A useful winners-and-losers analysis identifies the pages and queries responsible for a meaningful change, explains which metric moved, and distinguishes a genuine Search change from demand, seasonality, reporting noise, or a tiny baseline.
Do not begin with percentage change. Begin with absolute click difference. A page that rose from one to four clicks gained 300%, but it contributed only three additional visits. A page that rose from 4,000 to 4,500 clicks gained 12.5% and contributed 500 visits.
Define the comparison before opening the report
Write down:
- The property and Search type.
- The current and comparison periods.
- Whether the goal is to explain clicks, visibility, or a specific segment.
- Any page, country, device, brand, or search-appearance filters.
- The minimum baseline needed for a row to be decision-worthy.
Use equal-length, complete periods. Match days of the week and relevant seasonality. For a retailer, November versus October can confuse seasonal demand with SEO change; November versus the same weeks last year is often more informative.
A strong review usually includes both:
- Recent comparison: the last 28 complete days versus the previous 28 complete days.
- Seasonal comparison: the same period year over year.
Google recommends weekly or monthly granularity for long comparisons because it smooths day-of-week effects.
Fast workflow in the Search Console interface
- Open Performance → Search results.
- Select the correct Search type.
- Set the Date control to Compare and choose equivalent periods.
- Turn on Clicks and Impressions. Add CTR and Average position as diagnostic metrics.
- Open the Pages tab.
- Sort the Difference column by clicks to find the largest absolute gains.
- Reverse the sort to find the largest absolute losses.
- Click a page to filter to it, then open Queries.
- Repeat the diagnosis by Country, Device, and Search appearance when the change is concentrated.
- Export when you need more rows, custom thresholds, or repeatable classification.
Google’s Insights report can provide a quick starting list of pages and queries trending up or down. Google orders those lists by the change in click count, not percentage change. Use the full Performance report for the explanation.
Start at the broadest useful level
Use this drill-down sequence:
Property → site section → page → query → country/device/search appearance
If the sitewide click delta is -10,000, identify which directories account for it before studying individual queries. If one directory contributes -8,000, focus there. If losses are spread proportionally across every directory, investigate sitewide demand, technical problems, migrations, manual actions, algorithmic changes, or measurement anomalies.
The interface cannot calculate arbitrary directory groups. Use URL filters for one section at a time, or export page data and assign each URL to a mutually exclusive template or directory.
The four-metric diagnosis
Clicks are the outcome. Impressions, CTR, and average position are clues.
Clicks down, impressions down
Likely questions:
- Did search demand fall seasonally?
- Did important queries disappear or lose visibility?
- Were pages deindexed, canonicalized elsewhere, redirected, or removed?
- Is the decline limited to one country, device, or Search type?
- Did the comparison period contain an event or campaign?
If average position is stable but impressions fall, reduced demand or a change in the query mix is often more plausible than a simple ranking loss. It is still only a hypothesis; Search Console does not report total market demand or competitor impressions.
Clicks down, impressions stable, CTR down
Investigate:
- The queries whose CTR changed.
- Shifts in average position within those queries.
- Search appearance and result-layout changes.
- Titles and snippets as actually shown.
- A mismatch between query intent and landing-page promise.
- Brand versus non-brand mix.
Do not optimize a pagewide average CTR without considering position, query, country, and device. CTR naturally differs across these contexts.
Clicks down, impressions stable, average position worse
This is compatible with a visibility decline, but average position can change because the mix of queries changed. Filter to the affected page, then compare individual queries with enough impressions. Check whether the same query family, country, and device declined.
Average position is the average topmost position for the selected property or page grouping, not a daily rank-tracker reading.
Impressions up, clicks flat
This is not automatically a loss. A page may be appearing for more early-stage or lower-position queries. Determine whether:
- New impressions are at positions unlikely to earn clicks.
- The page expanded into less relevant query families.
- CTR fell only because the visibility mix broadened.
- The new queries indicate content worth improving.
Clicks up, impressions stable
Possible explanations include improved CTR, better positions within existing demand, a more attractive result appearance, or a shift toward higher-click query/device segments. Confirm by query rather than assuming a title change caused it.
Position improves while clicks fall
This apparent contradiction is common when demand or query mix changes. High-volume queries may have lost impressions while smaller queries improved position. Weighted averages can move in a favorable direction even when total clicks decline.
Calculate absolute and relative change correctly
For a metric M:
absolute change = current M - previous M
percentage change = (current M - previous M) / previous M × 100
Percentage change is undefined when the previous value is zero. Label a row new instead of inventing an infinite percentage. Similarly, a row present only in the previous period is lost, but it can also be absent because of privacy filtering or top-row truncation.
For CTR, report the difference in percentage points:
CTR change = current CTR - previous CTR
A move from 2% to 3% is +1 percentage point and +50% relative, two different statements.
Do not average row-level CTR or position without appropriate weights. Recalculate CTR as total clicks divided by total impressions. Search Console’s average position is already impression-weighted and aggregation-sensitive.
Use minimum baselines
Thresholds should fit the site, but every recurring report needs them. Examples:
- At least 100 previous-period impressions before labeling a CTR loser.
- At least 20 previous-period clicks before prioritizing percentage decline.
- At least 50 absolute lost clicks for an editorial review.
- A change sustained across two weekly periods before calling it a trend.
These are workflow examples, not Google rules. The goal is to prevent a long tail of volatile rows from crowding out material changes.
A useful priority score can combine business impact with evidence:
priority = absolute lost clicks × business weight × confidence factor
Keep each component visible. A single opaque score should not replace the underlying metrics.
Diagnose a losing page by query
After selecting a losing page:
- Open Queries.
- Sort by click difference.
- Record the few query families explaining most of the loss.
- Compare impressions, CTR, and position for each.
- Check whether another page from the property now appears for the same query.
- Segment by device and country.
- Inspect the live result and page only after the data identifies a focused hypothesis.
If many related queries declined together, evaluate the page’s overall relevance, freshness, intent match, internal linking, and competitive landscape. If only one query declined, avoid rewriting the whole page before understanding that query’s result set.
Diagnose a winning page responsibly
Wins deserve the same skepticism as losses. Ask:
- Is the increase sustained or a one-day spike?
- Did impressions grow because search demand changed?
- Did a campaign increase branded demand?
- Did one country or device create most of the gain?
- Is the page newly indexed, newly canonical, or replacing another URL?
- Did the query family change?
- Does GA4 show useful landing-page engagement or key events?
Search Console can associate timing and Search metrics with a change; it cannot prove causality. Google explicitly notes that news, user sentiment, competitors, and other events can contribute to performance changes.
Important aggregation and data limitations
Chart versus Pages table
The chart is property-aggregated. When you group by Page, the table is page-aggregated. Multiple pages from the same property can therefore make table totals differ from the chart. Use page rows to allocate change, not to reconstruct the property total without accounting for aggregation.
Canonical URL attribution
Most Performance data is assigned to Google’s selected canonical URL. A migration or canonical change can make one URL look like a winner and another like a loser even when underlying content demand is similar. Check URL Inspection and redirect history.
Hidden and truncated queries
Some queries are omitted for privacy, and only top rows are exposed. Query-filtered totals also exclude anonymized queries. Your visible query losers will not always explain every click in the chart.
Preliminary data
The newest data can change while Google finishes processing it. Use complete, finalized periods for formal reporting.
Search Console anomalies
Before diagnosing an isolated date, check Google’s data anomalies page. A logging or processing issue can create a chart movement without a real Search change.
Export workflow for a stronger analysis
Export both periods or use the Search Analytics API when you need to:
- Join rows that appear in only one period.
- Apply minimum thresholds.
- Classify URLs by section or template.
- Rank by absolute clicks and percentage change.
- Calculate cumulative contribution to the sitewide delta.
- Store a repeatable snapshot.
The API returns top rows and does not guarantee every row. Its request limit is up to 25,000 rows per call, with offsets available, but internal Search Console limits still apply. BigQuery bulk export is the better source when a large site needs the most complete available performance dataset.
A practical exported table includes:
| Field | Purpose |
|---|---|
| Page or query key | Stable row identifier |
| Current and previous clicks | Outcome and absolute delta |
| Current and previous impressions | Demand/visibility clue |
| Current and previous CTR | Click-behavior clue |
| Current and previous position | Visibility clue |
| Absolute click delta | Materiality ranking |
| Percentage delta | Scale context |
| Section/template | Ownership and pattern detection |
| Device/country | Concentration evidence |
| Annotation | Releases, migrations, campaigns, incidents |
Questions the analysis can and cannot answer
It can answer:
- Which reported pages or queries gained or lost the most clicks?
- Which segments account for most of the reported change?
- Did reported impressions, CTR, or average position move alongside clicks?
- When did the change begin?
It cannot prove by itself:
- Why Google’s ranking systems changed.
- Whether total market search demand changed.
- Which competitor gained the lost visibility.
- Whether page changes caused the outcome.
- Whether the additional clicks produced revenue or conversions.
Use result inspection, release logs, GA4, competitive data, and technical diagnostics for those questions.
Review checklist
- Equal, complete periods are compared.
- Search type and active filters are documented.
- Absolute click change is prioritized before percentage change.
- Tiny baselines are excluded or labeled.
- Pages are drilled into queries and relevant segments.
- Canonical and migration changes are considered.
- Chart/table aggregation differences are respected.
- Hidden queries and top-row limits are disclosed.
- Known Google reporting anomalies are checked.
- Findings are labeled as evidence, hypothesis, or confirmed cause.