In a keyword combiner, the order setting picks whether list A is treated as the outer loop with list B nested inside it, or the reverse — so the same 10 base terms × 5 modifiers always produce 50 pairs, but in a sequence you can predict, sort, and rerun. Both A-then-B and B-then-A produce the same Cartesian product; the choice changes only the iteration order and therefore the order of the final phrases. Because the loop order is deterministic, switching the toggle from A-then-B to B-then-A is the mechanical equivalent of transposing a matrix — every pair still appears, but the base terms now run down the columns instead of the rows. That predictability is the actual reason an order control exists: it lets you arrange output for human review, for ad-group drafts where grouping by modifier is convenient, or for version comparison between two runs of identical inputs. To control that order directly in your browser, paste your two lists into the Keyword Combiner and use its A-then-B and B-then-A toggle.
What the A-then-B vs B-then-A setting controls
The order toggle in a keyword combiner is not a filter, sorter, or relevance switch — it is a loop selector. The list you place in the outer position determines the major rhythm of the output, while the list in the inner position cycles through completely before the outer list advances. In practice that means a list A of three colors combined with a list B of two product names under A-then-B will iterate as color-1 plus product-1, color-1 plus product-2, color-2 plus product-1, color-2 plus product-2, color-3 plus product-1, color-3 plus product-2. Under B-then-A the same six pairs appear, but the cycle runs product-1 plus color-1, product-1 plus color-2, product-1 plus color-3, product-2 plus color-1, product-2 plus color-2, product-2 plus color-3.
Three properties follow from this mechanical definition. First, the count is identical either way: three × two is six regardless of which list is outer. Second, the phrase content differs between the two orderings: the same two terms produce different strings when the outer list changes, so `red` and `shoes` with a space become `red shoes` under A-then-B and `shoes red` under B-then-A. With an empty separator, distinct pairs can also collapse to the same final string — for example, `ab` plus `c` and `a` plus `bc` both produce `abc` — and the tool keeps the first-seen occurrence. Third, the sequence is reproducible — same input, same toggle, same output text, byte for byte. That third property is what makes the order choice worth thinking about instead of leaving it on the default.
Picking the right order for your keyword workflow
Choosing between A-then-B and B-then-A is essentially a grouping decision. Ask which axis you want the reader of the output list to scan most easily. If you are building an ad group where each base term is its own campaign and the modifier is the targeting layer, pick A-then-B and put the base terms in list A. The output will read like a per-base-term block, which makes it easy to copy a contiguous slice for a single campaign. If you are instead building a modifier-first taxonomy — perhaps a topical map where each color, size, or material is a section heading with every applicable product underneath — pick B-then-A and put the modifiers in list A.
Two more considerations push the choice in one direction or the other. The first is downstream tooling. Some keyword tools and spreadsheets expect a specific column layout; for example, a content brief template that reads "[intent] [topic]" reads better when the intent term is fixed and the topic cycles, which favours placing the larger, more stable list in the outer position. The second is audit ergonomics. When two SEOs need to compare their combined lists, an identical toggle setting on both sides is the only way to keep the output diffable. Choosing A-then-B on one machine and B-then-A on another would force a sort step before any line-by-line comparison could work.
Order is a mechanical property, not a relevance signal. The combiner does not score phrases, attach search volume, estimate competition, or rank candidates, so every group it produces still needs your own review against landing-page fit and any first-party or paid keyword evidence you use elsewhere in the workflow.
Set keyword order in the combiner
- Open the Keyword Combiner in your current browser tab; the combination logic runs locally, so nothing is uploaded while you work.
- Paste one base term or phrase per line into list A, with each line trimmed at the outer edges and empty lines left out of the count.
- Paste one modifier or second term per line into list B, following the same one-per-line rule; exact case-sensitive duplicates inside list B are removed after trimming, keeping the first occurrence.
- Choose A-then-B if you want list A to drive the outer loop, or B-then-A if list B should drive it; the toggle also dictates the order of the phrases in the final output block.
- Enter the exact separator — a space, hyphen, plus sign, empty string, or another short literal of up to 20 Unicode code points with no line break or control character; the separator you pick is treated as a literal with no hidden spacing.
- Run the combination; the tool computes the pair budget against the 10,000-pair ceiling before generation, so a request that would be too large fails the whole run rather than producing a partial sample.
- Review the summary panel — normalized list sizes, theoretical pair count, removed input duplicates, removed output collisions, and unique output count — to confirm the order produced what you expected.
- Copy the deterministic one-per-line output as a single block into your spreadsheet, ad-group draft, content brief, or taxonomy sheet.
How order changes phrase shape and output counts
Order alone does not change the number of combinations. It changes three other things: the visual shape of each phrase, the sequence of phrases in the output, and the set of output collisions that the tool has to deduplicate. The table below shows the same three colors and two products running through both orderings with a single space as the separator.
| Order setting | Iteration sequence | First three phrases | Last three phrases |
|---|---|---|---|
| A-then-B | color then product | red shoes, red hats, blue shoes | blue hats, green shoes, green hats |
| B-then-A | product then color | shoes red, shoes blue, shoes green | hats red, hats blue, hats green |
With a non-empty separator the two orderings contain identical word sets; they differ only in sequence. With an empty separator the phrases can collapse, because concatenation loses its boundary. For example, A-then-B with no separator on `red` and `shoes` becomes `redshoes`, while a different pair like `reds` and `hoes` would also collapse to `redshoes` and trigger an output collision. The combiner keeps the first-seen occurrence and reports the later collision in the summary, so switching order can shift which pair wins the collision. If your lists are short and the separator is empty, run both orderings and compare the summary collision counts to see how stable the result is. For separator choice guidance beyond the order decision, the separator guide walks through space, hyphen, plus sign, and empty-string behaviour in more detail.
Verifying the order with the combiner's summary
The summary panel is the cheapest way to confirm the order toggle did what you wanted. Five numbers matter for an order audit. The normalized list sizes show how many unique case-sensitive terms survived trimming and dedupe on each side. The theoretical input-pair count is simply the product of those two sizes and represents the budget the tool checked against the 10,000-pair ceiling. The removed input duplicates count tells you how many repeated lines were dropped from your pasted lists. The removed output collisions count is the most order-sensitive figure: it tells you how many final phrases collided after concatenation and were deduped, which is the place where A-then-B and B-then-A can produce different numbers if your separator is empty or short.
Use these numbers together. If the unique output count is much smaller than the theoretical input-pair count, the difference is split between input duplicates and output collisions — read both lines of the summary to see which dominated. If the unique output count matches the theoretical count exactly, the toggle change is a pure reordering with no deduplication effect, which is the cleanest case for diffing two runs against each other. If the count surprises you, narrow the source lists before copying, because the tool never silently truncates, paginates, samples, or replaces a too-large result with an ellipsis.
Getting repeatable results across runs
The point of choosing an explicit order is reproducibility. Identical input lists, identical separator, identical toggle, and identical browser tab produce identical text on every run, because the loop order is stable and the output deduplication keeps the first-seen occurrence. That property makes the output suitable for version comparison, regression tests on a keyword workflow, and reproducible content briefs. Store the source lists as plain text — one term per line, no trailing whitespace, no smart quotes that might survive trimming differently on a future paste — and the toggle decision you make today will reproduce tomorrow.
Two practical habits keep the runs aligned. First, decide the toggle once per workflow and document it next to the source lists, so the next person who runs the same combination does not accidentally switch the order and break the diff. Second, when iterating on a list — adding a new modifier, removing an irrelevant base term — rerun the combination and compare the new unique output count against the previous one; a count that matches the previous count exactly means your edit did not introduce any new collisions and the order is still behaving predictably. These two habits, combined with the summary audit, turn the order setting from a checkbox into a reliable part of the keyword pipeline.
For a deeper look, see Keyword Density Checker for Beginners: A Friendly Start.