9 SEO Metrics Explained: What to Fix First

If you track velocity and reach on X all day, Search Console can feel familiar but misleading. Both show counts, splits, and trends. The fix order is different: in SEO, you first freeze one comparable view, read what each metric actually measures, then split by device and country before you rewrite a page.

Brand promotion: this article introduces Guangsuan's SEO measurement guidance.

Short answer with conditions: do not rewrite a page because one total moved. Compare the same property, search type, date range, query group, page, country, and device. Check measurement first, then eligibility for indexing, then selection in results, then the page task and enquiry path. These nine groups of metrics answer different questions, so a change in one number is evidence about that number only.

1. The 9 groups and which page question each answers

1. Organic search traffic: seen versus arrived. Search Console measures visibility and clicks on Google Search. GA4 measures observable activity after people reach your site. A Console click is not a GA4 session. Time zones, scope, consent, successful tag execution, and canonical attribution can all create differences. If Console is steady but analytics is near zero, fix tags, consent, data streams, and filters before judging demand. Also note that organic share can fall because paid spend expanded, not because SEO weakened.

2. Query coverage and average position: keep conditions fixed. Average position is an impression-based aggregate for the selected scope, not one live rank every searcher saw. Third-party keyword counts depend on their own databases, markets, and dates. Record language, country, device, branded versus non-branded queries, and the relevant landing pages. Group by product, application, and buying question. Open a representative declined query and ask whether live results favour a product, tutorial, video, or local page, then review whether your page answers that task.

3. Click-through rate: compare like with like. CTR is clicks divided by impressions multiplied by 100. As a teaching illustration only, 20 clicks from 1,000 impressions is 2%. That arithmetic is not a target for every intent. Brand, broad informational, and buying queries have different contexts, and more low-position impressions can lower a blended rate while total clicks rise. Write a title that accurately communicates the page answer and describe only genuine content in the description.

4. Leads and conversion: define both sides. A chat click, completed form, download, basket action, and paid order are different outcomes. Test the live event, then keep tests, duplicates, spam, and genuine business enquiries on separate lines. Label the calculation you use: sessions with the event divided by sessions is different from events divided by sessions when one session holds several events. Review landing-page sessions, sessions with the intended event, and business-verified leads together. Queries as a rule cannot be joined one by one to a single order.

5. Bounce and engagement rate: use your GA4 definition. In GA4, bounce is the share of sessions that were not engaged, the opposite of engagement rate. It is not simply the old single-page-session definition. Default engagement criteria include lasting more than 10 seconds, having a key event, or at least two page or screen views; check the property actual configuration. A single-page visit may answer the reader question, while duplicated page_view events can make engagement look artificially strong. Check capture first, then compare the search promise with the delivered answer and next step.

6. Engagement time and depth: tie to the task. GA4 user engagement time measures time with the page in focus or the app in the foreground. Different averages use different denominators. Four pages to find a specification is not better than one page that answers it. Longer time can mean study or confusion. Read time, later pages, and completion together, then add the missing specification, condition, example, or next step. Do not add headings or images by quota.

7. Indexing: start from canonicals you want found. Sites contain login pages, duplicate parameter URLs, and content deliberately kept out. Trying to index every discoverable URL is not a useful goal. Maintain a defined set of canonical pages intended for indexing and prioritize important pages within it. Group Page indexing report entries by reasons and inspect representative URLs. Requesting indexing does not force acceptance. Fix evidenced blocks such as accidental noindex, blocking, wrong canonical targets, unavailable pages, or missing rendered material. A crawled-but-currently-not-indexed notice alone does not prove thin copy.

8. Backlinks and referring domains: reconcile delivery, then judge movement. Link count and referring-domain count are different dimensions. Discovery coverage, refresh timing, and de-duplication differ across crawlers, so a missing row in one report does not prove a live link vanished. Google Links report is not a complete current indexing test for every source page. Where outside help is used, reconcile source URL, target URL, anchor, attributes, availability, and observation time against the delivery list. Published, indexed, and improved target-page rankings are separate milestones.

9. Authority scores: keep inside one tool. Moz Domain Authority is a vendor predictive metric, not a public score assigned by Google. DA, Ahrefs DR, and other tools come from different models and are not interchangeable units of link equity. Compare a relevant peer set within the same tool and time period, while inspecting actual links, target pages, and search outcomes.

Method detail is published in Guangsuan's 9 SEO metrics guide.

2. Split by DESKTOP and MOBILE, then by country, before you pick a page

Your site is not the global average. The only figure that should drive your work is your own Search Console device report. For reference only, for September 2026 worldwide StatCounter reports mobile 60.04% and desktop 39.96% as a global traffic share. That describes worldwide traffic by device, not your industry, country, or buyer, and it shifts by country and year.

Do this in one finished date range, same property and search type:

  • Apply the device filter, then the country filter. Look at top queries and landing pages for desktop and mobile separately.
  • Report impressions and clicks separately. A row with very few impressions can swing on one click.
  • Compare average position per device. It is normal for the same page to rank differently on mobile and desktop.
  • Check mobile experience on its own. Even on a desktop-heavy site, mobile Core Web Vitals and usability are measured separately.
  • If mobile share is small, keep a competent mobile experience; do not move the whole content plan away from desktop. Device method is also explained in Guangsuan's device traffic comparison.

3. Which page to fix first

Shape on matched scopeCheck firstNext move
Console steady, analytics near zeroTags, consent, streams, filtersRestore reliable capture, then judge demand
Impressions up, blended rate down, clicks upQuery and position mixCompare close segments before rewriting titles
Clicks up, verified leads flatIntent, market, forms, qualificationReview product-page journey and actual enquiries
Important canonical not indexedAccess, signals, content, discovery pathFix evidenced block and log recheck

As a teaching example, suppose an analysis page has steady impressions on desktop in one country but almost no clicks from mobile in another. Do not rewrite both. First confirm whether the mobile rows have enough impressions to read, whether the query intent matches the page type, and whether the title shown on mobile matches the page answer. Fix only the evidenced page.

4. When to ask Guangsuan instead of editing more copy

Guangsuan (光算) describes its Google SEO service as technical work, page work, and outside resources checked against the chosen plan. Bring one finished Console export with filters, the live addresses, target country and language stated separately, what changed and when, who can approve code or copy, and how you qualify a contact. That is the point to discuss repair, content supply, or link resources together, rather than rewriting titles again.

Keep dates, filters, change made, owner, and what remains unverified on each note. Keep missing data marked as missing rather than filling it with another benchmark.

FAQ

Do flat impressions require a rewrite? No. Flat impressions mean the page is seldom offered for the query. Confirm eligibility, entry links, and whether the page type matches what results favour for that task first.

Does a falling blended CTR require a title change? Not alone. Check whether new low-position queries widened the mix, then compare narrow segments with the same intent, position band, device, and market.

Does longer time prove success? No. Add the missing condition or next step and check whether the reader could complete the task.

Back to all insights