What real charging sites actually do
Every number below comes from session-level data at operating US charging sites: tens of thousands of sessions across a dozen sites and a small-town network, over several years. Names, locations, hosts and operators are removed and details are generalized. The numbers are real.
This is what actually happened, not what we think will happen. Every figure below comes from real charging sessions at real operating sites. Nothing is modeled, projected or rounded up to look good.
The data has been anonymized: names, addresses, hosts, operators and drivers are removed, and details like location are generalized to a region. It has also been cleaned: scrambled dates fixed, quick retries merged so a driver trying twice counts once, and the most recent twelve months used wherever a site has that much history. How the data was cleaned.
Real doesn't mean guaranteed. These sites are what happened at their addresses, with their traffic, prices and electricity rates. Yours will differ. Your mileage may vary, and at a charging site that's meant literally.
What this data says
- The spread is 40 to 1. The best site does 5.2 sessions per port per day. The worst does 0.12. Same equipment, same country, same years. Nothing else in this business has that much leverage, which is why the forecast matters more than the hardware quote.
- Free does not create demand. The travel plaza charges drivers nothing, has plenty of ports on a highway, and still runs about one session per port per day. If free cannot fill the stalls, price was never the problem.
- The winner is a convenience store near an airport. Rideshare drivers charge every day, and this store sits right where they wait between trips. A steady stream of high-mileage drivers beats a busy road, and it beats demographics every time.
- Cars arrive around 30% and leave around 85%. That held across every DC site here, which is why 40 kWh is a better planning number for a session than anything derived from pack size.
- Busy sites flatten. The busiest six hours carry 34% of energy at the best site and 55% at the quietest one. Quiet sites are spiky, busy sites are smooth, and storage sizing should use the site you have rather than the one you want.
- Utilization is lower than anyone expects. The best site here occupies its ports 23% of the time. That is a good site. Most are under 5%.
- Failed sessions are a revenue line. Maintained sites came in at 1% to 6%. The small-town fleet ran 17%, with individual sites near 30%. A 17% failure rate is 17% fewer sessions than your forecast, and drivers who fail once tend not to come back.
- Uptime and success are different numbers. The small-town fleet reported uptime in the high 90s while roughly one in five attempts failed. Uptime says the charger answered the network. Success says a car left with electricity in it. Only one of those is revenue.
- Slow DC ages badly. The garage site has declined year over year as faster options opened nearby. Charging speed is competitive, not absolute.
- Scarcity is real estate too. The only fast charger in a remote small town runs 2.75 sessions per charger per day. The high-end shopping center runs 2.81. Being the only option for miles works as well as being in the best neighborhood in town. More on the real estate page.
- Great real estate carries a premium price. The shopping center charges $1.00 per kWh, with strong idle fees, and is still the busiest per charger in the newer data. The remote town charges about $0.58, below what the site planner says it needs to break even, while having no competition at all. Pricing power is part of what a location is worth.
- Destination sites follow their destination. The ski resort site ran about eight times more sessions in January than in April. Underwrite the whole year, not the best month.
- Most failed attempts are retries. Raw logs in the newer data show 24% to 38% of attempts failing, but many are followed by a successful retry within fifteen minutes. The drivers who actually left with nothing ranged from 6% to 23%. Measure the second number, because that's the one that costs you customers.
- More power isn't more business. The ski town and the corridor restaurants each run two 400 kW chargers, 800 kW per site. The small-town site that keeps pace with the big-city shopping center has a single 60 kW charger. Power beyond what the traffic needs shows up as hardware cost and demand charges, not sessions. Size for the drivers you have, and leave conduit for the ones you expect.
- Two connectors rarely means two cars. On the newer dual-connector chargers, both connectors were busy at once only 0% to 9% of the time. Plan and compare per charger, not per connector.
How to use these
Load any row into the tools with the button on it. That fills in ports, power, sessions per port per day, energy per session, and price, so you can run the pro forma against a site that actually exists instead of a number you hoped for.
The honest caveat: a dozen sites and one network is a sample, not a census. It now spans the South, Midwest, Northeast and West, and it spans everything from 30 kW chargers to 400 kW chargers. The corridor restaurants have only about eleven weeks of data. Treat these as reference points for what plausible looks like, not as the distribution of American charging.
If you operate sites and want to add anonymized data to this set, that is the contribution that would improve this site most.