Start with the full opportunity.
We evaluate locations across a partner’s portfolio and the wider charging landscape.
Every site earns investment.
Over 850M Data Points.
Finding a place to build is easy. Finding a site worth building takes evidence, a rigorous forecast, and the discipline to pass on locations that do not make the grade.
Data science narrows the field.Diligence makes the case.
Each pass narrows the field. Our data science informs the search; site diligence and investment review determine what moves forward.
Pressure-test demand.
Forecasts, historical charging evidence, and uncertainty checks remove weak candidates from consideration.
Examine what can be built.
Site design, utility, access, and commercial review further narrow the field.
Commit to the strongest sites.
The best options advance to a shared partner and investment decision.
Portfolio opportunityIllustrative
Candidate fieldBroad search
The average charging site is not our target.
Charging demand varies sharply from property to property. Network and market averages blend strong locations with weak ones; they cannot tell us which site merits a capital commitment.
That is why we forecast individual locations, evaluate the range of possible outcomes, and bring only the strongest cases into site diligence.
Same market.
Different outcomes.
Different outcomes.
An illustrative distribution of candidate locations. We invest in a site case, not a market mean.
Data science with an investment standard.
Our models are built to sharpen judgment, then withstand challenge from the people who have to build and operate the site.
Signal over noise.
Feature selection and correlation clustering keep overlapping signals from distorting the forecast.
Test outside the training set.
Held-out historical sites reveal how predictions compare with outcomes the model has not seen.
Quantify uncertainty.
Forecast ranges and bias checks show where a promising case deserves closer examination.
Bring in human review.
Data science informs site, financial, operational, and partner decisions. No single score makes the investment for us.
Our models are
based in reality.
These site prediction model outputs come from testing on held-out historical charging sites. We measure the size of the miss, the direction of bias, and how well the model ranks one site against another.
4.15pp
Mean absolute error
Average gap between predicted and observed utilization.
17%
Lower error
Improvement over the previous model on the same test set.
0.904
Rank correlation
How closely the model’s site ranking matched observed results.
−1.91pp
Mean bias
Predictions leaned slightly conservative in this historical test.
Vikta Site Prediction V3, September 2026. Held-out evaluation of 178 historical sites across 8-, 12-, and 16-stall configurations.
Tested by the biggest EV network on earth.
In a same-site comparison run in conjunction with Tesla, Vikta’s utilization forecasts came in 15–30% below (better) Tesla’s estimates across the set.
1530%
below Tesla’s forecasts
The model keeps learning.
As Vikta sites come online, their observed performance informs later model revisions and future investment decisions. Each live site adds a new test of the original forecast.
01Forecast
Set a site expectation.
02Build
Put the selected site into service.
03Observe
Compare activity with the forecast.
04Refine
Inform the next model revision.
Let’s find the locations worth building.
Share your portfolio. We’ll evaluate where a charging investment could make sense.