The branch is declared before entry, never retro-fitted after.

The rule

Selected before entry, in section three, alongside the regime it must match.

The checklist's Regime & Branch section verifies the ATR Normal/Trend state, the selected branch, the match between them, the exit model, and the break-even rule — all before the Setup section is even reached. The page states the constraint plainly: the branch is selected before entry, never retro-fitted after the trade starts moving. And the pairing with regime in the same section is deliberate, because the two are one decision. A branch whose regime does not support it is a branch/regime mismatch, which the hard-failure list names explicitly.

The corruption

Retro-fitting produces branch statistics that describe outcomes, not decisions.

Consider what happens when the label is assigned after the fact. A trade that ran to 4R gets recorded as Trend No-Partial; one that stalled at 1.2R gets recorded as Normal. Nothing dishonest was intended and every individual label is defensible. But the resulting branch populations are now sorted by outcome rather than by intent, which means every branch EV in the system is measuring the same thing: how the trades that happened to do well did. Trend No-Partial's expectancy will look excellent because losers were filed elsewhere. Normal's will look flat. The blend will be meaningless, and nothing about the numbers will look wrong.

The downstream reach

Almost everything the system claims to know depends on this one field.

The branch label is not one field among many. It routes the trade into branch counts, branch EV, the blended expectancy that serves as the system's master health read, quota compliance for Trend No-Partial and Overflow, the MAE/MFE Lab's branch-analytics diagnosis, variant attribution, and the Scorecard's weekly probabilities. The QA rules make the consistency requirement explicit — keep branch labels consistent across TradeZella, paper sheets, CP3, the Weekly Scorecard, SDE, and Regime tools. A label that was honest at declaration and consistent across all six surfaces is what makes every one of those outputs mean something.

• Branch counts and branch EV — the direct consumers.
• Blended EV, quota governance, and profile weighting — all inherit the label.
• Branch_Analytics in the MAE/MFE Lab judges each branch against its own identity, which requires the identity to be real.

The honest version of the temptation

Mid-trade reclassification is usually a real observation, arriving too late.

It is worth being fair about why retro-fitting happens, because it is rarely cynical. An operator declares Normal, the trade breaks into clean expansion, and the observation that this is behaving like a trend trade is often correct. The problem is not the observation — it is that acting on it converts an entry-classification error into a data-integrity error, and buries the evidence that the error occurred. The system's answer is to let the mislabeled trade stand as declared and let the pattern surface in review, where a run of Normal trades exhibiting very high MFE with low capture is precisely the branch-integrity finding the MAE/MFE Lab is built to raise.

What it protects

The declaration is what makes a losing branch diagnosable.

The deepest reason for declaring first is that it preserves the system's ability to be wrong in a locatable way. When labels are honest, a branch that stops working produces a visible, attributable deterioration — that branch's EV falls, its checkpoint conversion weakens, and the review can descend into it. When labels track outcomes, a failing branch simply stops receiving winners and quietly hands its losses to whichever branch is being used as the catch-all. The failure still happens. It just becomes untraceable, which is the one outcome the entire lifecycle architecture exists to prevent.

The consistency requirement

One label, six surfaces, and no translation layer.

The QA rules ask for branch labels to be kept consistent across TradeZella, the paper sheets, CP3, the Weekly Scorecard, the SDE, and the Regime tools — six places, no automatic reconciliation between them. This is why the approved label set is fixed and short: Normal, Trend Partial, Trend No-Partial, Overflow, and All Blended where applicable. A trade recorded as “Trend” on paper and “Trend Partial” in CP3 will not error anywhere; it will simply be counted in one place and missed in another, and the discrepancy surfaces as a small unexplained gap between two workbooks that both look internally consistent.