zevOS

01Tool

Charging site ROI calculator

Most charging business plans project revenue. This one works backwards: what utilisation does the site need to cover its fixed costs? If the location cannot plausibly reach that number, no pricing change will save it.

The site
Commercial terms
Costs

Result

7.1%

Break-even utilisation

At 8% utilisation this site covers its fixed costs and contributes ₹3,386 a month.

Energy delivered
4,608 kWh/month
Revenue
₹92,160/month
Variable cost
₹13.46/kWh
Contribution
₹6.54/kWh
Demand charge
₹18,750/month
Total fixed cost
₹26,750/month
Monthly profit
₹3,386/month
Payback
472 months (39.4 years)

Nothing you type here leaves your browser. This is a planning model, not financial advice — the demand-charge treatment in your state and your real utilisation will move these numbers more than anything else.

How to read it

Three things people get wrong

The model is simple. The assumptions are where the errors live.

The full unit-economics walkthrough

Rated power is not delivered power

A 60kW DC charger rarely averages 60kW across a session. Vehicle taper, arrival state of charge and thermal derating mean 35–45kW is a more honest planning figure. Using the nameplate rating overstates revenue by roughly a third.

Demand charges do not scale with usage

They are levied on sanctioned or peak demand regardless of units sold. This is why an under-utilised DC site loses money at a healthy margin per unit — and why reducing sanctioned load through smart charging is a permanent monthly saving.

Utilisation is a property of the location

New public sites run 3–6%. Mature urban DC reaches 10–15%. If your break-even needs 18% and the location does not plausibly support it, the answer is a different site, not a different price.

Want the model behind this, with the assumptions written down instead of buried in cells? It is worked through in the CPO business toolkit.

Run your real numbers with us

Bring a site you are considering. We will go through the DisCom questions to ask and where the model is most likely to be wrong.