Close Variant
A close variant is a search query Amazon treats as a near-match of your exact or phrase keyword — plurals, misspellings, reorderings, and function-word changes. They expand reach automatically but quietly inflate ACOS when the variant CVR diverges from the canonical keyword's.
A close variant is a search query that Amazon's match engine treats as equivalent to your bid keyword even though the text differs. For an exact-match keyword running shoes, Amazon will match — and bill — the queries running shoe, runing shoes, shoes running, running-shoes, and the running shoes as close variants. The same expansion applies to phrase match.
This is a quiet expansion. The campaign reports show the canonical keyword as the targeting term, but the underlying search terms include every variant Amazon mapped to it. The only way to see them is the search-term report.
What counts as a close variant
| Type | Example (target: running shoes) |
|---|---|
| Plural / singular | running shoe |
| Misspelling | runing shoes, runninng shoes |
| Reordering | shoes running, shoes for running |
| Function word | the running shoes, running the shoes |
| Stem / accent | runnig shoes, course running shoes |
| Hyphenation | running-shoes |
Amazon does not treat synonyms (sneakers, trainers) or category siblings (hiking shoes) as close variants under exact match. Those are separate keywords and require their own targets — or broad match to capture.
Why close variants matter for ACOS
The canonical keyword and its variants almost always have different conversion rates. A typical pattern:
| Search term | Clicks | Orders | CVR | ACOS |
|---|---|---|---|---|
running shoes (canonical) | 412 | 41 | 9.9% | 14% |
running shoe | 38 | 5 | 13.1% | 11% |
shoes running | 27 | 1 | 3.7% | 51% |
the running shoes | 19 | 0 | 0% | ∞ |
The report rolls all of this up as "exact-match running shoes, 12% ACOS, 47 orders" — which hides that the reorderings and function-word variants are dragging the average up. Without close-variant isolation, bid changes target a blended number that doesn't exist for any actual query.
How to control them
- Read the search-term report weekly. Filter to your exact-match targets and group by underlying customer search term. The variants jump out.
- Negative-exact the bad variants. A negative exact on
the running shoesblocks that specific query without touching the canonical keyword. - Promote the good variants. Variants with strong CVR get their own exact-match keyword and bid — they earn dedicated bid headroom instead of inheriting the canonical's bid.
- Don't try to disable close variants entirely. Amazon does not expose a switch. Negatives are the only lever.
Close variant ≠ "close match" in auto-targeting
Two unrelated Amazon concepts share similar names:
| Close variant | Close match (auto-targeting) | |
|---|---|---|
| Where it applies | Exact + phrase manual targeting | Auto campaign only |
| What expands | Plurals, typos, reorderings of your keyword | Queries Amazon judges highly relevant to your ASIN |
| Signal source | Keyword text | ASIN catalogue data |
| Control lever | Negative keywords | Auto-targeting group bids + negatives |
Confusing them leads to the wrong fix — negatives in the wrong campaign type.
Common mistakes
- Optimising the keyword's blended ACOS without looking at variant-level CVR. The blended number always lies.
- Negative-exacting the canonical when one variant is the problem. Use negative-exact on the exact variant text.
- Assuming exact match means exact. Since 2018, Amazon exact match is "exact-ish". Treating it as literal will silently overspend.
- Conflating close variants with close-match auto-targeting. Different campaign type, different lever.