Documentation
Advanced analysis
Six Beta reports that put statistics on your Search Console data: CTR curves, opportunity scores, entropy, decomposition, change points and anomalies.
Advanced analysis is the Advanced section of a property's sidebar: six reports that apply a statistical method to your Search Console data. Three work on queries and ask which ones earn too few clicks, which deserve work first and which spread their impressions over several pages. Three work on the daily series and separate trend from weekly rhythm, date lasting shifts and flag unusual days. All six are marked Beta and run only when you press Run Analysis.
Which report answers which question
| Report | The question it answers | Data it reads | Minimum |
|---|---|---|---|
| CTR Curve Modeling | Which queries get fewer clicks than their position earns on your own site? | Queries | 10+ impressions, positions 1-20 |
| Opportunity Scoring | Which queries just behind the top results are worth the most effort? | Queries | 10+ impressions, positions 3-30 |
| Cannibalization Entropy | Which queries share their impressions between several of your pages? | Query and page pairs | 2+ pages, 10+ impressions |
| Time-Series Decomposition | How much of the movement is trend, and how much is the day of the week? | Daily totals | 14 days |
| Change Point Detection | On which dates did traffic settle at a new level? | Daily totals | 14 days |
| Anomaly Detection | Which single days fell outside their normal range? | Daily totals | 21 days |
The Advanced entry at the top of the section opens an index of the six, titled Experimental Analytics, with the method and the data each one needs.
How to run an analysis
- Open the property and pick the report under Advanced in the sidebar.
- Choose the date range and any filters: query, page, country, device, search appearance, search type, and branded or non-branded traffic. Branded terms are the ones you list under Property Setup.
- Press Run Analysis. Once results are on screen the button reads Re-run Analysis; press it again after you change the range or the filters.
Controls that only change how the results are read apply instantly, without fetching again: the entropy threshold, the chart size in Opportunity Scoring, the metric in the three daily reports and the sensitivity in Change Point Detection and Anomaly Detection. Switching the model in CTR Curve Modeling fits the curve again.
How much data a run uses
- Query reports read up to 25,000 rows for the period, most clicked first. That matches the most the Search Analytics API returns in one request.
- Daily reports always analyse at least 90 days. If your range is shorter, it is extended backwards and a note above the results gives the dates actually used. A day without Search Console data counts as zero, so a property younger than the window starts with a run of empty days.
What these reports cannot tell you
- Rows by query leave out anonymised queries, the rare searches Google withholds for privacy, so the query reports see less than the property total. Google explains this in its deep dive on performance data.
- Position is Search Console's average of the topmost position your site held in each search, which blurs a query that moves around.
- A flag is a pattern in the numbers, not a cause. Check the query or page itself, your annotations and what changed on the site before acting.
- Beta means the methods can still change. Each report's page here describes the current version.
Exporting and sharing results
Export in the header saves the main table as CSV, as it stands under the active tab and search box, and each chart has its own PNG button. Save & share report and Send to Slack in the toolbar behave as they do on every report: see Exporting and Sharing.
How this ties in with the content engine
For articles AutoPSEO writes and manages, the content engine already watches the same data each week: it queues a refresh when an article's clicks drop by 40% or more, and an optimisation when it ranks between 8 and 20 but is clicked under 60% as often as that position usually is. See keeping articles fresh. Use the advanced reports for the rest of your site, or when you want to see the numbers behind a decision.
In this section
CTR Curve Modeling: expected CTR for each position
Fits a click-through curve to your own queries, flags the ones clicked less than their position predicts and estimates the clicks they leave behind.
Opportunity Scoring: rank queries by untapped clicks
Scores queries at positions 3 to 30 by impressions, missed clicks and distance from the top, then estimates the clicks a move to position 3 could add.
Cannibalization Entropy: find queries split across pages
Measures how evenly each query's impressions are shared between your pages, in bits of Shannon entropy, and bands the results from Healthy to Critical.
Time-Series Decomposition: trend, weekly pattern and noise
Splits daily clicks or impressions into trend, weekday pattern and residual, labels the direction of the trend and flags days the model cannot explain.
Change Point Detection: date the shifts in your traffic
Finds the dates where daily clicks or impressions moved to a new level, using CUSUM and binary segmentation, with the size and confidence of each shift.
Anomaly Detection: spot unusual days in Search Console data
Compares each day with the same weekday around it using robust statistics, flags spikes and drops by severity and adds an EWMA chart for slow drifts.