Most Amazon PPC automation advice is a list of rules with confident-sounding numbers and no explanation of where the numbers came from. That is how sellers end up pausing keywords after eight clicks, or raising budgets on campaigns that were never profitable in the first place.
This is the full set we run — grouped by job, with the threshold logic behind each one, and an honest note on the three categories where Amazon’s native rules cannot do what people assume they can.
First, derive your own thresholds
Two numbers make almost every rule below concrete. Without them you are copying someone else’s account.
Click threshold — how long before you judge a keyword
A keyword that has had five clicks and no sale has told you nothing. The number of clicks you should wait is approximately 100 ÷ your conversion rate percentage. At a 5% conversion rate, that is 20 clicks before a single sale is even statistically expected. Judge earlier and you are pausing keywords for being unlucky.
Break-even ACoS — the line between profit and volume
Break-even ACoS equals your contribution margin percentage. If 30% of the selling price survives COGS, referral and closing fees, fulfilment and a returns allowance, then a 30% ACoS is exactly break-even. Below it you make money on the incremental unit; above it you are buying volume on purpose. We covered the derivation in detail in what is a good ACoS on Amazon.
Rules that scale spend
| Rule | Conditions | Action |
|---|---|---|
| Proven Winner Push | ACoS < target AND orders ≥ 5 (14d) AND impression share < 30% | Bid +15–20% |
| Efficiency Buffer Scale | ACoS < 50% of target AND clicks ≥ 20 (14d) | Bid +10% |
| Budget-Starved Winner | Budget utilisation > 90% by mid-afternoon AND ACoS < target | Budget +20–25% |
| Top-of-Search Opportunity | TOS impression share < 20% AND ACoS < target AND TOS CVR > account average | TOS modifier +25% |
| New Keyword Runway | Keyword age < 7 days AND impressions > 1000 AND clicks < 5 | Bid +10% |
| Event Scale-Up | Scheduled date range (Prime Day, BFCM, festive) | Budget +50–100% |
New Keyword Runway is the odd one out and worth flagging: it is not a performance rule. A keyword with impressions but almost no clicks has not earned a bid increase — you are buying data, and the rule should expire once the keyword has enough clicks to judge normally.
Rules that cut losses
| Rule | Conditions | Action |
|---|---|---|
| Click-Waste Cutoff | Clicks ≥ your threshold AND orders = 0 | Bid −30% |
| ACoS Breach | ACoS > target AND clicks ≥ threshold | Bid −30% |
| Sustained Overspend Trim | ACoS > target for 14 consecutive days | Bid −10% |
| Declining Trend Guard | 7-day ACoS > 30-day ACoS × 1.3 AND spend > ₹200 | Bid −15% |
| High-Spend Zero-Order Kill | Spend > 2× break-even CPC equivalent AND orders = 0 | Pause keyword |
| Dead Keyword Pause | 30–35 clicks with no conversion AND orders = 0 | Pause keyword |
Note the difference between ACoS Breach and Sustained Overspend Trim. The first reacts to a clear breach with a sharp cut; the second shaves gently at campaigns that are persistently a little over. Running a −30% cut on every daily fluctuation will strip a campaign of impression share within a week.
Budget rules, and the trap in them
The second trap is that event budgets never come back down. Amazon’s native budget rules only increase. After Prime Day or BFCM the elevated budget sits there until somebody notices, which is usually after a fortnight of overspending. Either diarise the reset or use a scheduler that reverts automatically at the end of the window.
Negative keyword rules
| Rule | When to fire | Action |
|---|---|---|
| Standard Negative Harvest | 25–30 clicks, zero orders | Negative exact at ad-group level |
| Fast-Reaction Spend Cutoff | Search-term spend ≥ 2× average CPC, zero orders | Negative exact immediately |
| Auto vs Manual Cannibalisation | Term converts in Manual, still active in Auto | Negative exact in the Auto campaign |
| Obvious Irrelevance Block | Wrong category or competitor brand, zero orders | Negative phrase, no click wait |
| Negative List Audit | Negative added more than 90 days ago | Flag for review |
Two of these deliberately skip the click threshold. If a search term is obviously the wrong category, waiting for 25 clicks to prove it is just paying tuition. And in a launch phase where velocity matters more than data, a spend cutoff at twice your average CPC is a reasonable substitute for click count.
The audit rule exists because negatives rot. A term you negated when your price was ₹1,499 and you had eleven reviews may convert perfectly well now. Re-examine the list quarterly.
Day-parting — where native rules stop
Conversion rates are not flat across the day. Most categories show peaks somewhere between 9am–12pm and 7pm–11pm, but that is a starting hypothesis, not a fact about your account — verify it against your own hour-of-day data before scheduling anything.
| Rule | Conditions | Action |
|---|---|---|
| Peak Window Boost | Hour inside your high-CVR window AND ACoS < target | Bid/budget +15–20% for that window |
| Dead-Hour Bid Cut | Hour in a low-CVR window AND CVR < 50% of account average | Bid −30% or pause |
A working day-parting rule also needs a condition, not just a time. “Raise bids 8–11pm” will happily raise bids on campaigns losing money in that window too. The time says when; a condition says whether.
Inventory rules nobody sets up
Advertising a product into a stockout is the most avoidable waste in an Amazon account. When cover drops below roughly three weeks at current sell-through, it is worth throttling ads and letting the remaining units go to organic demand, which costs nothing.
Coming back is equally delicate. After a restock following a long pause, resume at around 70% of the previous bid and climb back over a week. A sudden full-bid restart after weeks of zero delivery tends to perform worse than a gradual ramp.
Keyword harvesting
| Rule | Conditions | Action |
|---|---|---|
| Auto-to-Manual Promotion | Auto search term with orders ≥ 2 AND ACoS < target | Add to Manual Exact at that term’s CPC + 10% |
| Match-Type Funnel | Broad keyword with orders ≥ 3 AND CVR > account average | Promote Broad → Phrase, then Phrase → Exact |
Promotion without isolation causes the cannibalisation problem above: the same term then competes with itself across two campaigns and you bid against yourself. Promote and negate in the same pass, and re-run the isolation check every two to four weeks.
How often should rules run?
| Rule type | Sensible cadence | Why |
|---|---|---|
| Bid rules (increase or decrease) | Daily | A 14-day lookback barely moves in an hour; shorter intervals chase noise |
| Budget pacing | Every 30–60 minutes | A campaign capping out at 2pm has to be caught the same day |
| Day-parting | Every 30 minutes | The schedule must catch hour boundaries and revert at window end |
| Negative harvesting | Weekly | Search-term reports settle slowly; daily churn produces false negatives |
| Negative audit | Quarterly | Listings, prices and reviews change on that timescale |
Roll them out safely
Every rule here can be wrong for your account. The sequence that avoids expensive surprises:
- Set your target ACoS first — per account, overridden per campaign where margins differ. Rules that compare against a target are meaningless without it.
- Run in notify-only for two weeks. Read what the rule would have done. This is where you discover that your click threshold is too low.
- Enable auto-apply on one rule at a time, scoped to a single portfolio, with a cap on any increase so a repeating rule cannot ratchet a budget upward indefinitely.
- Keep an undo trail. Every automated change should be reversible, and you should be able to see who or what made it.
ListingPilot ships these as a named rule library — installed disabled, so you switch on the ones that suit your account rather than inheriting somebody else’s automation. Day-parting schedules apply and revert automatically, including the decreases Amazon’s native tool will not do.