METHODOLOGY
How an Account's Ordering Baseline Is Built
The baseline is simply how many days normally pass between an account's orders. This piece covers the clustering rule that treats orders placed within three days of each other as one purchase, the four-order minimum before any baseline is claimed at all, and why nothing is ever compared to a category average.
CATEGORY ANALYSIS
Long-Cycle Accounts and the False-Alarm Problem
Accounts that buy quarterly generate the most false positives in any system built on fixed thresholds. This analysis covers why the gap has to be measured against each account's own baseline, what happens to accounts with too little history to judge, and how the list stays short enough that a rep actually works it.
PRACTICE NOTE
Why There Is Nothing to Configure
Allodial Predict asks for no thresholds, no rules, and no tuning. This note explains the design decision: hand-set thresholds systematically override what an account's own history already says, and every knob added is one more thing to maintain as the book changes underneath it.
DRIFT FRAMEWORK
Watch, Slipping, Gone Quiet: What the Levels Mean
Three named levels, assigned by arithmetic on the gap between an account's last order and its own baseline. This piece defines each level, explains what moves an account from one to the next, and makes the case for naming the levels rather than numbering them.
PRACTICE NOTE
Logging Outcomes and Attributing Saves
A rep works an account, logs what happened, and the account leaves the list. When that account orders again, the save is attributed to the call that preceded it. This note covers what counts as a save, what does not, and why the outcome is the only thing anyone is ever asked to type.
