The dashboard visualises the log. The log is the source of truth.

The direction of dependency

One tab holds facts, one tab holds pictures of them, and the arrow runs one way.

The log is the historical audit record; a helper table feeds the visuals; the dashboard displays them. Nothing flows back. That single-direction dependency is the reason the dashboard can be redesigned, extended or broken without any risk to the record, and it is also the reason a disagreement between the two is never genuinely a disagreement. If a chart shows something the log does not support, the chart is wrong — either the helper table's ranges have drifted, or the visual is aggregating differently than assumed. The instruction attached to the helper table is not to edit it except to repair formulas, which is the same posture the regime engine takes toward its mirrored inputs: a derived layer is maintained, not authored.

What it is genuinely for

Patterns across cycles are invisible in a table and obvious in a chart.

The dashboard is not a lesser version of the log; it does something the log cannot. A record of thirty cycles as rows is close to unreadable as a shape — tier usage drifting upward over six weeks, exposure drag increasing steadily, promotions clustering in particular conditions. Those are the questions the visuals answer, and the review uses named for them are exactly of that kind: detecting overuse or underuse of risk tiers, detecting hidden leverage stacking, evaluating the discipline of tier promotions. Each is a statement about a series rather than about a cycle, and each is nearly impossible to see by reading entries. The dashboard's contribution is compression across time, and compression across time is the one operation that genuinely requires a picture.

The failure

A chart makes a claim look settled that the underlying entries may not support.

The hazard in any derived visual layer is that rendering confers authority. A trend line through nine cycles looks like a finding whether or not nine cycles is enough to support one, and a bar that sits above its neighbours reads as significant regardless of what separates them. The dashboard also aggregates, so a single unusual cycle can move a series without being identifiable in it. This is why the pattern-recognition framing is precise rather than modest: the dashboard's correct output is a place to look, and the looking happens in the log entries that produced the pattern. An operator who concludes from a chart and never opens the underlying rows has drawn a conclusion from a summary of a summary.

The KPI strip

The most prominent element carries the least weight.

The latest-cycle strip shows gate, tier, pool, per-trade risk, directive, open drag, agreement and quality for the most recent cycle, and its stated use is fast situational awareness. That is a real job and a narrow one. It describes a single cycle, which is the sample size the whole system treats as insufficient for any structural conclusion — the same reason daily expectancy cannot promote and one week's contradiction reading is noise. So the strip is best read as orientation before a review rather than as a finding within one: it says where the system currently stands so that the charts and entries below can be read in context. The general principle applies here as everywhere in the workbook: the most visually prominent element is rarely the most authoritative one.

The review sequence

Strip for orientation, charts for patterns, log for adjudication.

The three surfaces compose into a short reading order that mirrors the discipline used elsewhere in the system. Begin at the strip to establish where things currently stand. Move to the charts to identify which series are moving and where the clusters fall. Then open the log entries behind whichever pattern looked real, and read the inputs, caps and effective decisions that produced it. The sequence matters because each stage narrows the next: the charts tell you which six cycles to read rather than all thirty, and the entries settle whether the pattern is what it appeared to be. Reversing it — reading entries first — is not wrong so much as inefficient, and stopping before the entries is where a review turns into an impression.

• Charts identify where to look; entries decide what is true.
• The dashboard aggregates, so a single outlier can move a series invisibly.
• The KPI strip describes one cycle — orientation, never a structural conclusion.

The key idea

Naming the source of truth is what keeps a derived layer honest.

Every system that renders its data eventually faces a moment when the rendering and the record disagree, and the outcome depends entirely on whether the hierarchy was settled beforehand. Stated in advance, the disagreement is a bug in the visual and gets fixed. Left unstated, it becomes an argument about which number to believe, and the more attractive surface tends to win. The single sentence declaring the log as the source of truth and the dashboard as pattern recognition costs nothing while everything is working and decides the case correctly on the day something is not.