No one passing a shop, passers-by not entering, and visitors unable to find an item are different problems. For a page, ask whether it appeared in search, whether anyone clicked, and whether visitors could finish the timer. One visit count cannot explain where things went wrong.
Do this one thing now
Write four lines: what I actually observed (for example, a visitor arrived but could not find Start) → what is still unknown → one specific change → how to recheck. Without data, interpret the example instead of inventing numbers.
Here is a completed example
Fictional observation: One tester asks, “Is this round icon Start?” Unknown: How many visitors came from search, or whether others had the same problem. One change: Add “Start timer” beside the icon. Check: Ask a willing tester to try again and record whether they find Start. This is a teaching example, not an actual user test.
Continue with your situation
I do not have a website
Identify that the example’s problem happens after arrival. Find one potentially confusing phrase in your draft, label your concern as a hypothesis, and rewrite it. You may ask someone to read it, but do not write “a user said” unless that happened.
I have a website
Only with Search Console permission, inspect your page’s search performance using a stated page and date range. Impressions are recorded result appearances under the report’s rules, not people. Clicks are clicks from search results to the site; they do not prove the timer finished. Check successful use separately with validated site records or real observations. Start with the clearest supported problem.
Keep it on your own paper or in your own document. No upload is needed. Mark work you have not done “not checked”; do not guess.
How do I know I am done?
With a site: explain one change using real evidence and record the change and recheck. Without a site: identify the example’s problem, rewrite one phrase in your draft, and label your concern as a hypothesis. State unknowns honestly; do not claim user feedback without a real reading session.
What can wait?
With no search data, do not rush to judge rankings. Check usability and real feedback first. Without a site, practice on the draft without installing analytics.
Open when you need more
How can I avoid misreading existing data?
Compare the same page, search type, country, device, and other filters across complete equal-length periods, noting other changes. Low impressions can have several causes; do not immediately blame the title. With too little data, keep observing without promising a deadline for results. Click-through rate is clicks divided by impressions in the same scope; it is not a universal score.
Complete work checklist, technical guides, templates, and sources
After the small exercise, consult this for your real project as needed. It retains the full scope of the original six-stage workflow.
Editorial sources reviewed · Not tested end-to-end on a live site
Entry requirement
Enter with a live page, a question, and available evidence. Without search data, start with technical checks and reader feedback.
Goal
Complete one documented improvement loop using real data and reader feedback, preserving uncertainty when data are insufficient.
Concepts first
- Impressions count search appearances; clicks count result clicks; CTR is clicks divided by impressions. Average position is an aggregate, not one fixed rank seen by everyone.
- A conversion is a predefined valuable action, such as completing a tool task or submitting an inquiry. Traffic, conversions, and revenue are not interchangeable.
- Missing data are not zero, and correlation is not causation. Seasonality, query mix, country, device, result formats, and updates may change outcomes.
SERP
The full results page, which can include organic results, ads, and other features.
Do not confuse: An ad is not an organic ranking; one screenshot is not everyone’s result.
See examples and evidence →Indexing
The engine processes a page and stores information in its search index.
Do not confuse: Being indexed does not guarantee appearance for a particular query.
See examples and evidence →Ranking
How the engine selects and orders results for a query.
Do not confuse: There is no permanent site-wide rank across all queries and regions.
See examples and evidence →Impressions
Your link received a recorded search-result impression under that result type’s rules.
Do not confuse: Impressions are neither unique people nor clicks.
See examples and evidence →Clicks
Recorded clicks from search results to your site.
Do not confuse: Clicks are not sessions, unique people, or purchases.
See examples and evidence →CTR
Clicks divided by impressions within the same reporting scope.
Do not confuse: No single good CTR applies to all positions, queries, and result formats.
See examples and evidence →Average position
GSC’s mean recorded position across impressions; property aggregation uses the topmost result.
Do not confuse: An average of 5 does not mean everyone saw you fifth.
See examples and evidence →Usage / conversion
A visitor completes a valuable action you defined in advance.
Do not confuse: A page visit is not automatically a conversion; a GA4 key event is not automatically revenue.
See examples and evidence →Page title / title link
Your title describes the page; Google may generate the result’s title from multiple sources.
Do not confuse: HTML title, visible H1, and search-result title are distinct.
See examples and evidence →Meta description / snippet
You can provide a description; Google may instead build the snippet from page content.
Do not confuse: Providing a description does not guarantee it appears verbatim.
See examples and evidence →Inputs
- Page URL, release/change log, authorized Search Console data, and optional existing analytics or user feedback.
- One primary outcome and an observation window chosen in advance; new sites can first check technical readiness and real reader feedback.
Actions
06.1Save a comparable baseline
In Search Console Performance, choose a page, search type, filters, and complete date range. Record clicks, impressions, CTR, position, and queries with export time and missing fields. A new page may have no baseline; do not manufacture one.
Output: A dated, filtered baseline or an explicit no-baseline statement.
06.2Diagnose in layers
Check page availability, then index status, then impressions/clicks, then task completion. Index problems go to stage 4; indexed pages with little visibility prompt stages 1–2; impressions with few clicks call for query, position, and title review; visits without completion prompt stage 3. Avoid conclusions from tiny samples.
Output: One evidence-based diagnosis rather than a generic “SEO failed” label.
06.3Make one interpretable change
Choose a specific hypothesis, such as a missing input example blocking completion. Write the change, expected observation, rollback point, and review date, then publish within authorization. Avoid changing everything and claiming causality. Fix urgent defects promptly and log them.
Output: An improvement card and deployment verification.
06.4Compare and choose the next loop
After the observation window, compare complete equal-length periods using the same filters and note concurrent changes and seasonality. If data are sparse, state insufficient evidence, revise observation, or seek reader feedback. Report observations without guaranteed-growth claims. Decide to keep, roll back, observe further, or return to a prior stage.
Output: Review card: observations, limitations, decision, owner, and next step.
Tools
- Google Search Console plus optional existing analytics, user interviews, and version history. Paid rank tracking is not required.
Tasks to delegate within this stage
Source reviewed · Live site NOT_RUN
Source reviewed · Live site NOT_RUN
Deliverables
- A baseline, improvement card, and review with source, time range, filters, and unknowns for every metric.
Copy this stage’s blank template →
Acceptance criteria
- You can explain the change, observation, limits, and next owner/trigger. A truthful, well-reasoned loop is complete even without growth.
Stop conditions
- Do not access private reports without permission, draw impact conclusions from incomplete data, or continue changes that harm usability or privacy; use the rollback plan.
Next step
Return to the relevant stage with evidence; every iteration need not restart all six stages.
Official sources
- Performance report: Search results ↗
Clicks, impressions, CTR, position, and filters. · 2026-10-10 - Get started with Search Console ↗
Site-owner reports for monitoring search performance. · 2026-10-10 - URL Inspection tool ↗
Stored index information, live tests, and property access. · 2026-10-10
Stage structure, checklists, output templates, and diagnostic sequence are original First Crawl teaching design, not a Google-certified curriculum or ranking formula.
I understand this step. See how AI can help.
Optional. Check the required input and expected result first. Installing a Skill does not complete the work or authorize publishing, messaging, or account access.
- Inspect Google Search Console
Read performance, indexed URL state, and indexing reasons after verifying property, dates, and filters.
- Evidence-led iteration
Compare compatible observations, preserve missing baselines and choose one evidence-backed next step.
The complete packages are source-reviewed; live-site execution is NOT_RUN. Private reports require existing access. Complete instruction packages are available in Chinese and English.
Start the next round where needed
Do not restart everything: return to lesson 1 for unclear needs, 2 for page layout, 3 for usability, 4 for access or indexing questions, and 5 for missing entry points.