Reading Results
The SEO Testing results dashboard, the per-test query performance panel, how to read a win or loss, and creating a revert test.
Once tests start completing, RankNest has two levels of results to read. The Results dashboard rolls every completed test for a client into one view covering what you've gained in total, which variables win most often, and which pages and queries moved. The individual test tells the detailed story: its before/after metrics and the Query Performance panel showing exactly which search queries were kept, gained, or lost.
This page covers both, plus what to do when a test loses.
Where to find it
Open the client's SEO Testing view and click Results in the toggle at the top right (the other side is Tests). Everything on the dashboard is built from completed tests only. Running and draft tests contribute nothing until they finish. Until the client has its first completed test, the dashboard shows a "No completed tests yet" message with a link back to the tests workspace.
Per-test results live in the experiment editor. Open a test from the Experiments list on the page detail pane, described in Managing Tests.
Cumulative impact
The strip at the top of the dashboard totals everything the client's completed tests have produced:
- Total Tests Completed, with the number of wins underneath.
- Total Clicks Gained: the sum of click changes across completed tests. It can be negative if losing tests outweigh winners.
- Total Impressions Gained: the same sum for impressions.
- Avg Winner CTR Lift: the average CTR improvement across the tests that won, with the overall win rate as the subtitle.
The count of completed tests sits to the right of the section heading, so you always know how big the sample behind these numbers is.
Win Rate Over Time
Win Rate Over Time charts the percentage of completed tests that improved CTR, ranking, or clicks, bucketed by month over the last twelve months. Above the chart is the latest month's win rate and its change from the previous month that had completed tests, in percentage points.
The bar rail under the line shows how many tests completed in each month. Read the two together: a 100% win rate on a month with one completed test is not the same finding as 60% across fifteen. Hover a bar to see that month's test count and win rate.
Variables Leaderboard
The leaderboard answers "what kind of change actually works on this site". Each row is a test variable (Title, Description, H1, and so on, including any custom variables you created) with:
| Column | What it shows |
|---|---|
| Variable | The variable name, with a Best badge on the strongest performer |
| Tests | How many completed tests used it |
| Wins | Win rate, with the raw "N of N" underneath |
| Avg CTR Lift | Average CTR change across those tests |
| Avg Rank Change | Average ranking change; positive means the page moved up |
Click any column header to sort by it, and click again to flip the direction. The Best badge only appears on a variable with enough completed tests to be meaningful, so a single lucky test never claims it.
Top Winners and Top Losers
These two lists rank individual completed tests by overall improvement, combining CTR, ranking, and click changes. Each row shows the test's title, the page path, and its headline metric with a secondary metric underneath.
Click any row to open that test in the experiment editor, where you can read the control and variant, the full before/after numbers, and the notes you wrote.
Read the losers list as carefully as the winners. A title rewrite that dropped clicks tells you something specific about how this audience searches, and it's the list you'll pull from when creating revert tests (below).
Query Movement (all tests)
Query Movement aggregates the query-level data from every completed test. Biggest Winners lists the non-brand search queries that gained the most clicks across those tests; Biggest Losers lists those that lost the most. Each row shows the query, its cumulative Clicks Δ, its Position Δ, and how many tests touched it.
Brand queries are excluded here on purpose. Someone searching the client's company name was going to click through regardless of what the title tag says, so leaving brand terms in would flatter every test.
Like What Wins on This Site, this card stays closed until the client has a handful of completed tests, and says so until then.
What Wins on This Site
The final card breaks win rate down by page attribute rather than by variable: Page Type, Word Count band, and Title Length band.
The page type here is the kind of page RankNest identified when it crawled the URL: home, blog, service, product, location, company, contact, and so on. It is not the link map's page type; that classification lives in Page Types & Clusters.
Two rules keep it honest, and both mean the card stays quiet early on:
- The whole section appears only after the client has a handful of completed tests. Until then it says so instead of charting noise.
- Within it, a bucket only appears once enough tests fall into it. A group with no qualifying bucket is hidden entirely.
Use it to aim your next round of tests: if short titles on Service pages win two-thirds of the time and long-form Supporting pages never do, test the pattern that's working.
Query Performance on a single test
Open any test and scroll to the Query Performance panel below the control and variant metrics. Click the header to expand it. The header itself shows the query count plus quick chips for how many queries are new or lost.
Inside, four summary cards frame the comparison:
- Baseline Queries: queries the page ranked for before the change.
- Results Queries: queries it ranked for during the test window.
- New Queries: queries that appeared only after the change.
- Lost Queries: queries the page ranked for before but not after.
The table below lists every query with clicks, impressions, and position, plus a change column for each once both snapshots exist. New and Lost badges mark queries that only exist on one side.
Controls above the table:
- Filter queries... searches within the list.
- All / Non-brand / Brand sets brand filtering, which defaults to Non-brand so brand searches don't drown out the queries you influenced.
- All / Retained / New / Lost appears once both snapshots exist.
- A Sort dropdown orders by impressions, or by clicks, position, or impressions change once there's a comparison.
Results are paged; use the arrows at the bottom to move through them.
If the panel says it's showing a baseline snapshot only, the test hasn't completed yet. The comparison half arrives when the test ends and its results sync.
How to read a win or a loss
A test counts as a win when the variant period improved CTR, ranking, or clicks without a matching regression elsewhere. In practice, work through it in this order:
- Start with CTR. For title and description tests, CTR is the metric your change can actually move. A rise in CTR with steady impressions is the cleanest possible win.
- Check impressions before you celebrate clicks. Clicks rise when impressions rise, which often means seasonality or a ranking change rather than better copy. If impressions doubled and CTR fell, the page got more exposure and your variant converted less of it.
- Check the rank change. Positive means the page moved up. A large ranking move usually means something outside the test changed too, and the CTR comparison gets much less trustworthy.
- Open Query Performance. Aggregate numbers hide the story. Losing three irrelevant queries while gaining two commercial ones can be a good outcome even when total clicks dipped.
- Write it down. Completed tests have Action and Notes & Takeaways fields in the editor. Fill them in while the reasoning is fresh. In six months these notes are the value.
Note: if Google Analytics 4 is connected, test cards also show a Key Ev. Δ figure. It's the change in GA4 key events on that page over the same window, shown as context only. It is not attributed to the test and never affects whether a test counts as a win.
Warning: a single test is evidence, not proof. Search results move for reasons you don't control: algorithm updates, competitors, seasonality. The leaderboard and win-rate trend exist because a pattern across many tests is far more reliable than any one result.
Restore the original after a losing test
When a test made things worse, the variant is still live on the page. RankNest can spawn a revert test that rolls the page back and measures the recovery, so you have evidence that undoing the change helped rather than just assuming it.
- Find the test. On its card in the Experiments list, click Restore; or open a completed, paused, or cancelled test and click Restore Original in the editor footer. (A running test's footer shows End Experiment instead.)
- Confirm the prompt. RankNest creates a new test with the control and variant swapped: what's live now (the losing variant) becomes the control, and the original version becomes the variant to restore.
- Put the original back on the live site.
- Open the new test and click Start Experiment.
The recovery test runs on a fixed 14-day window and arrives pre-filled with a hypothesis and notes referencing the test it reverts, so the pair stays readable later.
A few limits:
- The card's Restore button is hidden on drafts, since nothing changed on the site yet, and on tests that are themselves reverts.
- Once a revert exists for a test, its Restore button reads Reverted and is disabled, so you can't stack duplicates on one test.
- Recovery is measured against the degraded numbers, not the original baseline. That's the point: it answers "did putting it back help?", not "was the original better than the variant?"
Tips
- Read the leaderboard before planning a quarter of tests. Doubling down on the variable with the best win rate on this site beats testing everything evenly.
- A negative Total Clicks Gained is a prompt to look at the losers list and revert what's still live.
- Query Movement is a keyword research source. Queries the site keeps gaining clicks on are worth their own pages.
- Sort the leaderboard by Tests to see where your evidence is thin. A variable with one test doesn't tell you anything yet.
Troubleshooting
The dashboard says "No completed tests yet". Nothing has finished. End a running test manually, or wait for it to auto-complete after its end date.
A completed test has empty variant metrics. Its results never synced. Open the test and click Sync Results next to Variant Performance to pull the test-window numbers from Search Console.
"Not enough completed tests yet to chart a trend." Win Rate Over Time needs at least one month with a completed test. It fills in as tests finish.
Query Performance is empty on a completed test. Query-level snapshots are captured when Search Console data syncs. If the page had no query data at baseline, there's nothing to compare. See Google Search Console to confirm the property is connected and assigned to this client.
The Restore button is missing on a card. It's hidden on drafts and on tests that are themselves reverts. If it reads Reverted and is greyed out, a revert test already exists for that test.
Related
Last updated 2026-07-25