jaygptpro Search Term Report Miner
A Search Term Report from a mature account is 50,000+ rows. Sellers download it, open Excel, scroll for 20 minutes, find 4 negatives, give up. Meanwhile real money leaks: terms with 30+ clicks and 0 orders, exact-match converters stuck in broad campaigns, and dead clusters that drain $200/day. This skill carves the file into 4 surgical buckets and outputs a Bulk Operations upload ready to submit.
When to use this
- Monthly PPC review and you want a clean negative sweep
- Account spending $3K+/month on PPC with ACoS drift
- New product 60 days post-launch ready for the first harvest pass
- Pre-Q4 cleanup. Push wasted spend out before peak
The framework. The Four Buckets
Pull the Search Term Report (60-day window). For every search term, route to exactly one bucket based on click and order thresholds.
| Bucket | Filter | Action |
|---|---|---|
| 1. Negative Exact | Clicks >= 10, Orders = 0, ACoS = blank | Add as Negative Exact at campaign |
| 2. Negative Phrase | 3+ terms share root, combined clicks >= 25, orders <= 1 | Add root as Negative Phrase |
| 3. Exact Harvest | Orders >= 2, ACoS < target ACoS, term not already exact | Move to new exact-match campaign |
| 4. Hero Scale | Impressions >= 1000, CVR well above your account average, ACoS < target | Increase budget + bid +20% |
Anything else: leave alone.
A note on the Hero gate: there is no universal "good" CVR. It varies hugely by category, price, and product. Do not hardcode a fixed number like 12%. Instead, compare each term to YOUR account or category baseline. A practical hero gate is roughly 1.5x to 2x your average CVR for that product. Compute your baseline first, then set the gate from it.
Step by step
Download the STR. Advertising > Reports > Sponsored Products > Search Term Report. Date range: last 60 days. Format: CSV.
Set the target ACoS. Usually 25-35% depending on margin. This is the threshold for bucket 3 and 4.
Apply Bucket 1 filter. Clicks >= 10 AND Orders = 0. Export the search term column. Format as Negative Exact rows for Bulk Operations.
Cluster for Bucket 2. Group remaining 0-1 order terms by shared 2-word root. If a root cluster has 25+ combined clicks, add the root as Negative Phrase.
Apply Bucket 3 filter. Orders >= 2 AND ACoS < target. Check each term is not already an exact keyword in any campaign. Format as new exact-match campaign rows.
Apply Bucket 4 filter. Impressions >= 1000 AND CVR well above your account average (roughly 1.5-2x your baseline for that product) AND ACoS < target. These are scale candidates. Generate bid +20% rows.
Build the Bulk Operations file. 4 sheets, one per bucket. Use Amazon Bulk Sheet template columns.
Run the quality check, then upload to Campaign Manager.
Output format
## STR Mining Output. [Account] [date range]
**Summary**
- Total search terms analyzed: [N]
- Bucket 1 (Negative Exact): [N] terms, wasted spend last 60d: $[X]
- Bucket 2 (Negative Phrase): [N] roots
- Bucket 3 (Exact Harvest): [N] terms, projected exact-match orders/month: [N]
- Bucket 4 (Hero Scale): [N] terms
**Bulk Operations file structure**
Sheet 1: Negative Exacts
Campaign | Ad Group | Negative Keyword | Match Type
[rows]
Sheet 2: Negative Phrases
[rows]
Sheet 3: Exact Match Harvest. New campaigns
Campaign | Ad Group | Keyword | Match Type | Bid
[rows]
Sheet 4: Hero Scale. Bid increases
Campaign | Ad Group | Keyword | Old Bid | New Bid
[rows]
Worked example (illustrative)
The figures below are an example to show how the buckets shake out, not a benchmark to expect. Your term counts, CVR, ROAS, and savings depend entirely on your account.
Account: home goods seller, 42 SKUs, $8,400/month PPC spend. 60-day STR has 38,200 unique search terms. After parsing:
Bucket 1: 184 negative exact candidates. Wasted spend = $1,847 over 60 days, or $924/month immediately saved.
Bucket 2: 12 phrase roots like "cat", "for dogs", "wholesale". 412 clicks, 8 orders. Adding negative phrases blocks 200+ wasted clicks/month.
Bucket 3: 67 exact-match harvest candidates. In this example the harvested terms convert well above the account average and move to exact campaigns at 1.2x bid. The incremental-orders and ROAS figures you would model here are account-specific. do not assume a fixed order count or a particular ROAS. run the projection off your own term-level CVR and AOV.
Bucket 4: 14 hero terms. +20% bid aims to move them from mid page 1 toward the top of search.
Total projected impact in this scenario combines the negative-keyword savings with the harvested-term upside. Both are illustrative. compute your own numbers from your account data rather than carrying these over.
Quality check
- Date range is 60 days, not 30 (need volume for cluster math)
- Target ACoS set before bucketing, not after
- Bucket 3 harvest terms confirmed not already exact in any campaign
- Negative phrases checked for accidental brand cannibalization
- Bulk file uses correct column headers per Amazon template
Common mistakes
- 30-day STR. Not enough volume per term. Bucket 2 cluster math breaks down.
- No target ACoS. Bucket 3 and 4 lose their gate, you scale unprofitable terms.
- Adding negative phrases without checking brand terms. Easy to accidentally negative-match your own brand variant.
- Harvesting terms that already exist as exact. Creates internal bid wars.
- Skipping Bucket 4. Most sellers focus on cutting waste and miss the scale upside, which is often the larger lever.
Built by Jay GPT Pro
Part of Amazon Pro Skills. Production-grade skills for serious Amazon sellers. Free and open. Built by Jay Margaliot.
I share a new AI play for Amazon sellers every week, free, in my WhatsApp group. Join here: https://chat.whatsapp.com/ILX65p1yWcaIG3c9WGHpTY