Keyword Harvester

Either works, and both is better. The bulk file from Bulk operations already contains a search term sheet, and it also brings your existing keywords, your negatives, and the campaign IDs the export needs — so on its own it does almost everything. The Search Term report from Measurement & Reporting adds one thing the bulk sheet lacks: the split between sales of the advertised SKU and sales of other SKUs. Drop them together and the tool works out which is which.

Because a single order tells you almost nothing. At a 7% conversion rate, a term with fifteen clicks has a better than two-in-three chance of producing at least one sale even if it is entirely average — and plenty of below-average terms will manage one by luck alone. Two orders is a low bar that nonetheless removes most of the noise. It is the difference between a keyword you chose and a keyword that happened to you.

Because ACOS is a bid problem, and you are about to set a new bid. A search term converting well at 90% ACOS inside a broad campaign is usually a term being bought at the wrong price, not a bad term — and moving it to exact with its own calculated bid is the whole point of harvesting. Filter it out and you have discarded a keyword for a fault you were in the middle of fixing. The tool flags high ACOS so you price it carefully, then leaves it in the list.

Amazon reports sales attributed to the exact SKU you advertised separately from sales of other products the shopper bought after clicking. A term where most sales came from other SKUs has not proved that it matches the product you were advertising — it proved the shopper liked your catalogue. Only the standalone Search Term report carries those columns, so if you loaded only the bulk file the column is blank and the criterion passes everything rather than silently failing it.

Because harvesting is two changes. Once a term runs as exact in a performance campaign, the auto or broad campaign that discovered it will keep matching the same query — so you bid against yourself, pay twice to reach one shopper, and split the term's data across two rows that both look worse than the truth. The export adds a negative exact in the source ad group only, never account-wide, which stops the competition without blocking the term. Untick the box if you would rather do it by hand.

Yes, and they are often your best rows. Shoppers paste ASINs into search, and auto and product-targeting campaigns match against them. The important part is that an ASIN cannot be harvested as keyword text — Amazon will accept the row and it will never serve. This tool detects ASIN-shaped terms and writes them as product targeting expressions instead, which is the entity that actually works.

RPC × target ACOS. Where the term has enough clicks behind it to trust its own revenue per click, that is used; otherwise the destination ad group's figure is, which is the lowest-risk of the standard starting-bid methods because it is grounded in performance you already have. Where a term's converting clicks came unusually cheap, the bid is flagged — a lucky auction price on a broad campaign is not something to build an exact keyword on.

Yes. It contains Create rows for the new keywords and product targets in the ad group you picked, plus the matching source negatives, with IDs written as text so 15-digit values survive Excel. Review it first as you should with any bulk upload — particularly the negatives, since those affect campaigns you did not otherwise touch.

No. Both files are read by JavaScript in your browser and discarded when you close the tab. There is no server and no account. Load the page, disconnect from the internet, and everything still works.