Adjust bids

Bid adjustment is the biggest lever for moving ACoS toward your target. This category covers 7 patterns: lowering and raising bids at the keyword level, weighted adjustments across keywords / ASINs / auto targets, following Amazon’s suggested bid, and placement multipliers.

When to use this category

• “I want to lower bids on keywords with high ACoS”
• “I want to adjust bids across keywords, ASINs, and auto campaigns together”
• “I want to follow Amazon’s suggested bid”
• “I want to raise bids where IS% has headroom”
• “I want to optimize placement multipliers by CVR per placement”

Register this first to save time

If you register a target ACoS in Monthly KPI targets, later prompts will reference it automatically — you won’t need to write “target ACoS X%” every time.

---

Prompt 1: ACoS-based bid reduction (keywords)

🎯 Purpose

Safely lower bids on keywords where ACoS is above target, with clear justification.

📋 When to use

• You want to lower bids on keywords running above target ACoS
• You want a list of keywords that have room to be lowered against your target ACoS
• You want to trim wasted ad spend

📝 Prompt

Against an ACoS target of {{TARGET_ACOS}}, recent ACoS is trending at {{RECENT_ACOS}}.
Using the most recent finalized week, with min_clicks 10 and excluding zero-sales keywords,
list keywords with room to lower the bid, including Suggested bids.

Placeholders:

Placeholder	Description	Default source
{{TARGET_ACOS}}	Target ACoS (e.g., 15%)	Auto-referenced from Monthly KPI targets
{{RECENT_ACOS}}	Recent actual ACoS (e.g., 18%)	Auto-referenced from dashboard KPI

⚙️ What the AI does

1. Filter target keywords — Finalized-week data × min_clicks 10 or more × exclude zero-sales
   • Why finalized week: Trend judgments on unfinalized weeks are unstable because data isn’t complete
   • Why min_clicks 10: Keywords with few clicks have low judgment accuracy and high risk of misclassification from insufficient data
   • Why exclude zero-sales: To avoid divide-by-zero accidents when sales = 0 (negative-candidate handling is covered in a separate prompt)
2. Compute the Suggested bid — Current bid × (target ACoS ÷ actual ACoS)
3. Floor at 1 JPY — Clamp results below 1 JPY to 1 JPY (prevents invalid bids)
4. Return an approval table — 4 columns: count, current bid, Suggested bid, expected ACoS

📊 Example output

Recommended bid changes: 12 items

Keyword	Current bid	Suggested bid	Expected ACoS
amino acid shampoo	¥95	¥78	14.2%
refill shampoo	¥110	¥88	14.8%
…(10 more)…

Estimated monthly ad-spend reduction of ~¥48,000 across all 12 items. Reply “execute” to apply. To approve only certain items, say e.g. “execute item 1 only”.

🔒 Safety guards

• Change magnitude is auto-clamped at a -30% floor (prevents abrupt swings)
• Nothing is applied unless you explicitly say “execute” (no changes until approved)
• The Suggested bid floor is 1 JPY (prevents invalid values)
• Each suggestion has a 1-hour validity window; after that you’ll need to regenerate
• Rollback can be restored from bid history (contact support)

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot	Available with just Picaro connection
Phase 2 (Q3 2026)	Daily execution	Monthly KPI target registered + saved-prompt enabled
Phase 3 (Q4 2026)	Auto-execute (approval-flow)	Slack 1-click approval, weekly, up to 100 items/week
Phase 4 (2027)	Fully automated	ML confidence ≥ 0.85, within guardrails

👉 Next steps

• Set monthly KPI targets — Skip having to write {{TARGET_ACOS}} each time
• Add negative keywords — Handle zero-sales keywords separately
• Why ACoS alone isn’t enough — How to combine with TACoS and ROAS

❓ FAQ

Q: How are zero-sales keywords handled? A: They are excluded by this prompt. To pause or negative them, use Prompt 3.

Q: Can I execute only part of the list? A: Yes. Specify by number, e.g. “execute only items 1 and 3”.

---

Prompt 2: Unified optimization across all targets (keywords + ASINs + auto, weighted)

🎯 Purpose

Review keywords, ASINs, and auto targets together, returning Suggested bids weighted by the confidence of each match type.

📋 When to use

• You want to adjust bids across keywords, ASINs, and auto campaigns together
• You want to push Exact match harder while easing back on Broad
• You want to optimize bids with different weights per match type

📝 Prompt

Using the most recent finalized week, against a target ACoS of {{TARGET_ACOS}},
review bids across all targets (keywords, ASIN targets, and auto targets).
With min_clicks 10 and min_orders 1 or more, weight Exact match at {{EXACT_WEIGHT}}x
and Broad match at {{BROAD_WEIGHT}}x and return Suggested bids.

Placeholders:

Placeholder	Description	Default source
{{TARGET_ACOS}}	Target ACoS (e.g., 15%)	Auto-referenced from Monthly KPI targets
{{EXACT_WEIGHT}}	Multiplier applied to Exact match (e.g., 1.1)	Default 1.1 (Exact has high confidence)
{{BROAD_WEIGHT}}	Multiplier applied to Broad match (e.g., 0.9)	Default 0.9 (Broad is for discovery)

⚙️ What the AI does

1. Aggregate across all targets — Unify keywords, ASIN targets, and auto targets into a single table
   • Why unify: ACoS plays out differently per target type even within the same campaign, so cross-target comparison lets you optimize allocation
2. Volume cutoff — Filter by min_clicks 10+ and min_orders 1+
   • Why min_orders 1+: Don’t raise bids on targets with zero orders (prevents over-investment)
3. Weight by match type — Exact = high confidence (push harder), Broad = exploratory (ease back)
4. Clamp change magnitude — Cap at +20% / -30% per adjustment
   • Why +20% / -30%: Prevents abrupt swings per adjustment and keeps impact observable
5. Return an approval table — 5 columns: target type, match type, current bid, Suggested bid, expected ACoS

📊 Example output

Recommended bid changes: 28 items (keywords 18 / ASINs 6 / auto 4)

Target	Type	Current bid	Suggested bid	Expected ACoS
amino acid shampoo (Exact)	Keyword	¥95	¥82	14.5%
B0XXXXXXXX	ASIN	¥120	¥96	15.1%
close-match (auto)	Auto	¥85	¥72	14.9%
…(25 more)…

Exact match weighted 1.1x and Broad weighted 0.9x. Reply “execute” to apply.

🔒 Safety guards

• Change magnitude is auto-clamped at +20% / -30% (prevents abrupt swings)
• Nothing is applied unless you explicitly say “execute”
• Suggested bid floor is 1 JPY
• Each suggestion has a 1-hour validity window
• Rollback can be restored from bid history (contact support)

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot	Available with just Picaro connection
Phase 2 (Q3 2026)	Weekly / daily	Monthly KPI target registered + saved-prompt enabled
Phase 3 (Q4 2026)	Auto-execute (approval-flow)	Slack 1-click approval, weekly, within cap
Phase 4 (2027)	Fully automated	ML confidence ≥ 0.85, within guardrails

👉 Next steps

• Add negative keywords — Surface zero-conversion targets for exclusion
• Promote auto to manual — Promote winning auto targets to manual campaigns

---

Prompt 3: Carve out zero-sales targets as negative candidates

🎯 Purpose

Surface targets that are getting clicks but no sales as exclusion candidates — rather than lowering their bid.

📋 When to use

• You want to stop targets that have clicks but zero sales
• You want to negative out search terms eating clicks for no return
• You want to clean up non-converting targets

📝 Prompt

From the most recent finalized week, with min_clicks {{MIN_CLICKS}}+ and zero sales,
carve out targets into a separate list of negative-keyword candidates.
Run zero_sales_action in negative_candidate mode.

Placeholders:

Placeholder	Description	Default source
{{MIN_CLICKS}}	Minimum clicks to include (e.g., 10)	Default 10

⚙️ What the AI does

1. Extract candidates — Finalized-week × min_clicks 10+ × sales = 0
   • Why min_clicks 10: With too few clicks there’s not enough signal — this prevents accidentally excluding targets that are actually working
2. Separate from bid adjustments — Carve out as a dedicated negative-candidate list
   • Why separate: If mixed into the default bid-reduction mode, they’d surface as “-30%” suggestions and the targets you should actually stop would linger
3. Hand off to negative keyword addition — Post-approval execution flows through the standard Add negative keywords workflow

📊 Example output

Negative keyword candidates: 8 items

Target	Type	Clicks	Spend	Sales
bulk shampoo (Broad)	Keyword	24	¥1,920	¥0
B0XXXXXXXX	ASIN	18	¥1,440	¥0
…(6 more)…

¥9,800 of ad spend in the most recent finalized week produced no sales. To convert these to negatives, proceed to Add negative keywords.

🔒 Safety guards

• Only carves out the list — does not auto-register negatives
• Lowering min_clicks increases false positives — verify item counts are reasonable before changing
• Each suggestion has a 1-hour validity window
• Rollback after negative registration is handled via the Amazon Ads console or Add negative keywords

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot	Available with just Picaro connection
Phase 2 (Q3 2026)	Daily execution	Monthly KPI target registered + saved-prompt enabled
Phase 3 (Q4 2026)	Auto-execute (approval-flow)	Slack 1-click approval, weekly, up to 50 items/week
Phase 4 (2027)	Fully automated	ML confidence ≥ 0.85, within guardrails

👉 Next steps

• Add negative keywords — Register the negatives from this list
• Promote auto to manual — Clean up zero-conversion auto-derived targets

---

Prompt 4: Follow Amazon’s Suggested bid trend

🎯 Purpose

For targets without enough actual ACoS data, propose stable trend-following bid changes based on the time series of Amazon’s Suggested bid.

📋 When to use

• You want to align with Amazon’s Suggested bid
• You want to judge by market bid trends rather than ACoS
• You want to move bids on new campaigns where actual ACoS isn’t available yet

📝 Prompt

Using recent weekly data, look at the time series of Amazon's Suggested bid
and propose candidates whose bid should follow the suggestion.
Use min_trend_points {{MIN_TREND_POINTS}}+ and volatility under {{VOLATILITY_LIMIT}};
skip anything with high volatility.

Placeholders:

Placeholder	Description	Default source
{{MIN_TREND_POINTS}}	Minimum data points required for trend judgment (e.g., 3)	Default 3
{{VOLATILITY_LIMIT}}	Volatility ceiling for skip judgment (e.g., 30%)	Default 30%

⚙️ What the AI does

1. Fetch the Suggested bid time series — Collect weekly history of Amazon’s Suggested bid
   • Why not look at actual ACoS: So this works on ramping campaigns too — judgment is based on a market-derived reference value
2. Data-point cutoff — Require min_trend_points 3+
   • Why 3: Excludes targets with too little data and keeps trend judgments reliable
3. Volatility check — Skip volatility ≥ 30%
   • Why skip: Following whipsawing targets would make your bid whipsaw too, defeating the stable-following goal
4. Stepwise following — For uptrends, propose gradual increases

📊 Example output

Suggested bid follow candidates: 9 items

Target	Current bid	Suggested bid (recent)	Trend	Recommended action
amino acid shampoo (Exact)	¥95	¥118	4 weeks rising	+15% raise
B0XXXXXXXX	¥120	¥102	Gentle decline	-10% lower
…(7 more)…

Skipped 3 items with volatility ≥ 30%. Reply “execute” to apply.

Details: Amazon Suggested bid trend analysis

🔒 Safety guards

• Change magnitude is auto-clamped at +20% / -30%
• Targets with volatility ≥ 30% are auto-skipped
• Nothing is applied unless you explicitly say “execute”
• Each suggestion has a 1-hour validity window
• Rollback can be restored from bid history

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot	Available with just Picaro connection
Phase 2 (Q3 2026)	Daily execution	Monthly KPI target registered + saved-prompt enabled
Phase 3 (Q4 2026)	Auto-execute (approval-flow)	Slack 1-click approval, weekly, within cap
Phase 4 (2027)	Fully automated	ML confidence ≥ 0.85, within guardrails

👉 Next steps

• ACoS-based bid reduction — Combine with this once actual ACoS becomes available
• Automation phases 1–4 — Automation scenarios for trend following

---

Prompt 5: Bid optimization by top-of-search IS% and CVR

🎯 Purpose

Combine missed top-of-search impressions (IS% headroom) and week-over-week CVR change to simultaneously surface keywords whose exposure should grow and those whose bids should shrink.

📋 When to use

• You want to raise bids on keywords with Impression share headroom
• You want to lower bids on keywords whose CVR is dropping
• You want to capture more top-of-search share

📝 Prompt

Using the most recent finalized week (Sun–Sat), compare top-of-search
Impression share and CVR week-over-week and propose bid changes.
With min_clicks 10 and min_impressions {{MIN_IMPRESSIONS}}+, raise bids on
keywords with IS% headroom and lower bids on keywords with falling CVR.

Placeholders:

Placeholder	Description	Default source
{{MIN_IMPRESSIONS}}	Minimum impressions to qualify (e.g., 50)	Default 50

⚙️ What the AI does

1. Force finalized week (Sun–Sat) — Exclude unfinalized weeks
   • Why finalized week: Prior-period data isn’t complete on unfinalized weeks, so trend judgment fails — WoW comparison only works on finalized data
2. Cutoffs — min_clicks 10+ and min_impressions 50+
   • Why min_impressions 50: Impression share calculations lose precision when the denominator is small
3. IS% × CVR combined judgment
   • Keywords with IS% headroom (missed impressions) → Raise bids for more exposure
   • Keywords with falling WoW CVR → Lower bids for better efficiency
4. Change magnitude clamp — Cap at +30% / -20%
   • Why +30% / -20%: Operationally, push harder for exposure gains while easing CVR drops gradually

📊 Example output

IS% / CVR bid adjustments: 14 items

Keyword	Current bid	Top-of-search IS%	CVR WoW	Recommended action
amino acid shampoo	¥95	32%	+5%	+25% raise
refill shampoo	¥110	78%	-18%	-15% lower
…(12 more)…

8 raises / 6 reductions. Reply “execute” to apply.

Details: Top-of-search Impression share optimization

🔒 Safety guards

• Change magnitude is auto-clamped at +30% / -20%
• Cannot run outside a finalized week (Sun–Sat) — a precondition for WoW comparison
• Nothing is applied unless you explicitly say “execute”
• Each suggestion has a 1-hour validity window
• Rollback can be restored from bid history

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot	Available with just Picaro connection
Phase 2 (Q3 2026)	Daily execution	Monthly KPI target registered + saved-prompt enabled
Phase 3 (Q4 2026)	Auto-execute (approval-flow)	Slack 1-click approval, weekly
Phase 4 (2027)	Fully automated	ML confidence ≥ 0.85, within guardrails

👉 Next steps

• Adjust budgets — Check daily budget capacity when raising exposure
• ACoS-based bid reduction — Combine with the ACoS axis for two-stage optimization

---

Prompt 6: Auto-tune placement multipliers from CVR ratio

🎯 Purpose

Compare CVR per placement (top-of-search, product pages, etc.) against the campaign average and propose multipliers that lean into the higher-performing placements.

📋 When to use

• You want to raise the top-of-search multiplier
• You want to weight by placement-level CVR
• Your product-page efficiency is poor and you want to lower its multiplier

📝 Prompt

Using recent daily data, compare CVR per campaign × placement against the overall
average and propose placement multiplier changes.
Use min_clicks {{MIN_CLICKS}}+ and skip anything whose recommended-vs-current difference
is under {{SKIP_THRESHOLD}}.

Placeholders:

Placeholder	Description	Default source
{{MIN_CLICKS}}	Minimum clicks to qualify (e.g., 10)	Default 10
{{SKIP_THRESHOLD}}	Skip threshold for small differences (e.g., 5%)	Default 5%

⚙️ What the AI does

1. Aggregate CVR by placement — Compute CVR per placement on daily data
   • Why daily data: Placement-level metrics require daily granularity
2. Compare against campaign average — Lean into placements with above-average CVR
3. Cutoff — min_clicks 10+
   • Why min_clicks 10: Low-click placements have noisy CVR — this prevents misclassification
4. Skip small differences — Skip if recommended-vs-current diff < 5%
   • Why skip <5%: Sub-5% changes are within learning-noise range and frequent micro-adjustments cause whipsaw
5. Multiplier range — Compute within 0–900%

📊 Example output

Placement multiplier adjustments: 6 items

Campaign	Placement	Current multiplier	Suggested multiplier	CVR vs campaign avg
SP - amino acid shampoo	Top of Search	0%	50%	+62%
SP - amino acid shampoo	Product Pages	30%	0%	-45%
…(4 more)…

Skipped 3 suggestions whose diff was under 5%. Reply “execute” to apply.

🔒 Safety guards

• Multipliers are auto-clamped within 0–900%
• Suggestions with diff under 5% are auto-skipped (noise filter)
• Nothing is applied unless you explicitly say “execute”
• Each suggestion has a 1-hour validity window
• Rollback is possible by setting 0% via Prompt 7

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot	Available with just Picaro connection
Phase 2 (Q3 2026)	Daily execution	Monthly KPI target registered + saved-prompt enabled
Phase 3 (Q4 2026)	Auto-execute (approval-flow)	Slack 1-click approval, weekly
Phase 4 (2027)	Fully automated	ML confidence ≥ 0.85, within guardrails

👉 Next steps

• Manual placement multiplier — When you want to set a specific multiplier directly
• Top-of-search IS% and CVR bid optimization — Combine with bid-level adjustments

---

Prompt 7: Manual placement multiplier / rollback

🎯 Purpose

Directly set the placement multiplier to an arbitrary value for one campaign × one placement (including rollbacks of past adjustments).

📋 When to use

• You want to reset the top-of-search multiplier to 0%
• You want to roll back a past placement adjustment
• You want to manually set a multiplier on a specific campaign

📝 Prompt

For account_id {{ACCOUNT_ID}}, campaign_id {{CAMPAIGN_ID}}, set the placement multiplier
of "{{PLACEMENT_TYPE}}" to {{PERCENTAGE}}%.
(Use 0 for rollback, 1 for testing.)

Placeholders:

Placeholder	Description	Default source
{{ACCOUNT_ID}}	Target account ID	User input (selectable from default accounts)
{{CAMPAIGN_ID}}	Target campaign ID	User input
{{PLACEMENT_TYPE}}	Placement display name or internal name (e.g., Top of Search / PLACEMENT_TOP)	User input
{{PERCENTAGE}}	Multiplier to set (integer 0–900)	User input

⚙️ What the AI does

1. Input validation — Confirm percentage is an integer in 0–900
   • Why this range: Amazon Ads defines placement multipliers as 0–900%
2. Placement-name resolution — Accept either display name (Top of Search) or internal name (PLACEMENT_TOP)
3. Scope limit — Only 1 campaign × 1 placement per execution
   • Why 1 item only: Keeps impact small enough that no approval step is needed and direct execution is safe
4. Direct execution — Apply immediately without a suggestion step

📊 Example output

Done.

Field	Value
account_id	acc_XXXXXX
campaign_id	cmp_XXXXXX
Placement	Top of Search
Multiplier (before)	50%
Multiplier (after)	0%

To roll back, run the same prompt with 50.

🔒 Safety guards

• This prompt is applied directly (no approval step)
• Scope is limited to 1 campaign × 1 placement (multiple items require separate prompts)
• Out-of-range multipliers (outside 0–900%) are rejected
• The previous value is logged — you can roll back by running the same prompt with that value

🚦 Execution mode

Phase	Status	Conditions
Phase 1 (current)	One-shot (direct execution)	Available with just Picaro connection
Phase 2 (Q3 2026)	One-shot (unchanged)	Small scope — not a target for automation
Phase 3 (Q4 2026)	One-shot (unchanged)	Same as above
Phase 4 (2027)	One-shot (unchanged)	Maintained as a rollback tool

👉 Next steps

• Auto-tune placement multipliers from CVR ratio — When you want to re-evaluate the value
• Support / feedback — For unexpected rollback needs

---

Prompt quick reference

What you want to do	Prompt to use
Back-solve keyword bids from ACoS target	Prompt 1
Unified keywords + ASINs + auto, with weighting	Prompt 2
Carve zero-sales targets into negative candidates	Prompt 3
Follow Amazon’s Suggested bid (new campaigns etc.)	Prompt 4
Adjust bids by IS% / CVR	Prompt 5
Auto-tune placement multipliers by CVR ratio	Prompt 6
Set placement multiplier manually / roll back	Prompt 7

Related categories

• Set monthly KPI targets — Pre-register target ACoS so you don’t have to specify it each time
• Add negative keywords — Where the candidates from Prompt 3 get executed
• Adjust budgets — Combine bid and budget optimization
• Promote auto to manual — Candidate extraction as a precursor to bid adjustment