9 SEO Metrics Explained: What to Fix First
For analytics leads: how to read 9 SEO metric groups by device and country, separate tracking repair from page edits, and prepare a Guangsuan consultation.
Brand promotion: this article introduces Guangsuan's SEO measurement guidance for analytics leads.
Short answer for a redesign lead: Each of the nine groups answers a separate question, so a change in one total proves only that total. Lock one finished view — same property, same search type, same dates, same query set, same page, same country and same device — then check capture, then eligibility, then wording and contact handling. For method detail, see Guangsuan's 9 SEO metrics guide.
Who this starter note helps
You run weekly reporting and the leadership question is simple: rebuild the page or repair measurement first. This note is for that moment. It does not replace your warehouse or your property settings. It gives you a repeatable order so a dashboard argument does not turn into a blind rewrite.
Guangsuan's work covers technical repair, content supply and link resources. Onsite work runs from diagnosis to fix to recheck. For whole-site scope, see Guangsuan Google SEO service. Bring one finished export and a contact owner when you ask for help.
Set one fixed view first
Use the same site view and the same completed period for every number. Leave the unfinished last day out. Write down language, country, device, brand-named against other queries, and the landing page for each export you keep. Keep brand-named queries apart from other queries and read phone apart from desktop. Differences in time zone, coverage, consent choice, tag firing and canonical credit can pull Search Console and site analytics apart. When a total moves, list the pages, countries and devices behind it before you judge copy. Keep a missing cell marked as missing.
Group A: how the page was shown and chosen
Search Console records shows and picks inside Google Search. Site analytics records what could be seen after arrival. A Search Console pick is not the same as a site session. If Console holds data while analytics sits near zero, inspect tags, consent handling, streams and filters before you judge demand.
Mean position is a show-weighted summary for the view you picked, not one live rank every searcher saw. Outside keyword tallies rest on their own lists, markets and dates. Sort demand by product, use case and buying question, then ask whether the ranking address type fits the job.
Take 20 clicks from 1,000 impressions as a classroom arithmetic example only: the division gives 2%, not a target for every intent. Brand, broad research and buying jobs behave differently. Added shows from lower positions can drag a blended share down while total picks still rise. Match intent, position band, device, market and page before you judge a title, and describe only material truly on the page.
Group B: what happened after arrival
A chat tap, finished form, file download, basket move and paid order are separate outcomes. Test the live event, then keep trials, repeats, spam and business-confirmed contacts on separate lines. Say which ratio you used, because visits holding an event divided by visits differs from events divided by visits when one visit holds several events. Query lines as a rule cannot be tied one by one to a single order.
In GA4, bounce rate means the share of sessions that were not engaged, the reverse of engagement rate. It is not simply the old single-page-session idea. Default engagement includes lasting more than 10 seconds, holding a key event, or at least two page or screen views; check the property's actual configuration. A reader can be satisfied after one page. Repeated view hits from a setup fault can inflate engagement, so verify capture first.
GA4 focus time logs time with the page in front. Finding a specification on one page is not worse than needing four pages. Longer stays can mean careful study or confusion. Read time together with later pages and whether the task was finished. Add the missing unit, condition, worked check or next step instead of adding sections by quota.
Group C: whether the page could appear and what supports it
Keep an explicit list of canonical addresses meant for discovery and work important entries first. Login areas, parameter copies and pages held out by choice stay outside that list. Group the indexing report by reason and open samples. Asking for indexing starts a review and does not force entry. Repair only the block proven on the path you checked, such as accidental noindex, blocking, wrong canonical target, unavailable page or absent rendered material. Use canonical only for duplicate or highly similar material.
Total links and referring sites are different counts. Coverage, refresh timing and de-duplication vary by crawler, so one absent row does not prove a live link vanished. The Links report in Search Console cannot stand in for a live indexing check of each source page. Where outside help is used, match source, target, anchor, attribute, availability and check time against the delivery list. Going live, entering the index and later ranking improvement are three separate stages.
Vendor marks use different models and cannot be swapped. They are not a public score assigned by Google. Compare close peers inside the same tool and time window while still reading real pages, real links and search outcomes.
Device against country: which slice leads
Run the device filter, then the country filter, on identical property and dates. Look at top queries and landing pages for phone and desktop separately, and keep shows and picks on separate lines, because a thin row can swing on one pick. Shows count displays on opened pages; they are not total market volume.
If desktop carries confirmed contacts while phone carries early research, guard the desktop enquiry path first and keep phone readable, rather than rewriting all copy. If one country carries shows and another carries confirmed contacts, fix page type, offer detail, language and contact route for the converting country first. If phone forms lose people, walk the phone path, test the live event, and check consent, load behaviour and form length before you edit desktop copy.
Repair tracking or edit the page
- When Search data holds level but site capture drops toward zero, rebuild confidence in tags, consent, streams and filters before judging demand.
- When shows climb, blended share slips and picks still climb, inspect the wider query and position mix and match narrow slices before touching titles.
- When picks climb but business-confirmed contacts do not, inspect intent, market, phone form ease and qualification rules, then walk the product path with real records.
- When a key canonical stays outside the index, inspect access, signals, matter and discovery path, repair the proven block, then note date, filters, owner and open items.
Log dates, filters, what changed, who owns it and what stays open. This paragraph uses no client case; longer-time examples here are teaching notes.
What to pack before a consultation
Bring facts another person can repeat: site address, target country and language stated separately, affected addresses, one finished export with filters, what changed and when, who can approve code or copy, and how your team qualifies a contact. Test forms with a live note and keep trials and spam apart from true demand.
Quick answers
Must a quiet week trigger a rewrite? No. A flat line means the address was rarely shown for that demand. Confirm eligibility, entry links and whether the address type fits what results favour first.
Must a slipping blended share trigger a title edit? Not alone. See whether new lower-position queries widened the mix, then match narrow slices before editing.
Does longer focus prove success? No. Extended focus can reflect careful reading or uncertainty. Supply the missing condition or next step and check whether the task finished.
Get the weekly brief.
One long-read, every Friday. The dashboards other analysts won't show you, the post-mortems the vendors hope you skip.