Payback vs NPV vs IRR: when the three disagree about the same site

Ask three investors whether a charging site is worth building and they may reach for three different numbers: how fast the money comes back, what the deal is worth today, and what rate of return the cash flows imply. Most of the time the three agree. The interesting cases are the ones where they don't — and DC fast charging, with its slow ramps and long holds, produces those cases naturally.

Here is a worked example where the same two sites rank differently on all three metrics. The numbers are deliberately simplified — flat and smoothly growing cash flows, no taxes or financing — to isolate how the metrics behave, not to describe any real site.

Two hypothetical sites#

Both cost $500,000 to build. They differ only in the shape of their cash flows over a 15-year hold:

• Site A produces a flat $120,000 a year from day one — think mature corridor traffic that is already there.
• Site B starts at $60,000 and grows 12% a year — think a developing area that compounds.

Payback says A, decisively. NPV says B, narrowly. IRR says A again. Same sites, same arithmetic, three rankings.

What each number encodes#

Payback answers "how long is my capital at risk?" — it counts years until cumulative cash equals the outlay and ignores everything after. Site B's best years, the ones growing past $200,000, are invisible to it. That blindness is also its honesty: payback is the only one of the three that says nothing about forecasts beyond the recovery point, which is why lenders and cautious operators keep it around.

NPV answers "what is this stream worth today, in dollars?" — every year's cash, discounted back at a chosen rate, minus the outlay. It is the only metric of the three denominated in money, and the only one that naturally rewards Site B's back-loaded compounding. It is also the only one that depends on an input the site itself doesn't supply: the discount rate. At 10%, B wins by about $19,000. Rerun the same table at 12% and the ranking flips — A's near-term dollars lose less to discounting than B's distant ones. The choice of rate is doing real work, which is a reason it deserves its own scrutiny rather than a default nobody read.

IRR answers "what rate does this stream pay?" — the discount rate at which NPV crosses zero. It is scale-free and time-weighted, which makes it comparable across deals of different sizes and the native language of anyone comparing against a hurdle rate. Its known quirk is reinvestment: a high IRR on early cash flows assumes those dollars keep earning at that rate somewhere, which the metric itself cannot guarantee. A's fast, front-loaded recovery is exactly the shape IRR flatters.

Why the disagreement is information#

The three metrics disagree here because the sites differ in shape, not quality: A is front-loaded, B is back-loaded, and each metric weighs time differently. That means the disagreement itself tells you what you are choosing between — capital velocity (A) versus terminal scale (B) — and which belief the choice rides on. Preferring B means trusting a 12% growth forecast for a decade and a half; preferring A means valuing recovery before year five more than dollars after year ten.

A single headline number would have buried that trade under a ranking.

What the model reports#

This is why a ForgeAsset scenario reports payback, NPV, and IRR together — alongside the cash-on-cash multiple and peak cash deficit — on top of the full 15-year statement, and lets the discount rate be an input rather than an opinion baked into the output. The returns pillar covers what ranges the model produces across site types; the wizard computes all five for a specific address and set of assumptions. Which number should decide? The model doesn't say — it shows where they agree, where they don't, and what assumption each verdict leans on.