How the weighted forecast is computed (and why it's lower than raw pipeline)

[cognis-digital]

A common first question: where does the weighted forecast number come from, and why is it lower than the raw open-pipeline value?

It's computed entirely from your own deal history — no magic constants:

1. For each open stage, dealflow measures the advance rate = advanced / entered (the fraction of deals that entered the stage and later reached a strictly later stage).
2. P(win | stage) is the product of advance rates from that stage through to the won stage. So probability is monotonic: a deal in proposal has a higher P(win) than one still in lead.
3. The weighted forecast = Σ amount × P(win | current_stage) over every open deal. Won deals contribute 0 (already closed), lost deals have P(win) = 0.

In the bundled 01-basic sample: raw open pipeline is $60,000, the engine discounts it to a $32,444 weighted forecast — a 46% haircut driven by the team's real conversion rates.

The full walkthrough (with mermaid diagrams) is in docs/ARCHITECTURE.md. Questions about the math, edge cases (re-entered stages, missing amounts, mixed date formats), or how to model your own pipeline — ask here.