Exploratory brief
From charging
visibility to a
fleet decision.
Public-charging performance data can show what infrastructure is doing. Fleet context can show what a particular vehicle needs. This walkthrough explores how the two could combine.
Prepared for exploratory partner conversations. No integration or partnership is implied.
Illustrative demonstration — not a live operational recommendation.
The complementary layers
Charging-intelligence layer · potential authorised inputs
- Location
- Connector
- Nominal power
- Operational status
- Historic reliability
- Delivered-power patterns
- Site and operator performance
- Relevant customer-experience signals
One Fleet context
- Vehicle
- Current state of charge
- Load
- Route
- Delivery schedule
- Weather
- Driver preferences
- Required arrival reserve
- Alternative stopping options
One decision
- Leave now
- Charge before departure
- Charge once en route
- Recommended stopping location
- Expected duration
- Expected arrival state
- Backup action
- Honest confidence explanation
Outcome loop · potential consented outcomes
- Charger reached
- Actual wait
- Actual delivered power
- Session completed or failed
- Time spent
- Departure state
- Arrival state
- Driver-reported operational issue
Any sharing of outcome information would require an agreed lawful basis, data minimisation, commercial terms and technical governance.
Worked example · fictional journey
Liverpool → Cambridge
- Vauxhall e-Vivaro 75 kWh
- Heavy load
- Starting charge: 74%
Can this van complete the journey and return to its next required location?
Not recommended
Candidate A — 41 miles in
Too early in the journey; the vehicle would leave with capacity it cannot use.
Recommended
Candidate B — near Stafford
Mid-journey, adequate historic delivered power for the schedule, and a second site within a short detour as backup.
Not recommended
Candidate C — 168 miles in
Historic delivered power insufficient for this schedule, poor backup availability, and arrival reserve falls below the required margin.
Not recommended
Candidate D — off-corridor retail park
Detour creates an unacceptable delay against the return leg.
Charge once near Stafford.
Illustrative demonstration — not a live operational recommendation.
Confidence: limited
Known
- Vehicle specification
- Illustrative route
- Demonstration weather
- Starting charge
Not yet known
- Live charger performance
- Current queue
- Vehicle battery health
- Verified load
- Actual traffic
- Driver-hour constraints
This should not yet be used as a live operational recommendation.
Pilot hypothesis
Could charging-performance intelligence improve a fleet’s next decision?
Stage 1 — Synthetic test
- Agree data definitions
- Use fictional journeys
- Compare recommendation logic
- Identify missing information
- Define safety boundaries
Stage 2 — Historic replay
Subject to permission
- Use pseudonymised historic charging events
- Replay genuine commercial journeys
- Compare predicted and actual outcomes
- Test whether the decision would have improved
Stage 3 — Controlled field observation
Subject to governance
- Small invited vehicle group
- Advisory recommendations only
- Driver remains responsible
- Backup planning required
- No automated vehicle control
- Independent outcome evaluation
Proposed measures · not results
- Recommendation feasibility
- Arrival-state prediction error
- Expected versus actual delivered power
- Expected versus actual wait
- Failed-session handling
- Backup success
- Additional driver interaction
- Driver confidence
- Schedule adherence
Questions for a partner
- Which performance signals are sufficiently reliable for journey decisions?
- What data may be shared, and at what level of aggregation?
- Which outcomes would be valuable to return?
- What should never be inferred?
- How should confidence and uncertainty be expressed?
- What is the smallest useful synthetic test?
- Who would own derived learning?
- How should commercial attribution work?