• AC charging
• DC charging
• The battery
• The handshake
• Why sessions fail
• Payments, software, hardware

AC charging: the charger is in the car

A Level 2 "charger" is not a charger. It is a switch with opinions. It delivers grid AC to the vehicle and the vehicle's onboard charger rectifies that AC into DC for the battery. The onboard charger is the bottleneck, and it varies by model: 7.7 kW on many cars, 11 kW on some, 19.2 kW on a few. Two cars on the same 11 kW unit will charge at different speeds and neither one is broken.

The wall unit and the car talk over the control pilot, a 1 kHz square wave on a dedicated pin. The duty cycle of that wave tells the car how much current the circuit can supply. The car answers by pulling the pilot voltage down through a resistor to say it is connected and ready. Only then does the relay close. A separate proximity pin tells the car a plug is in the socket, so it won't drive off with the cable attached.

Practical consequence: Level 2 hardware is cheap and simple, and almost all of the cost of a Level 2 project is in the concrete and the copper.

DC charging: the charger does the work

A DC fast charger skips the car's onboard charger entirely. The cabinet takes AC from the grid, rectifies it into DC with a stack of power modules (commonly 15 to 75 kW each), and feeds that DC straight to the battery pack through the contactors. More modules, more power, more money, more heat, more things to fail.

Because the modules are shared, a four-dispenser site may not deliver full power to four cars at once. Power sharing is a feature, not a defect, but it is also why "four 150 kW stalls" sometimes means "600 kW split four ways when busy." Ask how power is allocated before you sign.

Communication is where it gets interesting. CCS and NACS/J3400 use HomePlug Green PHY power line communication over the pilot pins, running DIN 70121 or ISO 15118. CHAdeMO uses CAN. ISO 15118 is what enables Plug and Charge, where the car presents a certificate and authorizes itself with no app or card at all.

Pack voltage matters too. Most EVs run a roughly 400 V pack, newer ones run roughly 800 V. Chargers have a voltage window. A car outside that window either charges slowly through a boost stage or doesn't charge at all.

The battery, in the terms that affect your revenue

A pack is a few thousand cells wired into modules, managed by a battery management system that owns every decision about current, temperature, and safety. Chemistry drives behavior. NMC packs hold more energy per pound and take fast charging well. LFP packs are cheaper and longer lived, tolerate a daily 100% charge, and tend to taper harder at the top and suffer more in the cold.

Charging follows constant current, then constant voltage. The pack takes a high, steady current while there is room, then the current tapers as it fills so cell voltage stays inside safe limits. That taper is why the marketing number is 10% to 80%, and why the last 20% can take as long as the first 60%.

• State of charge is the single biggest driver of session speed. A car arriving at 60% will never see peak power, no matter how big your charger is.
• Temperature is the second. A cold pack accepts very little current until it warms. Cars that precondition on the way to a charger do far better, which is why navigation-triggered preconditioning matters.
• C-rate is power divided by capacity. A 75 kWh pack at 150 kW is charging at 2C, which is demanding. The same 150 kW into a 150 kWh pack is a gentle 1C.
• Degradation comes from heat, time spent at high state of charge, and repeated fast charging at high state of charge. It is real, and it is slower than the internet believes.

The handshake, step by step

A DC session is a negotiation, and every step is a place to fail.

Why sessions fail

Not actually failures

• Taper. The car arrived at 70% and is doing what it should.
• Cold pack. Power climbs as it warms. Nothing to fix.
• Power sharing. Two cars, one power cabinet.
• The car's onboard limit on Level 2. You cannot sell 11 kW to a 7.7 kW car.

Real failures, roughly in order of how often they bite

• Payment authorization. The terminal cannot reach the processor, the pre-auth is declined, or the card is a type your terminal was never certified for.
• Connectivity. The cellular modem sits in a metal box at the edge of a parking lot. Signal there is worse than you assumed on site visit day.
• Communication timeouts between car and charger, often specific to one vehicle model and one firmware version. These are the hardest to diagnose and the most common cause of "it works for everyone but my car."
• Isolation and ground faults, usually water in a connector or a cable that got run over.
• Cables and connectors. The single most abused component on the site. They are consumables. Budget for them.
• Contactor and power module failures, which take the stall down until a technician with parts arrives.
• Firmware updates that fix one vehicle and break another, or that break the payment terminal driver.
• Blocked stalls, whether by a gas car or by a driver who finished an hour ago.

Payments, software, and hardware: where the journey actually breaks

Three vendors usually meet inside one steel cabinet: whoever built the hardware, whoever runs the charge management system (CMS), and whoever processes payments. The driver does not know or care. The driver knows the screen said "authorizing" for ninety seconds and then nothing happened.

Where it breaks

• Any hop that needs the network fails when the network is down, and the network is down more often than the site survey implied. Decide in advance what the charger does offline: refuse sessions, or start them and hope you can bill later. Both answers cost money. Pick one on purpose.
• Pre-auth holds look like charges to drivers. Expect calls about the $75 you did not actually take.
• Terminal certification is tied to a specific processor. Changing processors can mean changing terminals, which can mean a truck roll to every stall.
• Missing meter values turn into billing disputes you will lose.
• Version mismatches between hardware firmware and CMS expectations cause subtle failures that each vendor blames on the other. OCPP 1.6J is common, 2.0.1 is where things are going, and "OCPP compliant" means almost nothing without a documented profile.
• Refunds and chargebacks have to be possible for whoever answers the phone, which is usually the site host who has no access to any of these systems.

• What you sell
• Hardware
• Revenue
• Costs
• Demand charges
• Ownership models
• Incentives
• Reliability
• How sites fail
• Glossary

What you are actually selling

Drivers think they're buying electricity. They're buying time. A DC fast charger sells thirty minutes back to someone halfway through a road trip. A Level 2 charger at a hotel sells a full battery by breakfast. Price and position the site around the time, not the kWh.

The business, meanwhile, runs on one number: sessions per port per day. Nearly every cost is fixed (hardware, construction, software, maintenance, demand charges) and nearly every dollar of revenue is variable. Low utilization doesn't mean low profit. It means real losses, because the fixed costs show up whether anyone plugs in or not.

Almost everything else in this business is a way of moving that one number, and the biggest lever on it by far is the address.

The hardware, in one pass

Match the charger to the dwell time. A 350 kW charger at an apartment complex is a monument to someone else's budget. A 7 kW charger at an interstate exit is a customer service incident.

Connectors: North America is moving from CCS1 to NACS, standardized as SAE J3400. New DC sites should plan for J3400, and most networks keep CCS1 alongside it through the transition. Level 2 uses J1772 or J3400.

A port is a connector that can charge a car at the same time as the others. A charger (or dispenser) may have one or two ports, and some split power between them. When you compare quotes, count ports and simultaneous power, not boxes.

Where the money comes from

• Energy sales, priced per kWh where state weights and measures rules allow it (most now do), otherwise per minute or per session.
• Idle fees for cars that finish charging and keep the spot. Less about revenue, more about turnover.
• Session fees, memberships, and subscription pricing.
• Fleet contracts. One delivery fleet committed to overnight charging is worth more than a hundred random drivers.
• Environmental credits, such as Low Carbon Fuel Standard credits in California, Oregon, and Washington. Real money, volatile price.
• Indirect value to the host: dwell time, foot traffic, room nights, tenant retention. This is often the actual reason a host says yes.

Where the money goes

Up front

• Hardware. The line item everyone negotiates hardest and the one that matters least by the end.
• Make-ready: trenching, conduit, switchgear, panels, pads, bollards, signage, accessible stalls. Where estimates go to die. Estimate yours.
• Utility service upgrades, such as a new transformer or a new service. Can cost more than the chargers and take a year or longer to schedule.
• Design, engineering, permitting, and commissioning.

Every month

• Electricity: energy charges plus demand charges. See below, and bring a drink.
• Network software and connectivity, usually priced per port per year.
• Maintenance and warranty. Budget for it on day one, because the chargers will not ask permission to break.
• Payment processing, host revenue share or lease, and insurance.

Demand charges, the silent killer

Most commercial electric bills have two parts. Energy is billed per kWh used. Demand is billed per kW of the highest power draw in the month, usually measured over a 15 minute window. A 150 kW charger that runs flat out for fifteen minutes, once, sets that month's demand at 150 kW.

At $15 per kW, that is $2,250 before you've sold much of anything. Spread across one 30 kWh session, the electricity cost you about $75 per kWh. Spread across 600 sessions, it's closer to $0.25. Same charger, same tariff, different business.

Ways to fight it

• EV-specific rate schedules. Many utilities now offer them, often with reduced or phased-in demand charges.
• Power management: capping site power or sharing it across ports so the peak stays lower.
• Battery storage to shave peaks. It works, but it's a second project with its own capex.
• Right-sizing. Building for the traffic you'll have in year five means paying demand on it in year one. Build the conduit for later, energize for now.

Who owns what

The question that picks the model: who is better positioned to survive a slow first three years? Whoever that is should own the utilization risk, and get paid for it.

Incentives, as of fall 2026

• The federal 30C tax credit is done for new projects. It doesn't apply to property placed in service after June 30, 2026.
• NEVI, the federal highway charging formula program, was paused and reissued under new guidance in 2025. The money flows through state DOTs, each on its own schedule. Ask yours directly.
• Utility make-ready programs are often the best money left, because they pay for the expensive part.
• State grants and rebates vary widely. The AFDC laws and incentives database and DSIRE are the fastest way to see what applies.
• Grants are free money with a price: reporting, uptime requirements, and wage rules. Price the strings before you celebrate.

Reliability is revenue

A broken charger is worse than no charger. It costs you the session, the review, and the driver's willingness to try you again. NEVI-funded ports have to hold 97% uptime, which is a reasonable bar for anyone.

• Know who fixes it and how fast, in writing, before commissioning.
• Stock spares for the parts that actually fail: cables, connectors, payment readers, power modules, communication boards.
• Parts shipped from overseas bring lead times and customs paperwork. A two-week repair becomes a two-month one without stock on the shelf.
• Monitor successful sessions, not just online status. A charger can report online all day and fail every session.

How sites fail

• Picked for the land deal, not the traffic.
• Make-ready and utility work estimated at a fraction of what they cost.
• Demand charges modeled as an afterthought, or not at all.
• Nobody owns maintenance.
• Priced below the all-in cost of a kWh because the competitor down the road did it first.
• Built for 2032 traffic in 2026.

Glossary

AADT

Annual average daily traffic. Vehicles passing a point on a road on an average day. State DOTs publish it.

CPO

Charge point operator. The company that runs the chargers and usually owns the relationship with the driver.

EVSE

Electric vehicle supply equipment. The charger.

eMSP

E-mobility service provider. An app or network that lets drivers start and pay for sessions on chargers it doesn't own.

OCPP

Open Charge Point Protocol. How chargers talk to back-end software. Insist on it, so you can change software vendors without changing hardware.

OCPI

Open Charge Point Interface. How networks talk to each other for roaming.

kW and kWh

kW is power, how fast. kWh is energy, how much. Demand charges bill kW. Drivers buy kWh. Most bad pro formas confuse the two once.

Make-ready

Everything electrical and civil up to the charger itself.

Utilization

Share of time a port is in use, or sessions per port per day. This site uses both.

Dwell time

How long the car is parked. Pick the charger speed to match it.

NACS, J3400, CCS1

Connector standards. NACS was standardized as SAE J3400 and is becoming the default in North America. CCS1 is the incumbent.

• Same box, four addresses
• Why location wins
• Scarcity
• Underwrite it like a development
• Site control
• Value beyond the plug
• Walk the site

Same box, four addresses

Here is the whole argument in one picture. Take two identical fast chargers with the same cost, the same price and the same everything else, and install them at four different addresses. The only input that changes is how busy each location actually turned out to be, using real utilization from the benchmark sites.

Why location wins

Look at what actually varies between charging sites. Hardware prices differ by maybe two to one. Prices to drivers differ by about the same. Utilization, in the real data on this site, differs by forty to one. When one input moves forty times more than everything else combined, that input is the business.

Location sets the ceiling. Hardware, pricing and operations decide how close you get to it. Good equipment can't fix a bad address, and a great address forgives a surprising amount.

What good looks like

A well-chosen urban site typically keeps its chargers busy 5% to 10% of the time. That doesn't sound like much, but on the box above it's the zone where the business works. With standard electric rates, a good urban site sits around break-even, and the top of the range earns back the build. Add an EV-friendly electric rate, which many utilities now offer, or price a little above the minimum, and most of the range makes money. Good real estate gets you into the game. Good execution wins it.

The newer benchmark data backs this up. The best-placed chargers in it, at a high-end shopping center and at the only fast charger in a remote small town, each keep a charger busy about 6% to 7% of the day. That's the band, measured.

Scarcity is real estate too

Prime real estate isn't the only way to win. The newer data has a remote small-town site in oil country, the only fast charger for miles, that is as busy per charger as a high-end shopping center in a major West Coast city. Drivers passing through have exactly one option, and they use it.

Scarcity works for the same reason prime locations do: demand with nowhere else to go. It comes with two conditions. There has to be real traffic passing through, since being the only charger in a place nobody drives to is just lonely. And you should price like the only option, fairly but not timidly, because that pricing power is part of what the location is worth.

Underwrite it like a development

Every question a real estate developer asks about a retail parcel has a charging version. Ask all of them before anyone orders a charger.

• Traffic and frontage. Volume on the adjacent road, which side of it you're on, and whether drivers can turn in without crossing three lanes. Awkward access has quietly killed more sites than bad chargers.
• Who drives past, not just how many. Rideshare, delivery and other high-mileage drivers charge every day. The best site in the benchmarks is a convenience store near an airport, where those drivers wait between trips. A steady stream of them is worth more than a big traffic count.
• Visibility. If drivers can't see the chargers from the road, or find them in the app photo, the chargers might as well not exist.
• Anchors and dwell. What is there to do for thirty minutes? Food, coffee, shopping, somewhere to sit. The anchor tenant of a fast-charging site is a clean bathroom.
• Access and circulation. Stalls you can pull through with a trailer, good lighting, and access around the clock. A gate that locks at 10pm is a business that closes at 10pm.
• Power at the parcel. Available capacity, the distance to the electrical gear, and the utility's timeline. It's the charging version of having utilities at the lot line, and the install estimator shows what distance costs.
• Competition. Who is nearby, how fast their chargers are, and what is being built. The garage in the benchmarks lost its drivers to faster chargers down the road.
• Growth. EV adoption in the area, new housing and new roads. Underwrite the market you'll have in year five, but build for the one you have today.
• Entitlements. Zoning, permits, parking minimums (turning required parking spaces into charging stalls can run into them), and any easements the utility will need.

Measure it like a developer

Developers judge a project by its yield on cost: the annual operating profit once the site has settled in, divided by everything it cost to build. The interactive above shows it for each address. The test is whether the finished site would be worth more than it cost to build. If it would, you created value. If it wouldn't, you built an amenity, whatever the spreadsheet called it.

Site control is the asset

In real estate, if you don't control the land, you don't really own the business. Charging works the same way. The lease is as important as the chargers, and it usually gets a fraction of the attention.

• Term. At least as long as the equipment's useful life, with renewal options. A five-year lease on a ten-year asset is a gift to the landlord.
• Access. Around the clock, for drivers and for technicians, in writing.
• Exclusivity. No competing chargers on the property, and ideally none within a set radius that the landlord controls.
• Relocation and redevelopment. What happens if the landlord redevelops the parcel, and who pays to move everything.
• Easements. The utility may need its own rights to bring in new service. Get the landlord's cooperation committed in the lease, not promised on a call.
• Rent structure. Flat rent, revenue share, or a mix. Revenue share aligns everyone at an unproven site. Flat rent is cleaner at a strong one.
• End of term. Who owns the improvements, who removes them, and what "restore the premises" means for a trench full of conduit.

The value beyond the plug

For a property owner, chargers are part of the real estate, not just a business sitting on it. They can fill hotel rooms, lengthen shopping visits, attract and keep tenants, and position a property for the next decade of drivers. Sometimes the charger earns less than the building earns because of it, and that is perfectly fine.

The only rule is to decide it on purpose. If the chargers are an amenity, fund them like one and measure what they do for the property. If they are a business, hold them to a business's standard. The expensive mistake is building an amenity and expecting it to perform like an investment.

Walk the site before you sign anything

• Drive in from both directions at the busiest hour of the day.
• Stand where the charger will go and ask what a driver does for the next thirty minutes.
• Find the electrical room, then walk the route to the stalls, counting every sidewalk and curb on the way.
• Check the cell signal at the stall, not in the parking lot. The connectivity check will grade it.
• Count the competing chargers within a few miles, and note how fast they are.
• Read the lease before you fall in love with the parcel.

• Paying cash
• Public programs
• Financing
• The capital stack
• SBA checklist
• It's a real business

Paying cash, from a business that already earns it

If you run a business with steady cash flow, such as a hotel, a store or a fleet, you can fund chargers the way you'd fund any other capital improvement. It's the fastest route and the simplest paperwork. It's also the one where a bad decision costs you the most directly.

• Buy what you need, not what you want. Build what the site can realistically fill. For most first projects that means a few Level 2 ports, or one or two DC fast chargers, with spare conduit in the ground so you can grow into it.
• Get the utility's money first. Many utilities run make-ready programs that pay for the expensive electrical work, and most require approval before you build. Apply first, then build.
• Understand the tax side. The federal 30C charger credit is gone for new projects, but depreciation is generous: 100% bonus depreciation is permanent for qualifying property bought and placed in service after January 19, 2025, and Section 179 expensing is available up to $2.5 million. If there's a battery, the 48E storage credit may apply, subject to the FEOC rules. Have a CPA sort out which parts of the project qualify for what.
• Treat it like any other investment. What's the payback, what's the yield on cost, and what else could that money be doing? The site planner answers the first two.

Public programs: generous, competitive, and slow

Grants and rebates can cover most of a charging project's cost, and in some programs all of it. The federal NEVI program funds fast chargers along major highway corridors through state transportation departments. States run their own programs. California's CALeVIP is the best known: its funding window from October 7, 2026 to January 14, 2027 can cover up to 100% of eligible installation costs, up to $100,000 per fast-charging port. Utilities add make-ready incentives on top.

Almost anyone can apply. Winning is another matter.

• You have to be ready before you apply. Programs increasingly want projects that are ready to build: site control through a lease or ownership, engineered drawings, a utility service design, permits, and real cost estimates. All of that costs money, and you spend it whether you win or not.
• There's no guarantee. The funding pools are small compared with the number of applicants, and plenty of good projects miss out.
• The money often comes after you build. Many programs reimburse after construction, so you still need a way to pay for the project in the meantime.
• It comes with strings. Uptime requirements, data reporting, wage rules, and for federal money, Buy America requirements on the equipment. Price them before you celebrate.

Financing: the route most people should take

Borrowing lets you build now, keep your cash for running the business, and pay for the chargers out of the revenue they earn. Three sources are worth knowing.

• Vendor financing. Many manufacturers and distributors offer equipment loans or leases with the purchase. It's fast and convenient. Compare the rate honestly, and read what happens to the lease if the vendor leaves the market.
• Banks. Equipment loans and lines of credit, usually for borrowers with a banking relationship, a cash flow history and collateral.
• SBA loans. Loans guaranteed by the Small Business Administration, made through banks and specialist lenders. The 7(a) program lends up to $5 million for most business purposes, with simpler processing for loans up to $350,000. The 504 program funds fixed assets, typically with a bank lending about half, a certified development company about 40%, and you putting in about 10%.

Why SBA loans suit charging

Charging works best as part of an operating business: a coffee shop, hotel, gas station or car wash. That's also what SBA lenders like to see. The host business's existing cash flow supports the loan payments while the chargers ramp up, which takes much of the risk out of the deal for everyone. Getting pre-approved before you commit to a site tells you what you can afford and makes you a serious buyer when you talk to vendors.

What lenders look for, as of fall 2026

• Cash flow that covers the payments. Lenders measure debt service coverage: cash available for debt payments divided by the payments themselves. For SBA small loans the floor is 1.1, on historical or projected cash flow. For larger loans to new businesses, projections need to reach 1.15 within two years. Most lenders are happier at 1.25 or better.
• Money in the deal. New businesses, meaning those with a year or less of revenue, must put in at least 10% of the total project cost. Established businesses adding chargers are judged case by case, but skin in the game always helps.
• Personal guarantees. Owners of 20% or more generally guarantee the loan personally.
• Eligibility. The business must qualify as small, operate for profit, and meet SBA ownership and citizenship requirements.
• The rules keep moving. A revised SBA rulebook took effect on October 1, 2026 and carried most requirements forward. Ask your lender what applies on the day you apply.

Check whether an SBA loan fits

The SBA checklist

What a lender will ask for on an SBA loan for a charging project. Tick things off as you gather them. Your progress stays in this browser.

It's a real business, not a side hustle

Whichever way you pay for it, a charging site is a business. It needs someone who owns the uptime, the pricing, the maintenance, the utility relationship and the driver who calls at 10pm. It's closer to a full-time job than a passive investment, and the sites that succeed are the ones someone is paying attention to.

None of that should scare anyone off. It should just set expectations. Fund it sensibly, run it properly, and put it in the right place, and it's a good business.

• What it's for
• Site or equipment
• Sizing
• What it can earn
• 48E and FEOC
• State programs

What a battery is actually for

• Cutting demand charges. The battery covers the peaks so the meter never sees them.
• Avoiding or deferring a service upgrade. Often the real prize. A battery that lets you build on an existing 200 amp service instead of waiting eighteen months for a new one is worth more than the demand savings.
• Building where the grid cannot serve you yet. Capacity arrives years after demand does.
• Energy arbitrage on time-of-use rates, charging at night and selling by day. Real, usually small.
• Resiliency and grid programs. Nice, and sometimes paid, but rarely the reason on its own.

Site level or equipment level

The deciding question is usually not cost. It is whether the expensive part of your project is the utility service. If it is, the battery is competing against a transformer and a year of waiting, and it usually wins. If your service is already adequate, the battery is competing against a demand charge line item, and that is a much harder fight.

Sizing one, roughly

Charging demand clusters. Take your daily energy, decide what share lands in the busy window (about half in six hours is a reasonable starting point), and compare that to what the grid can supply during that window at your chosen limit. The gap is the battery's job.

Usable energy is what matters, not nameplate. So is discharge power: a 300 kWh battery that can only push 100 kW will not cover a 400 kW peak, however much energy it holds. Both numbers have to clear the bar.

Then check the recharge. The battery has to refill between busy windows at whatever power the grid limit leaves over. A site that is busy fourteen hours a day does not give a battery much time to breathe.

What a battery can actually earn

A battery at a charging site is a small power plant that happens to live next to your chargers. It can earn in several ways at once, which the industry calls stacking and everyone else calls "wait, it does what?"

• Demand charge savings. The battery covers your peaks, so the meter sees a smaller number. Usually the biggest and most reliable line.
• Capacity value. In markets like PJM, which covers Illinois' ComEd territory and New Jersey, part of your bill is set by how much power you draw during a handful of system-wide peak hours each year. Discharge during those hours and your capacity charge drops the following year. PJM's last two auctions cleared at the price cap, around $330 per MW-day, so the hours that set that tag are worth real money.
• Demand response and virtual power plants. Utilities and aggregators pay you to discharge when the grid is stressed. The payment is usually per kW you promise to deliver, per year or per season.
• Energy arbitrage. Charge when power is cheap, use it when it's expensive. Real on time-of-use rates, usually the smallest slice.
• Ancillary services. Frequency regulation and reserves pay well, but mostly to larger batteries on the grid side of the meter. For a charging site, treat this as upside, not a plan.
• Resilience. Keeping the lights on during an outage has value that doesn't show up on a spreadsheet, and the chargers only stay up if the battery's inverter is built to run the site as an island.

The federal tax credit: 48E, and the FEOC rules that can zero it

With the 30C charger credit gone for new projects, the storage credit is often the only federal money left at a charging site. It is also the one most likely to disappear in diligence, so it's worth understanding properly.

What 48E pays

• A standalone battery qualifies for the section 48E investment tax credit: 30% of the eligible cost, if the project pays prevailing wages and uses registered apprentices, or is under one megawatt. Otherwise the base is 6%.
• Bonus credits can add 10 points for domestic content and 10 points for siting in an energy community.
• Storage kept its credit through the 2025 federal budget law, with full value for projects that begin construction through 2033 before it phases down.
• Credits can be sold for cash to a tax-paying buyer, which matters if you don't have the tax appetite yourself. Buyers will do diligence on everything below.

The FEOC rules, in plain English

The same 2025 law added restrictions on "prohibited foreign entities," which in practice means companies owned, controlled or heavily influenced by China, Russia, Iran or North Korea. There are three tests, and failing any of them means no credit at all.

1. Who owns the project. The taxpayer claiming the credit can't be a prohibited foreign entity itself.
2. Who controls it. The owner can't be "foreign-influenced," which includes certain licensing and control arrangements. For 48E, making those payments within ten years after the battery goes into service can claw back the entire credit.
3. Where the parts came from. This is the one that bites. The project calculates a material assistance cost ratio: the share of its equipment cost that did not come from prohibited entities. It has to clear a minimum that rises every year.

Battery cells are roughly half of the equipment cost in the IRS safe harbor tables. That means cells from a prohibited manufacturer make the threshold very hard to reach, however American the rest of the system is. A system assembled in the US around prohibited cells is still, for this purpose, mostly prohibited.

What to actually do

• Ask every supplier for written certification of where the cells, modules and major components were made, and by whom. Build the paper trail before you buy, not before you file.
• Treat cell origin as the first procurement question, not the last. It decides the math on its own.
• If you plan to sell the credit, talk to buyers early. They will set the documentation standard, whatever the law technically requires.
• Watch for guidance. Treasury issued interim rules in early 2026 (Notice 2026-15) and has promised proposed regulations and updated safe harbor tables. Details will move.
• Get a tax advisor who has done this. This page is a map, not a tax opinion.

State programs worth knowing

State money changes faster than anything else on this site, so treat these as a starting point and check the current rules before you model them. As of fall 2026:

Illinois

• The Clean and Reliable Grid Affordability Act took effect June 1, 2026. It targets 3 GW of storage, adds a rebate for standalone batteries, and requires the utilities to run virtual power plant programs that pay batteries to discharge.
• ComEd and Ameren already offer a distributed generation rebate of up to $300 per kWh for storage paired with solar. ComEd's tariffs for the new standalone rebate and its VPP program were filed and are awaiting approval.
• ComEd territory sits in PJM, so the capacity value described above applies on top.
• For larger projects, the Illinois Power Agency runs long-term procurements for utility-scale storage.

New Jersey

• The Garden State Energy Storage Program targets 2,000 MW of storage by 2030. Phase 1 covered large grid-scale projects, with the first awards in March 2026.
• Phase 2 covers smaller batteries on the distribution grid and behind the meter, with a mix of upfront and performance incentives. It is still being designed.
• An August 2026 straw proposal for small behind-the-meter batteries floated up to $200 per kW per year for ten years in exchange for discharging during utility-called events. Commercial terms may differ. Watch the Board of Public Utilities.
• New Jersey is also in PJM, so capacity value applies.

Elsewhere

California, Massachusetts, Connecticut, New York and others run their own storage incentives and demand response programs, and they vary widely in size and rules. The fastest way to see what applies at an address is the DSIRE database, then a call to the utility's program team.

The things people forget

• Cycles and warranty. Daily cycling is fine, twice-daily cycling on a warranty written for one cycle a day is not. Read the throughput terms, not the years.
• Degradation. Size for the capacity you need in year eight, or budget for augmentation.
• Round trip losses. You buy roughly 12% more energy than you sell through the battery. Small, but it belongs in the model.
• Fire code, setbacks, and permitting. Storage brings its own inspection path, and it is not the charger's.
• Interconnection. Storage that can export triggers a different, slower utility review than storage that cannot. Non-export designs get through faster.
• Tax credits. The FEOC rules decide whether you get 30% or more back, or nothing. See 48E and FEOC above.

When a battery does not pay

• Your tariff has no demand charges, or has an EV rate that already suppresses them.
• Utilization is so low that the demand charge, while brutal per kWh, is small in absolute dollars against the battery's cost.
• You need a service upgrade for daily energy anyway, so the battery saves you nothing on the utility side.
• The site is a candidate for fewer or smaller chargers instead. That fix is free.

The site planner has all of this built in. Set a grid limit, add a battery at the site or inside the chargers, and it will tell you what the demand savings are worth and whether they cover the battery.

Open the site planner

• What depends on it
• The options
• Tier 1 IoT providers
• When it drops
• Check a site

What depends on the network

Almost everything a driver experiences as "the charger" is actually a conversation between the charger and a server somewhere else. When the connection is weak, the charger doesn't fail cleanly. It fails intermittently, which is worse, because it looks like a hardware problem and gets diagnosed like one.

• Authorization. Every app start, RFID tap, and Plug and Charge handshake asks the back office for permission. No connection, no answer, no session.
• Payments. The card terminal sends a pre-authorization to the processor before the session and a capture after it. A drop in between leaves you with a session you delivered and may not be able to bill.
• Uptime and monitoring. The software only knows what the charger tells it. An unreachable charger looks down even while it charges cars, and a charger that is online but failing every session can look fine. Weak connectivity makes your uptime numbers wrong in both directions.
• Remote fixes. Most faults can be cleared with a remote reset or a configuration change, if you can reach the charger. Every unreachable fault becomes a truck roll.
• Firmware updates. Updates are large, and large downloads over weak links fail. An interrupted update can leave a charger needing a site visit to recover.
• Pricing and configuration. Price changes, idle fees, and access rules only take effect when they arrive.
• Load management. Commands that cap site power to keep demand charges down are only as good as the link that carries them. If the command doesn't arrive, the peak does.
• Roaming. Sessions started from another network's app cross yet another connection before the charger ever hears about them.

The options

Buy for the network's future, not today's

US carriers shut down their 3G networks in 2022, and chargers with 3G modems went dark until the modems were replaced. Ask what cellular technology the modem uses, how long carriers are expected to support it, and whether the modem is a swappable part or a board replacement.

A modem in every cabinet, or one router for the site

A modem inside each charger means independent failures, but also a small antenna inside a metal box. One industrial router with an external antenna and two SIMs, feeding every charger over Ethernet, usually gets a much better signal. It is also a single point of failure, so buy one with dual modems or dual SIMs, and put it on the same maintenance plan as the chargers.

The payment terminal may have its own modem

Many payment terminals connect separately from the charger. That is two modems, two SIMs, and two ways to fail. A charger can be online while its terminal is not, and the driver will blame the charger either way. Check the terminal's signal too.

Why a tier 1 IoT provider matters

A SIM from a phone store is built for a person who moves around and tolerates the odd dropped call. A charger sits in one spot for ten years and needs to answer every time. The difference shows up at exactly the wrong moments.

What tier 1 means

• Direct relationships with the major carriers, or the carrier itself.
• Multi-carrier SIMs that switch automatically when the primary carrier degrades.
• Private network options (a private APN and static IPs), so your chargers aren't sharing the public internet path with everyone's phones.
• A management platform showing usage, connection status, and alerts for every SIM.
• Pooled data across your fleet, so one chatty charger doesn't hit an overage cliff.
• Industrial SIMs, often soldered to the board, rated for heat, cold, and vibration.
• Support that picks up and can see your SIMs when you call.

What goes wrong with the cheap option

• Deprioritization. Consumer and low-cost plans get pushed down the queue when a tower is busy, which at a gas station is exactly when drivers want to charge.
• SIMs deactivated for inactivity or unusual usage patterns, with nobody watching when it happens.
• Carrier network address translation that silently closes idle connections. Chargers then drop off, reconnect, and drop again, which looks like a flaky charger.
• No visibility. When a site goes quiet, you can't tell if it's the charger, the SIM, or the carrier.

Examples, not endorsements. Carriers: AT&T, T-Mobile, Verizon, all of which run IoT divisions. IoT connectivity providers: 1NCE, Hologram, KORE, Soracom. Industrial routers: Cradlepoint, Digi, Peplink, Sierra Wireless, Teltonika.

Questions for your IoT provider

Which carriers does the SIM use, and does it fail over automatically?

Listen forAt least two major carriers, with automatic switching based on signal and service.

Do you offer a private APN and static IPs?

Listen forYes, and a straight answer on how it helps keep chargers reachable.

What happens when a SIM hits its data limit or goes quiet?

Listen forPooled data, alerts before anything is cut off, and no silent deactivations.

Is the modem certified on the carriers you use?

Listen forCarrier certification for the exact modem in your charger. Uncertified modems can be refused by the network.

What can I see, and what can you see, when a charger goes offline?

Listen forPer-SIM connection history and signal readings, available to you directly.

When the network drops anyway

It will. Decide in advance what the site does, because the default is usually "whatever the firmware happened to do."

• Local authorization. Chargers can keep a list of known RFID cards and members so regulars can still start sessions offline.
• Store and forward. OCPP chargers can queue session records and send them when the link returns. Confirm yours does, and that your software accepts late messages without mangling the bill.
• Card payments. Most terminals need to reach the processor to authorize. Choose between refusing sessions and allowing limited free or deferred sessions, and know what each costs.
• Keep-alive settings. Heartbeat and connection keep-alive intervals need tuning for your network, or carrier equipment will quietly close idle connections.
• Monitor the link itself. Track reconnections per day and signal readings over time, not just online or offline. A charger reconnecting forty times a day is telling you something before it fails.

• Ground rules
• For site hosts
• For operators
• Questions by vendor
• When a vendor leaves
• Walk-away signs

Ground rules for everyone

Uptime guarantees are theater

Whether a charger works for a driver depends on the car, the cellular network, the payment processor, the back office, the utility, the weather, the last person who ran over the cable, and the vendor. No vendor controls that list. The ones who guarantee an uptime number define it until it means nothing: network outages excluded, "customer-caused" faults excluded, scheduled maintenance excluded, and a stall that is online but fails every session counted as up.

If the pitch leads with uptime, walk away. Ask about service and support instead, and make them walk you through their internal timelines, what triggers a dispatch, and who does the work.

Hold vendors to what they actually control

• Response time. How fast a qualified person starts working the problem, remotely or on site.
• Time to repair. How fast the stall is charging cars again, with the clock starting when the fault is reported, not when they get around to diagnosing it.
• Spare parts on hand. Which parts are stocked, where, and how many. A two-day repair becomes a two-month repair when the power module ships from overseas.

A tight service SLA built on those three numbers is worth more than any uptime guarantee, because it is something a vendor can actually be held to.

Price it all in

The quote should include everything the charger needs to take a payment and charge a car: payment terminal, cable length, connectivity, commissioning, and the first term of software. If the POS is an add-on, the longer cable is an upgrade, and the modem is a line item, the real price is higher than the headline and the vendor knows it. All-inclusive pricing, or keep shopping.

Payment terms should be tight and fair

• A modest deposit to reserve production, not most of the order up front.
• The balance at net 30 after delivery.
• Shipped DDP, delivered duty paid. The vendor carries freight, customs clearance, and import duties to your site. On hardware built overseas, the gap between DDP and anything less is tariff exposure you did not price.

Prefer in-house service

A vendor whose own technicians service its own equipment has every incentive to fix it fast and learn from the failure. A vendor that subcontracts service has a margin to protect and a subcontractor with ten other customers. Ask who shows up. Get a name, not a network.

Speed is your leverage

Vendors bend for a buyer who can sign this quarter. They do not bend for tire-kickers, and they remember who wasted their time. Don't negotiate hard during ideation. Get indicative pricing, do your homework, and start procurement when you are actually ready to move. That is when a vendor will give up terms, because a real order is on the table. This goes for hardware and software alike.

If you are a site host buying turnkey

Turnkey means one contract and one phone number, which is the appeal. It also means five companies hiding behind one logo: whoever builds the hardware, runs the software, processes payments, installs the site, and fixes it. You are only as protected as the weakest of the five.

• Make them name every subcontractor, by layer. If the answer is vague, so is the accountability.
• Find out who answers when a driver calls, and what you are expected to do while you wait.
• Get the service SLA in writing, with the three numbers above and a remedy when they're missed.
• Ask what happens to your chargers if the provider exits, sells, or you want to leave. Can the hardware move to another network, and who pays for that?
• Know who owns the chargers, the data, and the driver relationship. Those can be three different answers.
• Read the renewal terms. A cheap first year with an expensive year four is a lease with extra steps.

If you are an operator buying the pieces

Buying hardware, software, payments, installation, and service separately gets you better pricing and more control. It also makes you the integrator, which means the gaps between vendors are now your problem.

• Get the hardware and software vendors on the same call before you buy either, and have them describe the full session together, from plug-in to payment capture.
• Confirm the OCPP version and the specific features each side actually supports. "OCPP compliant" without details means nothing.
• Decide who owns each failure mode before it happens. A charger that won't start could be hardware, software, payments, or connectivity, and each vendor will suggest it's one of the others.
• Test the ugly paths on purpose before launch: pull the antenna, use a declined card, stop mid-session, unplug during precharge.

Questions to ask, by vendor

Examples of vendors in each category are listed to show what the category covers. Inclusion is not an endorsement.

Hardware

Examples: ABB E-mobility, Alpitronic, Autel, BTC Power, Kempower.

Who services your equipment, and are they your employees?

Listen forNamed in-house technicians in your region, or a specific, accountable partner with its own SLA to the vendor.

What is your response time and time to repair, in writing?

Listen forHours and days with remedies attached. Not "best efforts."

Which spare parts do you stock in North America, and where?

Listen forA specific list: power modules, cables, connectors, payment readers, controllers, contactors. Stocked domestically.

Is the price all-inclusive?

Listen forPayment terminal, cable, connectivity, commissioning, and first-term software all included.

What are your payment and shipping terms?

Listen forModest deposit, net 30, DDP.

Will you provide service manuals and parts access to a qualified third party?

Listen forYes, at least for basic repairs, and in writing. You'll need it if they ever leave.

What does the warranty exclude?

Listen forA short list. Cables and connectors excluded as "wear items" is common and expensive.

Charge management software (CMS)

Examples: AMPECO, Driivz, Monta.

Which OCPP version and features have you certified or tested with this hardware?

Listen forNamed hardware models, a version number, and a list of supported features. Ideally, live sites running the same combination.

What is the support model?

Listen forA clear tier 2 role for you, the operator, with defined response times. Be wary of anyone who says they will "handle driver support" as a headline feature.

How do I move my chargers to another platform?

Listen forA documented process to change the OCPP backend address and credentials, and no fees or lockouts for doing it.

Who owns the session data, and can I export all of it?

Listen forYou own it, and there's a bulk export in a standard format.

What happens to my sites if you are acquired or shut down?

Listen forNotice periods, data export, and help migrating. Not a shrug.

What does pricing look like in year three?

Listen forPer-port pricing with a cap on increases.

Payments

Examples: Nayax and other unattended payment terminal providers.

Which processors is this terminal certified with?

Listen forMore than one. Changing processors should not mean changing terminals.

How large is the pre-authorization hold, and can I set it?

Listen forConfigurable. Drivers call about a $100 hold far more than a $25 one.

What does the terminal do when connectivity drops?

Listen forA defined offline policy you choose, not one you discover.

Who handles refunds and chargebacks, and how fast?

Listen forA process your tier 1 team can actually use.

Installer or EPC

How many charging sites have you built, and can I call two of the owners?

Listen forReal references at similar sites. General electrical experience is not the same thing.

Who manages the utility, and what is your assumption for their timeline?

Listen forSomeone who has worked with your specific utility and quotes the schedule in months, not weeks.

What is excluded from the price?

Listen forA short exclusions list. Trenching distance, soil conditions, and utility work are where change orders come from.

Who commissions the chargers, and who signs off that they work?

Listen forA defined handoff with the hardware vendor, ending in real test sessions on real cars.

Service and maintenance

What triggers a dispatch, and how fast after that?

Listen forSpecific triggers, like a failed-session threshold or a fault code, and a clock that starts automatically.

Do your technicians carry parts, or order them after diagnosis?

Listen forCommon parts on the truck. The second visit is where time to repair goes to die.

Is preventive maintenance included, and what does it cover?

Listen forScheduled visits with a checklist: cables, connectors, filters, cooling, firmware.

How do you report?

Listen forPer-ticket records with times and root cause, not a monthly uptime percentage.

When a vendor leaves the market

It isn't hypothetical. In October 2024, Enel X Way closed its North American business on a few days' notice and said its software would be discontinued, warning that commercial stations would lose functionality without software continuity. Owners of networked commercial chargers were left scrambling, and it was independent OCPP software providers that ended up migrating sites so they could keep running.

Earlier that year, DC charger maker Tritium entered voluntary administration and its business was sold to Exicom. The new owner committed to supporting existing warranties, while acknowledging it wasn't necessarily obliged to. That's the lesson: a warranty is a promise from a company, and it's only as good as the company.

Negotiate these before you need them

• Access to spare parts, directly or through a named distributor, if the vendor exits or stops supporting the model.
• Service manuals for basic repairs: cables, connectors, payment readers, filters, and module swaps.
• Documented steps to change the OCPP backend address and credentials, so you can point your chargers at a new platform without the old one's help.
• Your data, exportable, on demand.

Then test the OCPP changeover on one charger before you need it. A migration plan you have never run is a theory.

Walk-away signs

• The pitch leads with uptime.
• They can't tell you who shows up to fix the charger, or it's "our partner network."
• The payment terminal, the longer cable, or connectivity is an add-on.
• Most of the order is due up front, or they won't ship DDP.
• "OCPP compliant," with no version, no feature list, and no named hardware it has run on.
• No answer on moving to another software platform, or a fee for doing it.
• The software vendor promises to take care of your drivers.
• References are all pilots or all from one friendly customer.
• Warranty exclusions longer than the warranty.

• The metric
• Know your vehicles
• The 80% rule
• Power and windows
• Reliability
• Operations

The metric that matters is throughput

For a public site, the question is how many strangers show up. For a fleet, you already know. The questions are how many vehicles you can charge in a day, whether each one is ready on time, and what each charge costs. Utilization is an output, not the goal.

A fleet charger sitting idle for an hour because nobody moved the car is a staffing problem, not a demand problem. That distinction runs through everything below.

Know your vehicles

• Every model has its own charging curve. The peak number on the spec sheet lasts for a short stretch in the middle of the battery, and only if the pack is warm. The battery section explains why.
• The car's onboard charger caps Level 2. An 11 kW car on a 19.2 kW charger is an 11 kW car.
• 400-volt and 800-volt vehicles behave differently on the same charger, and some chargers can't give both full power.
• In a mixed fleet, plan around the slowest vehicle you charge often, not the fastest one in the brochure.
• Software updates change charging curves. What you measured last year may not be what you get today.

The 80% rule

Charging slows sharply as the battery fills. On a typical fast-charging curve, going from 80% to 100% can take as long as going from 20% to 80%. For one driver, that's an annoyance. For a fleet, it's a whole vehicle you didn't charge.

• Set a charge-to target by use. Most rental returns and dealer service turns only need 70% to 80%.
• Stop the session at the target automatically, through the charger, the software, or the vehicle's own charge limit, then move the car.
• Take cars to 100% only when the next trip genuinely needs it, and do it where time is cheap: overnight, on Level 2.

Power on site, and the charge window

• Power. Total site power decides how many chargers can run flat out at once. Shared power slows everyone down when the lot is busy, and managed charging keeps demand charges in check.
• The window. This is when cars are actually available to charge. A depot might have eight to ten hours overnight. A rental return lot sees a steady stream all day. A dealer service lot has whatever time falls between drop-off and pickup.
• Match the charger to the window. If a car sits for ten hours, Level 2 will finish it cheaply. If it has to go back out in an hour, only DC fast will get it there.

Reliability, success rate and compatibility

Fleets feel failures more than anyone. A failed public session loses one sale. A failed fleet session means a car isn't ready, a porter made a wasted trip, and the whole schedule slips.

• Track success rate by vehicle model, not just by charger. Compatibility problems are usually specific to one model on one software version.
• Test every model in your fleet on the exact hardware and software you plan to buy, before you buy it. Bring the cars to the vendor, or the vendor to the cars.
• Control firmware updates. An update that fixes one model can break another, and a fleet finds out at 6am.
• Budget for retries. A failed start costs a swap and someone's time, and the throughput tool counts it.
• Don't forget the network. A charger that can't reach its software can't start a session. See Connectivity.

Operations, where fleets are won

• Staff the moves. The people swapping cars set the ceiling as often as the chargers do. Count swaps per hour, not just chargers.
• Design the lot for flow. A queue lane, pull-through stalls, and cables that reach every model's charge port.
• Charge in priority order. The car leaving soonest charges first, and only to what its next trip needs.
• Use telematics. Know each car's charge level before it arrives, and stop sessions at target without anyone walking out to do it.
• Schedule around the rate. Shift what you can into cheap hours and cap site power at peak.
• Measure what matters. Vehicles ready on time, vehicles charged per port per day, success rate by model, and cost per charge.

Model your fleet

How the forecast works

Traffic-driven sites (highway, retail) start from the road. EVs passing per day equal traffic times EV share. Sessions equal EVs passing times a capture rate, adjusted for site quality and for competing ports nearby (each competitor cuts demand by roughly 13%, a factor of 1 divided by 1 plus 0.15 per competitor).

Destination sites (hotel, workplace, multifamily) start from the people already there. Sessions equal population times the share present on a given day, times vehicles per person, times EV share, times the share of those EVs that charge that day.

Both are capped at practical capacity: ports times operating minutes divided by session length, times a maximum utilization. Past that cap, drivers queue or leave, and the forecast says so.

EV share grows at a constant annual rate, capped at 60% of vehicles. That cap is a safety rail, not a prediction.

How the site planner works

• Capex is hardware, plus make-ready (per kW for DC, per port for Level 2), plus any utility upgrade, plus design and permitting as a percentage (with a minimum), plus contingency. Grants and make-ready rebates come off the top.
• Grid energy equals delivered energy divided by charger efficiency. You pay for the losses, the driver doesn't.
• Billed demand equals installed kW times a billed demand percentage, every month, all year. Real peaks depend on how sessions overlap. At low utilization, assume they overlap more than you'd like.
• Sessions grow at a flat annual rate, capped at 24 per port per day for DC and 8 for Level 2.
• Results are in today's dollars, pre-tax, unlevered, with no depreciation, no hardware replacement, and no residual value. Chargers that last ten years exist. Plan as if yours are not among them.
• Break-even sessions scale DC and Level 2 sessions together until net present value hits zero.

Live data, and why it may not work here

The forecast can pull two things automatically: traffic counts from state DOT ArcGIS map services, and public port counts within a radius from the federal Alternative Fuels Data Center. Both are open, both are free, and one needs a key you can get in a minute at developer.nlr.gov/signup.

The hosted copy of this page runs under a content policy that blocks requests to other sites, so the lookups will report that they were blocked. Download the page and serve it yourself, or open the file locally, and they work. That is a hosting limit, not a data limit.

The traffic lookup queries an ArcGIS feature layer for count sites within your radius, reads whichever field carries volume, and takes the median. Field names differ by state, which is why it matches on the name rather than assuming one. If your state is not in the list, paste its layer URL: most DOTs publish one, and searching your state plus "AADT ArcGIS REST" finds it. The built-in list is a convenience, not a guarantee. Verify against the source map.

Counts are per segment and sometimes per direction. A median of nearby readings is a starting point, not the traffic passing your driveway.

How the battery math works

• Daily grid energy is split into a busy window (hours and share of daily energy, both inputs). Everything above what the grid limit can supply in that window is the battery's job.
• The battery covers up to the lesser of its usable energy and its power times the window length. Whatever it cannot cover is spread across the window and added back to billed demand.
• Billed demand becomes the grid limit plus any shortfall, and never more than the site would have drawn without a battery.
• Round trip losses are charged as extra energy on everything that passes through the battery.
• Battery cost is per kWh and per kW for a site battery, or a premium per charger for battery-integrated units. Replacement, if you set a year, lands as a cash outflow in that year.
• The result is compared against the identical site with no battery, so the NPV difference is what the battery is worth on its own.
• What this does not model: recharge scheduling between windows, time-of-use arbitrage, degradation curves, grid services revenue, and the utility interconnection review that storage triggers. All four matter. None are one-liners.

Running it offline with a local data file

The repository includes build-data.mjs, a dependency-free Node script that downloads open data and writes compact JSON into a data folder next to the page. Serve the site and the page loads whatever is listed in data/manifest.json at startup. Or load the files by hand in the forecast, which works even on a hosted copy that blocks outside requests, because reading a file you chose is not a network call.

Once a file is loaded, lookups run in your browser: nearest or median traffic counts, and port counts inside your radius. No API key, no network, no rate limits, and the page remembers the data between visits.

The formats are deliberately boring, so you can generate them from anything:

• Traffic: type "aadt", rows of latitude, longitude, and daily volume.
• Stations: type "stations", rows of latitude, longitude, DC fast port count, and Level 2 port count.

What about PlugShare

PlugShare has the best crowd-sourced charging data in North America and no open API. Access is a commercial license through Recargo, and their terms do not permit scraping or redistributing the data. A site that tells people to respect their vendor contracts should not ship a scraper, so this one does not. If you have a PlugShare license, the local file format above is three columns wide and you can fill it yourself.

The open substitutes are the Alternative Fuels Data Center, which is federal data and free to use, and Open Charge Map, which is community data under a share-alike license. Neither has PlugShare's check-in history, which is the part you would actually have wanted.

Utility rates

The planner can pull current commercial tariffs from the US Utility Rate Database, which is maintained on behalf of the Department of Energy and served through OpenEI. It lists EV-specific schedules first, averages time-of-use energy prices into one number, and takes the highest first-tier demand charge. That simplification is fine for screening and wrong for a final model. Real tariffs have seasons, tiers, ratchets and minimums, and the schedule you land on depends on your service size, which the utility decides.

Where the benchmarks come from

Session-level logs from operating sites, published with permission and stripped of anything that identifies a site, host, operator or driver. Figures are trailing twelve months where a site has that history and lifetime otherwise. A session counts when it delivered at least half a kWh. Failed sessions are attempts that delivered nothing.

How the newer benchmark data was cleaned

• Some session exports had their dates partly converted by spreadsheet software, which read day-first dates as month-first whenever the day was 12 or under. Those dates were swapped back before any analysis. Without the fix, some sites appeared to have more than a year of history when they had eleven weeks.
• Sessions per port count each connector. Sessions per charger count each physical unit, which is the fairer comparison because the dual-connector units in this data almost always charged one car at a time.
• "Drivers who left with nothing" counts failed attempts that were not followed by a successful session at the same site within fifteen minutes, with clusters of failures counted once. It's a better measure of lost customers than the raw failure rate, which counts every retry.
• Figures are for the most recent twelve months where a site has that much history, and for the full period otherwise.

Where to get real inputs

• Traffic counts: your state DOT's traffic count map (search your state name plus "traffic counts"), or the FHWA Highway Performance Monitoring System.
• EV registrations and adoption: AFDC maps and data and Atlas EV Hub.
• Competing chargers: the AFDC station locator.
• Incentives: AFDC laws and incentives, DSIRE, and the 30C credit entry for the federal status.
• Electricity rates: your utility's commercial tariff book, and ask specifically about EV rate schedules.
• Costs: real quotes. Hardware from at least two vendors, make-ready from a local electrical contractor who has done a charging site before, and utility upgrade costs from the utility itself.

Limits

This is a ballpark tool for learning and early screening. It is not an engineering study, a traffic study, an interconnection estimate, or investment, legal, or tax advice. Use it to decide which sites deserve a real study, not to skip the study.

License

Written content is released under Creative Commons Attribution 4.0. Code is released under the MIT license. Copy it, fork it, fix it, and credit the source. Corrections to assumptions are the most useful contribution you can make.

Data credits: Alternative Fuels Data Center, US Department of Energy. Open Charge Map contributors, CC BY-SA 4.0. State departments of transportation for traffic counts. If you publish a copy that ships their data, carry their credits with it.

Last updated October 2026.