Demand response requires an assessment of how consumption would have changed without a dispatch event. Storage and generation have their own measurement boundaries and product assessments. These services should not inherit one assumed calculation method. At project initiation, record the locally accepted data sources, intervals, baseline method, adjustment conditions and appeal window, then check the metering and systems against them. Baseline design is product adaptation, not a discretionary choice of the most favourable algorithm after settlement.

ZHICE separates evidence into raw readings, operating context and calculated results. Readings retain timestamps, device identity and quality flags. Context records shutdowns, shift changes, weather or faults that may alter consumption. Calculations retain the rule version, exclusions and reasons for adjustments. Results without supporting material cannot resolve disputes. Large volumes of raw data without a defined calculation method are equally inadequate for an independent reviewer seeking to reproduce accepted delivery.

Missing data should neither be treated automatically as zero response nor silently filled as normal operation. The appropriate treatment depends on accepted substitute data and correction procedures. Internal exception tickets should separate repairable timestamp alignment issues from gaps that prevent proof of delivery. Any adjustment involving judgement should preserve before-and-after versions and approval. Model estimates may support investigation, but they cannot replace settlement evidence where the applicable rules do not accept them.

Before launch, ask a reviewer who did not participate in dispatch to reconstruct an event using only archived materials. Failure to reach the same conclusion identifies a gap in the record chain. This exercise demonstrates commercial readiness more directly than a live dashboard. Changes in a user’s production plan require a named reporting owner, a deadline for updating available capability, and a determination of whether the original baseline remains applicable. These decisions should precede repeated negotiations over data after the event.

Capability should be refreshed continuously rather than fixed indefinitely after initial testing. Production schedules, equipment and communications change. Historical response supports the next decision only under sufficiently comparable conditions. ZHICE recommends operating profiles for resource groups that identify reliable conditions, conditions requiring reduced commitments, and conditions requiring manual confirmation. The profiles should guide offers and substitutions, rather than becoming static scores used only in presentations.

Reviews should distinguish foreseeable deviations from unexpected events. Seasonal production changes can inform availability in advance; communications failures call for a review of resilience and recovery; measurement disputes require scrutiny of evidence rules. Combining all deviations into one failure rate prevents a decision on whether forecasting, maintenance, contracts or data need improvement. Management need not discuss every device alarm, but it should understand the recurring causes affecting settlement and the cross-functional decisions needed to resolve them.

Before changing strategy, record what will change and which result should improve. Observe the revised configuration within a limited resource set, compare delivery and acceptance under comparable conditions, and then consider wider use. Comparisons should account for differences in product, user status and observation window. Revenue gains in a high-price month do not establish technical effectiveness. Where evidence remains limited, preserve that uncertainty and use fewer exceptions and better explainability as interim objectives.

The loop also supports account management. Reliable resources with modest revenue may need a better product fit. Frequently dispatched resources with inadequate compensation may leave. Occasional failures that are inexpensive to correct may justify targeted upgrades. Retention decisions should consider contribution, the user relationship and improvement potential. Connection size alone is insufficient, and low short-term receipts do not automatically mean a low-value user if that user provides valuable availability at critical times.