Site Selection for EV Charging Stations, With Data

Station counts say nothing about demand. How utilization, power mix, network concentration and peak shape reveal the real capacity gap.

Where the stations are is public information. Which of them are full is not.

Most site selection decisions ignore that asymmetry: look at a map, count the stations in an area, build where there are fewest. The problem with the method is that a station count says nothing about demand. A district with forty stations sitting idle half the day is a worse place to invest than one with eight that all fill up in the evening.

Density misleads, utilization does not

Competition is measured in used capacity, not installed capacity. Where socket count is high, two things can be true at once: the market is saturated, or the market grew and supply has not caught up. The count does not separate them. Utilization does.

The value to look at is average peak-hour utilization across the sockets in the area. Below roughly forty percent, new capacity divides existing demand. Pressing above eighty and flattening at the peak, there is unserved demand and it can be quantified. The band in between deserves the most care: read the direction of travel rather than the level.

For any of this to be usable, utilization has to have been measured consistently. Whether it was computed per socket or per station, and whether fault time was stripped out, changes the answer directly; we covered how utilization is measured separately.

The unit of analysis is the catchment, not the district

Administrative boundaries do not describe driving behaviour. A station at the far end of a district is not a practical alternative even when it looks close on a map, and one on the wrong side of a river is no alternative at all.

The right unit is the catchment: the set of stations reachable within a given drive time. Five to seven minutes in a city, and the spacing between exits on a highway, are reasonable starting points. Competition is defined inside that set and utilization is aggregated over it. Run the same analysis by district and stations that never compete land in one pool, making the area look saturated when it is not.

Four signals worth reading

The utilization profile. The daily and weekly utilization curve across the catchment's sockets. The curve, not the average — because the decision is driven by the peak.

Power mix. If an area has twenty AC sockets and a single DC one, the gap there is not in socket count but in power type. A profile where DC is constantly full and AC constantly empty states directly what the next investment should be.

Network concentration. If every socket in the catchment belongs to one operator, there is no price competition there and users are locked into that network's app. For a second entrant that is an opening — particularly where that network's tariff runs above neighbouring areas. We covered how to read tariff differences between networks separately.

The shape of the peak. Fifty percent utilization all day and one hundred percent during a two-hour peak produce the same average and call for different investments. The first is a new-location problem; the second is a socket-count problem at an existing one.

Calculating the capacity gap

Those four signals reduce to one number. Put the catchment's total socket-hours in the denominator and observed occupied socket-hours in the numerator; the peak-hour portion of the difference gives an upper bound on unserved demand.

It has to be called an upper bound, because utilization data never shows turned-away demand. A driver who arrives at a full station and leaves appears in no series. In a catchment that saturates at peak, real demand always sits above the utilization you measured, so estimates there stay conservative and the error usually runs in your favour.

The inverse also holds: in an empty area, utilization reflects real demand exactly, because nobody is being turned away. The data is cautious where saturated and precise where empty — and a decision should account for that asymmetry.

After the investment: measure your own site too

The most commonly skipped part of site selection is what happens after opening. When a new station goes live two things happen at once: its own utilization settles, and the utilization of everything around it changes. Without measuring the second, you cannot tell whether your site's performance came from market growth or from your neighbour's share.

Watching the same catchment before and after opening makes that distinction. Surrounding utilization falling while yours rises is displacement; both rising together is growth. That distinction is what gets the second and third site of a chain right.

What the dataset looks like

What this analysis needs is not complicated, but it has to be continuous: timestamped status per socket, plus location, power, connector type, operator, and tariff per channel. Collected at a fixed cadence, catchment calculations, utilization curves and the capacity gap all derive from that.

The hard part is not the schema, it is the continuity. An investment case usually looks at a window of a few months, and the gaps in that window land exactly where you are trying to read seasonality. Which is why collection freshness has to be a commitment rather than a feature — the same argument as in data freshness is an SLA.

Conclusion

Empty on a map is not the same as demand. Site selection becomes defensible when catchment-level utilization, power mix, network concentration and peak shape are read together; decided on station counts, it is largely luck.

Our EV charging stations usage data is the layer that feeds those analyses: fifteen-minute sampling, socket-level status, and tariffs per network. Where to build cannot be decided without measuring what is already there.