Memory & Decision Ledger

Your company stops forgetting.

Companies pay for the same lesson every year: the lost customers found again, the failed channel tried again, the pricing mistake with new fonts. The company's memory lives in whoever hasn't quit yet.

Mechanism

Four memories.

Facts

What's true right now.

History

What happened: runs and events.

Decisions

What was chosen, and why.

Lessons

What works here: plays drawn from checked results.

What’s kept

One decision, sealed.

ledger · sealed

Expected

8–14

Actual

11

where things stood: Aster, week 2 — before launch
options: Start with services · Win design partners first · Run a paid launch
prediction: 8–14 good conversations (model v1.4)
decision, why, and who: Win design partners first — the founder and Cortex, together
missions: 3 missions set up
result: 11 good conversations
vs. the prediction: inside the expected range
lesson: more trust in v1.4's mid-budget forecasts

on record · sealed

where things stood: Aster, week 6 — after the pilot
options: Hold price · Raise 10% · Tiered pricing
prediction: about the same or better on conversion (model v1.3)
decision, why, and who: Raise 10% — the founder, though Cortex leaned neutral
missions: 1 mission set up
result: conversion flat, revenue per account up
vs. the prediction: I underestimated the revenue effect
lesson: next pricing forecast gives revenue per account more weight

on record · sealed

where things stood: Aster, week 9 — after the agency signal
options: Agency partnership · Direct outbound · Content
prediction: agencies are more willing to pay (flagged)
decision, why, and who: Direct outbound — the founder, following the flag
missions: 2 missions set up
result: outbound beat the agency baseline
vs. the prediction: the flagged guess held up
lesson: "agencies are more willing to pay" saved as a lesson

Three sealed entries. Click a spine to bring it forward.

Worked example

Aster's D-104 closes.

Expected 8–14, got 11. The agency signal becomes a new lesson (“agencies are more willing to pay”), and the next recommendation changes.

Kept honest

A fix adds a new version. It never erases the old one.

Predictions are kept word for word · memory follows who’s allowed to see what, and stays inside your company · a correction adds a new version and never overwrites the old one.

UnderstandDecideDeliverVerifyLearn

Results get sealed here, and the next recommendation gets sharper.

The hard question

Does my data train your model?

By default, your company’s data and results improve YOUR recommendations. Helping improve the shared model is something you choose, with your consent, and it’s de-identified. The exact rules are in the Trust Center, not in a marketing page’s fine print.

See Cortex Why this compounds