlying verdict — not worthwhile on metered savings alone — held at both ends of a fairly wide range. When a decision’s conclusion is sensitive mainly to whether you count the soft, non-monetized benefits at all, and barely sensitive to a few points of discount rate either way, that is useful information: it tells you where to spend your effort refining the analysis, and the discount rate is usually not it.

What the payback period does and doesn't tell you

The tool’s payback period — the first period where cumulative undiscounted flows turn non-negative — is a simple, popular number precisely because it is easy to explain: “this pays for itself in five years.” It is worth being clear about what it leaves out, though. It ignores discounting entirely, so it can make a slow-payback investment look identical to a fast one even when the timing of the money is quite different in present-value terms. It also stops counting the moment the flows turn positive and says nothing about what happens afterward — a project with a slightly longer payback but much larger benefits in the years beyond it can easily be the better choice despite “losing” on payback period alone. Treat payback period as a rough, intuitive gut-check for how long your money is tied up, and treat net present value as the more complete answer to whether the decision is worthwhile overall; the two will not always point the same direction, and when they disagree, the NPV is the more trustworthy of the two.

Where the units stop matching

That verdict is completely honest about what it measured — the metered utility savings against the invoiced installation cost — and completely silent about everything else the upgrade might be worth. A better-insulated home is typically quieter, more evenly heated room to room, less prone to condensation and the maintenance problems that come with it, and arguably a little kinder to the environment through reduced energy use. None of those benefits arrive as a number on a utility bill. They are real, but they are not naturally denominated in dollars, and the cost-benefit analyzer has no field for “quieter” or “more comfortable” — it only accepts a benefit number and a cost number per period. This is not a flaw specific to this tool; it is a structural feature of cost-benefit analysis as a method, and it shows up anywhere a decision mixes something easily metered (a utility bill, a purchase price) with something that is not (comfort, time, stress, risk, goodwill).

Two bad habits tend to follow from this gap, in opposite directions. The first is silently dropping the non-monetary factors because they are inconvenient to quantify, which is exactly what the calculation above did — it is honest as far as it goes, but it understates the real case for the upgrade by construction, not because the upgrade is actually less valuable than a fuller accounting would show. The second, and more dangerous, habit is forcing a dollar figure onto a soft factor just to make the spreadsheet balance, without flagging how shaky that number is. Once a guessed value sits in the same column as a metered utility bill, it visually inherits the same authority, even though one number came from a meter and the other came from a shrug.

Watching a soft assumption flip the verdict

To see how much leverage a single soft number can have, suppose you decide the comfort and reduced-maintenance benefits are worth an extra $150 a year to you — a genuine guess, not a metered figure. Adding that to the $650 in metered savings gives $800 a year, and re-running the exact same six-year, 5%-discount analysis with that adjusted figure returns a net present value of +$60.55, a benefit-cost ratio of 1.02, and a payback period of 5 years. The verdict has flipped from clearly not worthwhile to marginally worthwhile, entirely on the strength of one estimated number that never touched a utility meter. That is not a criticism of including the comfort value — it may well be a completely reasonable number, and ignoring it entirely biases the analysis in the other direction just as much. The point is narrower and more important: whoever reads only the final NPV has no way of knowing that the entire verdict is currently balanced on top of a single $150-a-year guess, unless the analysis is presented in a way that makes that visible.

A more honest way to run the numbers

Three habits keep a mixed-unit cost-benefit analysis honest rather than quietly misleading. First, always compute and report the “hard floor” version — the NPV using only the costs and benefits you can point to a receipt, a meter reading, or a contract for. That is your defensible baseline, and in the example above it is the −$700.80 figure. Second, list every non-monetized factor separately, in its own terms, rather than skipping it or silently folding it into the dollar figure — “quieter home,” “fewer repair callouts,” “lower carbon footprint” can sit next to the NPV as explicit, named considerations rather than disappearing into it. Third, if you do want to see what a monetized estimate of those soft factors would do to the verdict, run the analysis a second time with your best-guess figure clearly labeled as an assumption, and report both numbers side by side — the hard-floor NPV and the assumption-adjusted NPV — so that anyone reading the result, including your future self, can see exactly how much of the case rests on a metered fact and how much rests on a judgment call.

This is not a reason to avoid quantifying soft factors when they matter to the decision; refusing to estimate them at all just substitutes one bias (treating them as worth exactly zero) for another (treating a wild guess as fact). It is a reason to keep the two kinds of numbers visibly separate rather than letting a spreadsheet’s uniform-looking columns erase the difference between what you measured and what you assumed. Run your own numbers through the cost-benefit analyzer for the hard-floor version first, then decide deliberately whether and how to layer a labeled assumption on top.