// decision log

Every row a reason. Every move reversible.

The decision log records every catalog change Magistry makes, with its trigger, its evidence and who applied it. In the last 30 days (measured 17 Oct 2026) that was 1,785 catalog writes applied to Shopify. Ad network changes and Brain actions are recorded the same way. Anatomy below.

Install on Shopify

How the audit plane works

Live A trigger on every row Catalog changes reversible Approvals recorded

catalog changes · example

#84193  DISCOUNT_TEST  SKU-LIN-228 (-15%)
        ACTIVE → ON_DISCOUNT       ✓ APPLIED

#84194  DRAFT          SKU-CER-019
        ON_DISCOUNT → DRAFTED      ✓ APPLIED

#84195  OPTIMIZE_LOSER SKU-WOO-441
        ACTIVE → OPTIMIZING        ✓ APPLIED

#84196  REVIVE_SEASONAL SKU-LIN-007
        SEASONAL_VAULT → PUBLISHED ⏸ WAITING
        trigger: SEASONAL_REVIVE
        reason: kill switch on since 09:14

cycle 2026-05-25T09:00Z

// anatomy

Read a row, understand the day.

Nine fields, the same on every row. The example below is a discount test for a single product, with the trigger, the evidence and the way back in one row.

catalog change · row #84193, example

row_id            : 84193
product_id        : p_88421
sku               : SKU-LIN-228-SND-L

from_state        : ACTIVE
to_state          : ON_DISCOUNT
trigger           : OPTIMIZE_CYCLE_FAILED

evidence          : {
  action            : discount_test
  performance_tier  : LOSER
  optimize_attempts : 1
  cost_tier         : A    (cost verified)
  margin_pct        : 41
  reasoning         : "loser, optimised once
                       already, margin OK →
                       last-chance discount"
}

revert            : restore price to 89.00;
                      stamps reverted_at / reverted_by

applied_to_shopify: true
applied_at        : 2026-05-25T08:02:14Z
approved_by       : operator, with biometric

status            : APPLIED
audit_chain       : signal → plan → approve → execute → log

↶ revert availablekept on record

product_id

uuid

The product the row pertains to. A variant id is added when the change is per variant.

p_88421

from_state

lifecycle_state

Where the product was before the change.

ACTIVE

to_state

lifecycle_state

Where it lands. A move outside the allowed lifecycle transitions is refused.

ON_DISCOUNT

trigger

Trigger enum

Why the change was proposed. One of a fixed list, never freeform text.

OPTIMIZE_CYCLE_FAILED

evidence

json

The move and the numbers behind it: performance tier, margin, cost tier, complaint signal, earlier attempts and the written reasoning.

{action: discount_test, cost_tier: A}

applied_to_shopify

bool

Whether the change reached Shopify. A call that changed nothing is not counted as a write.

true

applied_at

timestamptz

When the executor wrote the change.

2026-05-25T08:02Z

approved_by

user

Who approved it, and whether they confirmed with Face ID or fingerprint in the mobile app.

operator, with biometric

reverted_at

timestamptz

When the change was reverted, and by whom. The original row stays on record.

null

// triggers

Why a row exists.

Every decision row carries a trigger: the reason the engine emitted the action. Triggers are enumerated; no row gets a freeform 'because we felt like it'.

AUDIT_CLASSIFICATION

Each pass classifies a product by performance tier, margin and complaint signal, and the result proposes a move.

OPTIMIZE_CYCLE_FAILED

A product rewritten once without lift goes to a discount test when its margin allows, and is drafted when it does not.

DISCOUNT_CYCLE_FAILED

A product still losing after its discount test is drafted.

COMPETITOR_REPRICE

A tracked competitor sells the same product for less. A discount test is proposed above the store's margin floor.

PROMOTION_APPLY

A promotion started and set the product's sale price. When it ends, the price is put back and recorded as PROMOTION_REVERT.

SHOPIFY_STATE_RECONCILE

Someone changed the product in Shopify directly. The change is recorded so the history matches the store.

MANUAL_OVERRIDE

A person made the change. Logged the same way as an agent's move, in the same shape.

// evidence

The evidence behind the move.

The evidence field changes shape per action type. For a catalog lifecycle move it holds the move, the numbers that chose it and the reasoning in a sentence.

// the move

actionthe lifecycle move, from a fixed list
current_statethe product's state when it was read

// the product

performance_tierwinner, mid or loser on this store
margin_pctmargin at the current price
cost_tierhow well the product cost is known (A to C)

// the history

optimize_attemptsearlier copy rewrites for this product
complaint_signalcomplaints from the helpdesk about it

// the reasoning

reasoningthe engine's reasons, in order

evidence · example

{
  "action":            "discount_test",
  "current_state":     "ACTIVE",
  "performance_tier":  "LOSER",
  "margin_pct":        41,
  "cost_tier":         "A",
  "optimize_attempts": 1,
  "complaint_signal":  null,
  "reasoning": [
    "loser, optimised once already,
     margin OK → last-chance discount"
  ]
}

// reversal model

Revert writes the old value back.

Reverting a change writes the previous value back to Shopify and stamps the original row with when it was reverted and by whom. A revert takes the same store lock and respects the same kill switch as the original write.

// original

row #84193
  action      : discount_test
  delta       : -15%
  before      : 89.00
  status      : APPLIED
  applied_at  : 2026-05-25T08:02Z

// after revert (same row)

row #84193
  action      : discount_test
  price       : restored to 89.00
  reverted_at : 2026-05-25T11:32Z
  reverted_by : operator (jane@store.com)

the original row stays on record

// why it is kept