Revenue leakage in charging is rarely fraud. It is sessions that delivered energy and produced no payment, and it persists because nobody is looking for something they never expected to arrive.
The amounts are individually small and cumulatively significant, and they are entirely detectable with the data most networks already hold.
Where energy and money separate
- Sessions delivered offline whose records never reached the platform.
- Backdated transactions rejected by a platform enforcing a recency window.
- Payment authorisations that succeeded but were never captured.
- Sessions with a start reading and no stop reading, so no billable quantity exists.
- Free or test sessions never marked as such and therefore never queried.
- Sessions billed at a tariff not in force when they occurred.
Offline delivery is the largest source
A charger operating offline delivers energy on trust and stores the record. That is only revenue-safe if the record arrives, is accepted, and is billed correctly.
The failure worth checking specifically is a platform rejecting transactions older than some window. Those sessions vanish silently, and the charger believes it delivered them successfully.
Authorisation is not capture
A payment authorised at the start of a session must be captured for the actual amount afterwards. Where capture fails, or where the session ended in a way the capture logic did not anticipate, the authorisation eventually expires and the money is never taken.
This produces a session that looks paid in the charging platform and never appears in settlement.
Detect by reconciliation, not by suspicion
Compare three records: what the charger reported, what the platform holds, and what settled. Every session should appear in all three, and every exception should be listed rather than absorbed.
Most networks compare two of the three, usually platform and settlement, which misses exactly the case where the charger delivered energy the platform never recorded.
Make it routine and small
Reconciliation run only when someone suspects a problem finds issues that are months old and whose evidence has rotated out of retention. Run on a schedule, exceptions surface while they are still resolvable.
The useful output is a count of unmatched records per period. A number trending upward indicates a systematic problem rather than isolated incidents.
Duplicates run the other way
The same mechanism that loses transactions also duplicates them, when a charger resends after an uncertain acknowledgement. Reconciliation should look for both, because billing a driver twice costs more in trust than the energy was worth.