Circuit-level review answers one question: does this result pass? It doesn't answer a different and often more useful one — why are twenty circuits in this section all producing similar values? Those are different analytical tasks, and commissioning test packs frequently undergo the first without the second.
The pass/fail layer is necessary but not sufficient. What the aggregate data communicates about installation quality, design-margin consumption, and shared upstream conditions is a separate layer of analysis — one that only becomes visible when results are read across a section rather than sequentially down a schedule.
What circuit-level review doesn't surface
A test pack reviewed circuit by circuit will identify any result that exceeds a threshold. It's less likely to flag a section where all results are clustering toward the threshold without crossing it. Individually compliant, collectively indicative — that distinction matters for a pre-energisation review where the question isn't just whether each circuit passed, but whether the installation is performing as engineered.
Pattern recognition in this context isn't searching for failures. It's looking for distributed characteristics that carry information about the installation condition before any individual result becomes a problem. The value is that it can bring an investigation to the right starting point before the first operational load cycle.
IR clustering at the lower end of tolerance
Insulation resistance values across a board or cable section are worth reviewing as a population, not just individually. Multiple circuits from the same distribution point all passing but clustering toward the lower end of acceptable range may suggest a shared environmental condition affecting insulation across that section — moisture ingress during installation, cable ends left exposed before termination, a shared damp routing that hasn't yet resolved.
The distinction is between a single circuit at 5 MΩ on new cable where the threshold is 1 MΩ — worth noting, possibly worth a retest after the environment stabilises — and eight circuits from the same board ranging 1.2 to 3 MΩ, all passing, none individually alarming. The latter is a distribution pattern worth investigating. It may indicate nothing more than installation sequence, or it may indicate a shared condition that degrades further under operational environment. Hard to say for sure from the data alone, but it warrants the question.
Zs elevated uniformly across a section
Where Zs values across a circuit section are all slightly elevated — within tolerance, but consistently closer to the limit than adjacent sections on comparable cable lengths — the pattern may indicate a shared upstream impedance contribution rather than individual termination issues. A common feed cable or earth path carrying higher resistance than design assumption will propagate a consistent offset to every downstream circuit. Circuit-level review sees each result passing. Section-level review sees all results elevated by the same margin relative to circuits drawn from elsewhere in the installation.
The useful comparison is against design intent, not just against the statutory threshold. Supply impedance is calculated during design from transformer characteristics and cable schedule data; a measured Ze above the design figure warrants investigation independent of whether individual circuits technically satisfy the disconnection-time requirement. Where design margin was assumed rather than verified, it may already be consumed by the time test data is available.
Section average comparison first. Plot Zs averages by section and compare against sections on similar protective device types and comparable cable runs. A consistent elevation across one section relative to others is a stronger signal than any individual result.
Ze at origin. Direct measurement at the supply origin for the affected section isolates upstream contribution from installation variables. A confirmed elevated Ze shifts investigation to the supply path rather than individual circuit terminations.
Main earthing terminal and bonding arrangement. Connection integrity and conductor sizing at the earthing terminal. A consistent Zs offset across a section often traces here rather than to individual circuit faults.
RCD trip times skewing toward the limit
RCD trip times clustered toward the upper end of acceptable range across a panel or section are worth noting as a population characteristic, even where each device individually satisfies the time limit. Where trip times from a common batch or a single installation crew skew consistently toward the maximum, the pattern may suggest consistent mechanical drag, pre-load on the mechanism, or batch quality variation that a single result wouldn't reveal.
RCDs tripping at 95–100% of the permitted time limit pass the test. Whether they remain within limit after years of operational environment, vibration, and thermal cycling is a different question. Commissioning data read as a pattern can sometimes hint at that trajectory. It's worth capturing the observation in the review record even where closure doesn't require immediate action.
Reading the section, not just the schedule
Good test pack review uses individual circuit results as primary data and then makes a second pass at section and system level — grouping by distribution board, cable section, installation crew, test date. Clustering, shared deviations, and trends that only become visible when results are aggregated are what circuit-by-circuit review systematically misses.
The section-level view is also what connects individual test results to upstream conditions — the kind of finding that collapses a verification sequence from twenty individual investigations to one. On an anonymised data centre pre-energisation review, a section-level Zs comparison across two distribution feeds from the same upstream assembly identified a supply-side impedance contribution that each individual circuit result would have taken considerably longer to locate. The pattern made the starting point obvious. Working circuit by circuit would have delayed it.