Record actions and measure their impact

Actions you take cannot be measured later unless they are recorded. This category covers:

• Action log: bid adjustments, budget changes, promotions, and other moves you make yourself
• External event: TV broadcasts, sales, buzz, stockouts, and other things that happen outside your control

The actions you record here are used by Prompt 3 “Before/After comparison” in View KPIs on the dashboard.

When to use this category

• “I want to record last week’s bid changes as an action”
• “I want to log the TV broadcast date”
• “I want to register the sale period”
• “I want to exclude a stockout period from analysis”

Pair it with Before/After comparison

Recording an action log only pays off when you measure it. 7-14 days after the action, pass the action ID to Prompt 3 “Before/After comparison” in View KPIs on the dashboard and the deltas in ACoS / ROAS / ad spend / sales are aggregated automatically.

---

Prompt 1: Register an action log

🎯 Purpose

Record an executed action (bid adjustment, budget change, promotion, etc.) in a form that can be used for Before/After comparison.

📋 When to use

• You want to record a bid change or budget change as an action
• You want to leave the date and details so you can run a Before/After measurement later
• You want to link the action to a proposal ID executed via the bid optimization prompt

📝 Prompt

Register an action log.
- Type: {{ACTION_TYPE}}
- Name: {{ACTION_NAME}}
- Date executed: {{ACTION_DATE}}
- Details: {{DESCRIPTION}}
- Status: {{STATUS}}
- Proposal ID: {{PROPOSAL_ID}}

Placeholders:

Placeholder	Description	Default source
{{ACTION_TYPE}}	Action type (cpc / budget / placement / promotion / negative / custom)	User input (required)
{{ACTION_NAME}}	Action name (e.g. Bid -22% on 18 high-ACoS keywords)	User input (optional)
{{ACTION_DATE}}	Date executed (YYYY-MM-DD)	User input (required)
{{DESCRIPTION}}	Detailed notes	User input (optional)
{{STATUS}}	pending / monitoring / completed / archived	Default monitoring
{{PROPOSAL_ID}}	Proposal ID executed via Adjust bids, etc.	Optional; leave blank if not applicable

⚙️ What the AI does

1. Normalize the type — Normalize to one of the six types (cpc / budget / placement / promotion / negative / custom)
   • Why six types: So that the comparison metric can be switched per action class at Before/After aggregation. For example, bid actions are read against ACoS movement, while budget actions are read against absolute sales change.
2. Validate the date — Accept past dates only (warn on future dates)
   • Why past dates only: Action logs are intended to record moves that have already been made. Forward-looking plans belong in Automation rules.
3. Link to proposal ID — When you specify a proposal ID from the bid optimization or similar prompt, the link back to the source data is preserved.
4. Default status — If unspecified, registered as monitoring (measuring impact).
5. Return the registration result — Present the assigned action ID (format evt_YYYY_MM_DD_<type>) and all fields in a table.

📊 Example Output

Action log registered

Field	Value
Action ID	evt_2026_05_18_cpc
Type	cpc (bid adjustment)
Name	Bid -22% on 18 high-ACoS keywords
Date executed	2026-05-18
Status	monitoring
Proposal ID	prop_xxxxxx
Details	Target ACoS 15% vs. recent 21.8%. Average -22% reduction approved and executed.

For impact measurement, on 2026-05-25 (7 days later) or 2026-06-01 (14 days later) pass evt_2026_05_18_cpc to Before/After comparison.

🔒 Safety

• Action IDs are auto-assigned at registration (no duplicates or overwrites to an incorrect ID).
• If you specify an existing action ID for registration, new creation is rejected and you are directed to Prompt 2 for updates.
• Future-dated executions return a warning and the registration is held.
• Registration events are recorded in the audit log (deletions still leave a trace in the audit log).

🚦 Execution mode

Phase	State	Conditions
Phase 1 (current)	One-shot	Available with Picaro connection only
Phase 2 (Q3 2026)	Stays one-shot	Each action’s content needs human review
Phase 3 (Q4 2026)	One-shot (with templates)	Save frequently-used action contents and recall them
Phase 4 (2027)	Auto-recorded	Action logs are generated automatically when Adjust bids etc. is executed

👉 Next steps

• Before/After comparison — Measure impact with this action ID 7-14 days later
• List / update / add notes to action logs — Leave observations during the monitoring period as notes
• Create reports — Bundle the action log into the monthly report

❓ FAQ

Q. Should I split multiple actions taken on the same day into separate logs? A. If the types differ, register them separately (e.g. bid adjustment and budget change as separate logs). If the type is the same and only the campaign or target differs, combine them into one entry and list the targets in the details. This makes it easier to isolate which one contributed during Before/After comparison.

Q. What if I don’t see an effect immediately after executing an action? A. For bid actions, the guideline is to measure with finalized weekly data at least 7, ideally 14, days later. By design, value swings during the learning period are not used as judgment material — keep the status at monitoring and proceed to Before/After comparison once the period is finalized.

---

Prompt 2: List / update / add notes to action logs

🎯 Purpose

Review the action logs you have previously registered, and add progress notes or update their status.

📋 When to use

• You want to see a list of past actions
• You want to filter only the actions currently being monitored
• You want to add notes such as “starting to see results” or “unexpected side effect” to an action
• You want to update a completed action’s status to completed

📝 Prompt

Operate on an action log.
- Operation: {{ACTION}}
- Target ID: {{EVENT_ID}}
- Filters: type={{TYPE_FILTER}} / status={{STATUS_FILTER}}
- Update fields: {{UPDATE_FIELDS}}
- Note body: {{NOTE}}

Placeholders:

Placeholder	Description	Default source
{{ACTION}}	One of list / get / update / remove / add_note	User input (required)
{{EVENT_ID}}	Target action ID (for get / update / remove / add_note)	Auto-referenced from Prompt 1 registration result
{{TYPE_FILTER}}	Filter by type (when listing, e.g. cpc)	Optional
{{STATUS_FILTER}}	Filter by status (when listing, e.g. monitoring)	Optional
{{UPDATE_FIELDS}}	Fields to update (name / status / description, etc.)	Update only
{{NOTE}}	Note body	add_note only

⚙️ What the AI does

1. Route by operation — Distinguish between the five operations: list / get / update / remove / add_note.
2. Default sort for list — Up to 20 entries shown newest-first by execution date.
   • Why 20: Returning every record makes the list hard to read; prompt the user to specify filters for more.
3. Confirm status transitions — Recommend the flow pending → monitoring → completed / archived, and warn on reverse transitions.
   • Why define a flow: Without team-wide alignment on status meaning, “actions still being monitored” can mix into report aggregations and break the judgment.
4. Append-only notes — Existing notes are not overwritten; new notes are appended with a date.
   • Why append: To preserve the chronological observation log of the action. You can track “Day 3: results starting” → “Day 7: ROAS +18%”.
5. Confirm before delete — Before deletion, present a summary of the target action and do not execute until the user explicitly says “delete it”.

📊 Example Output

Action log list (status=monitoring / last 30 days / 5 entries)

Action ID	Type	Name	Date executed	Status	Days elapsed
evt_2026_05_18_cpc	cpc	Bid -22% on 18 high-ACoS keywords	2026-05-18	monitoring	0 days
evt_2026_05_11_budget	budget	Pre-Prime Day budget +30%	2026-05-11	monitoring	7 days
evt_2026_05_04_negative	negative	Exclude 12 zero-conversion keywords	2026-05-04	monitoring	14 days
… (2 more) …

evt_2026_05_04_negative, which has passed 14 days, is ready for Before/After comparison. To measure impact, reply with “Run Before/After for evt_2026_05_04_negative”.

🔒 Safety

• Operations on non-existent action IDs are rejected (prevents overwrites against incorrect IDs).
• Deletes require a confirmation step (not executed until the user explicitly says “delete it”).
• Reverse status transitions such as completed → monitoring return a warning.
• Notes are append-only; existing notes cannot be overwritten (to protect the history).
• The audit log records every update and deletion operation.

🚦 Execution mode

Phase	State	Conditions
Phase 1 (current)	One-shot	Available with Picaro connection only
Phase 2 (Q3 2026)	Daily execution	Check the list of monitored actions every morning via “saved prompts”
Phase 3 (Q4 2026)	Auto-delivered	Actions at 7 / 14 days elapsed are pushed to Slack automatically
Phase 4 (2027)	Auto + status update	After Before/After aggregation, status auto-updates based on the verdict (final approval still human)

👉 Next steps

• Before/After comparison — Measure impact starting with actions that have enough elapsed days
• Register an action log — Record a new action
• Create reports — Bundle a summary of monitored actions into the monthly report

---

Prompt 3: Record an external event

🎯 Purpose

Record external factors — TV broadcasts, sales, buzz, stockouts — that you did not cause yourself, so they can be referenced later as reasons for KPI movement.

📋 When to use

• You want to log a TV broadcast date
• You want to register a sale period (Prime Day, Black Friday, etc.)
• You want to record a day on which buzz occurred
• You want to exclude a stockout period from analysis

📝 Prompt

Record an external event.
- Type: {{EVENT_TYPE}}
- Start date: {{START_DATE}}
- End date: {{END_DATE}}
- Label: {{LABEL}}
- Details: {{DESCRIPTION}}

Placeholders:

Placeholder	Description	Default source
{{EVENT_TYPE}}	tv / sale / buzz / stockout / other	User input (required)
{{START_DATE}}	Start date (YYYY-MM-DD)	User input (required)
{{END_DATE}}	End date (for ranged events)	Optional; leave blank for single-day events
{{LABEL}}	Display name (e.g. Prime Day 2026 Spring)	Optional
{{DESCRIPTION}}	Detailed notes (program name, target ASINs for the sale, etc.)	Optional

⚙️ What the AI does

1. Normalize the type — Normalize to one of the five types (tv / sale / buzz / stockout / other).
   • Why five types: To separate “change driven by your own move” from “change driven by external factors” when interpreting KPI movement. tv / sale are read as sales uplift, while stockout is read as sales drop.
2. Validate the range — Confirm that start date ≤ end date.
3. Check for duplicate events — Prevent duplicate registration if an event of the same type and same period already exists.
4. Store separately from action logs — Because these are not your own actions, they are stored in a different table from action logs.
   • Why separate: When you want to see “the effect of your own action” in Before/After comparison, mixing in external-factor events skews the judgment.
5. Return the registration result — Present the event ID (format ext_YYYY_MM_DD_<type>) and all fields in a table.

📊 Example Output

External event registered

Field	Value
Event ID	ext_2026_06_11_sale
Type	sale (sale period)
Label	Prime Day 2026 Spring
Start date	2026-06-11
End date	2026-06-12
Details	All ASINs, 20% OFF coupon distributed concurrently

When you run a Before/After comparison covering this period, the event will be annotated as an external factor influencing KPI movement.

🔒 Safety

• If an event of the same type and same period already exists, duplicate registration is prevented and the existing event ID is returned.
• An error is returned when the end date is earlier than the start date.
• External events are annotated as “possibly influenced by external factors” in Before/After comparison, but the period is not auto-excluded (judgment is left to the user).
• The audit log records every registration and update operation.

🚦 Execution mode

Phase	State	Conditions
Phase 1 (current)	One-shot	Available with Picaro connection only
Phase 2 (Q3 2026)	One-shot	Continues manual registration each time
Phase 3 (Q4 2026)	Semi-auto	Auto-import the Amazon sale calendar + confirmation
Phase 4 (2027)	Auto	Stockouts and sales are auto-detected and registered from external data; TV / buzz remain manual

👉 Next steps

• Before/After comparison — Review the verdict carefully for periods that overlap external events
• Register an action log — Record your own actions tied to the external event (e.g. front-loaded budget for the sale) separately
• See sales contribution via N-gram analysis — Check the shift in search terms driven by buzz or TV broadcasts

❓ FAQ

Q. Can I exclude a stockout period from Before/After comparison? A. Auto-exclusion is not performed. During Before/After comparison, you receive an annotation such as “a stockout event (ext_xxxxxx) overlaps the After period”, and you decide whether to trust the verdict or re-specify the period. The impact of a stockout varies greatly by product and category, so a mechanical exclusion would cause misjudgments.

Q. Should a sale period and a budget-front-loading action be registered together? A. Registering both is standard practice. Put the sale period under external events (Prompt 3), and the corresponding budget front-loading under the action log (Prompt 1). This makes it easier to separate “external factor” vs. “own action” contributions in Before/After comparison.

---

Prompt quick reference

What you want to do	Prompt to use
Register an action log (for later Before/After comparison)	Prompt 1
List / update / add notes to action logs	Prompt 2
External event (TV / sale / buzz / stockout)	Prompt 3

Related categories

• View KPIs on the dashboard — Use recorded actions in Prompt 3 “Before/After comparison”
• Adjust bids / Adjust budgets — Standard practice is to combine proposal execution with leaving an action log
• Set monthly KPI targets — Register target ACoS / TACoS in advance as the impact-judgment baseline
• Create reports — Bundle action logs and external events into the monthly report