EV Charging → Question 6

Where Does Fidelity Change the Decision?

146 unique locations across three models. 72 all agree on. The other 74 — 39 A-only and 35 B&C-only — are where fidelity changes the decision.

72
All Models Agree
39
A-Only
35
B&C Only
Cross-Model Agreement Map — 146 Unique Locations
All models agree (49%) A-only (27%) B&C only (24%)

72 of 146 locations (49%): all three models agree. These are the obvious ones — clear demand, good grid access, no controversy. For these, Model A is sufficient.

39 of 146 (27%): unique to Model A. These are city-center sites where Model A's population-based scoring puts chargers. Model B/C's optimization finds these duplicate home charging and redirects to higher-value corridor and suburban locations.

35 of 146 (24%): unique to Model B/C. These are corridor and suburban sites that Model A's scoring formula misses entirely. From Model A's 111 stations, 72 overlap with B/C (65%), while 39 are unique to A (35%). From B/C's 49 stations, most overlap with some A picks, plus 35 locations A never considered.

Model Agreement Breakdown — 146 Unique Locations
Finding
All three models agree on 72 of 146 unique locations — the obvious ones with clear demand and good grid access. Model A uniquely picks 39 city-center sites that duplicate home charging. Model B/C uniquely identify 35 corridor and suburban sites that A's scoring formula misses entirely.
Most sites do not need the grid-aware model. For the 30 where substations run 65–78% loaded, Model C saves an average of $2.3M per station in NPV. The expensive part is not building the model—it is knowing which 30 stations to run it on.

The Right Model for the Right Question

Question
Right Model
Why
Q1: Gap Analysis
Model A sufficient
237 uncovered tracts found. Population/traffic screening finds the obvious gaps.
Q2–Q3: Placement + Sizing
Model B necessary
88.5% vs 76.1% coverage. Demand-based optimization relocates 35% of stations and right-sizes ports per station.
Q4–Q5: Grid + Cost
Model C essential
$68M NPV savings. 30 of 49 stations can't get full power. Grid check using 4,151 real substations prevents 294 stranded ports.
Q6: Fidelity Map
All three, in sequence
Start simple. Escalate where the decision changes. The pattern: screen → optimize → constrain.

Cross-model comparison: pairwise location matching with 5-mile tolerance.

← Prev Q5: True Cost Study Back to EV Charging Next → Q7: Model Validation