Decision Log

The decision log is Adlyse’s system of record for what changed and why — every pause, budget bump, creative swap, anomaly, and measured outcome, timestamped per ad account.

It’s load-bearing for the impact-analysis card on the dashboard: when you see “paused campaign X → conversions up 12% next week,” that’s the decision log joining a past edit to a later metric shift.

Entry types#

Type	When it’s written
`decision`	A deliberate change made by a user or agent (“paused under-performing creative”)
`outcome`	A measured result of a prior decision
`anomaly`	An unexpected metric shift caught by sync
`edit`	A direct edit to campaign / ad-set / ad configuration

Who writes entries#

Sync pipeline — automatically writes anomaly entries when metrics shift past a threshold and edit entries when it detects configuration changes not attributable to an in-Adlyse action.
In-app assistant — when Adlyse AI pauses a campaign or bumps a budget, it writes a decision entry.
MCP write tools — update_campaign_status and update_campaign_budget each auto-write an edit entry on successful apply, tagged created_via="tool" with a fingerprint of the calling API key.

Fields#

Each entry has:

Decision text — short human-readable description (“paused ad 123 via Ads Manager”)
Entry type (above)
Outcome JSON — structured metric delta, {metric, before, after, window}
Tags — free-form classification, e.g. ["meta", "tofu", "us_fitness"]
Source ref — platform identifier (ad_id, campaign_id) when applicable
Metadata — for MCP-written entries, includes mcp_caller_fingerprint and the raw platform response
Created via — ad_sync, admin, or tool

Why this matters#

Decision logging is a big reason to route your AI workflows through Adlyse rather than a point-to-point “AI-plus-Meta-API” script. Adlyse’s impact analysis reads from this log when it correlates future metric movements with past decisions, so richer logs produce better future recommendations. Changes applied through the MCP write tools (update_campaign_status, update_campaign_budget) are logged here automatically.

Decision Log

The decision log is Adlyse’s system of record for what changed and why — every pause, budget bump, creative swap, anomaly, and measured outcome, timestamped per ad account.

Type

When it’s written

decision

A deliberate change made by a user or agent (“paused under-performing creative”)

outcome

A measured result of a prior decision

anomaly

An unexpected metric shift caught by sync

edit

A direct edit to campaign / ad-set / ad configuration

Who writes entries#

Sync pipeline — automatically writes anomaly entries when metrics shift past a threshold and edit entries when it detects configuration changes not attributable to an in-Adlyse action.

In-app assistant — when Adlyse AI pauses a campaign or bumps a budget, it writes a decision entry.

MCP write tools — update_campaign_status and update_campaign_budget each auto-write an edit entry on successful apply, tagged created_via="tool" with a fingerprint of the calling API key.

Fields#

Each entry has:

Decision text — short human-readable description (“paused ad 123 via Ads Manager”)

Entry type (above)

Outcome JSON — structured metric delta, {metric, before, after, window}

Tags — free-form classification, e.g. ["meta", "tofu", "us_fitness"]

Source ref — platform identifier (ad_id, campaign_id) when applicable

Metadata — for MCP-written entries, includes mcp_caller_fingerprint and the raw platform response

Created via — ad_sync, admin, or tool

Why this matters#