MLZP · Product Discovery · Thinking Canvas

Deepening the Member Association Model

Not a status page. The left is what we've settled, kept short. The right is what's actually open, which is what we're here to think through. Flip a card's state and jot notes as we go; it saves to a shared store, so anyone on the link sees the same board.

synced
Settled — context, not for debate

Here so we don't relitigate it. If any of it turns out to be wrong, that becomes an open card.

▪
Direct attribute matching is liveLIVE. Shared email, phone, name, device, bank, or wallet links two accounts and feeds the counts our checks read.
▪
The gap it leaves. Direct matching only catches one hop. Rings that share no single identifier, or move money between accounts, stay invisible.
▪
Direction: transitive linkage. If A links B and B links C, treat A and C as linked. Plus money-flow as a signal no shared attribute reveals.
▪
Enforce detection-first. Feed investigators as alerts first; gate bonuses only once precision is measured and trusted.
▪
Confidence admits edges; it doesn't score clusters. Weighting decides whether a link forms. Once formed, membership is yes/no, not a probability.
▪
We have ground truth already. Felix's SOP plus the Salesforce case history is a labeled set of adjudicated device-groups.
Open

The real material. Each card is a question we don't have a settled answer to. The short line under it is why it's hard or why it matters, so we start from the same place.