Reading Data: Let Numbers Decide

Your site is live. Now what? Gut feeling is a terrible product manager. Data isn’t perfect either, but it’s honest — if you ask each tool the question it was built to answer. GeFei’s rule is simple: no weekly data review, no progress. This article explains which tools do what, and the weekly routine that turns raw numbers into a to-do list.

What Each Tool Actually Tells You

Beginners collect dashboards like trophies. You need four tools, and each answers one question only:

Google Search Console (GSC): “What does Google think of my site?”

GSC is the only source of truth for your search performance: which queries show your pages, your average position, impressions, and click-through rate (CTR). Two things to remember: it only covers Google search (not direct traffic, not social), and the data lags by a day or two. It’s a rear-view mirror, not a dashboard.

GA4 / Microsoft Clarity: “What do visitors do on my site?”

Once someone lands, GSC goes blind. GA4 shows where visitors come from, how long they stay, and how many pages they view — in near real time. Clarity adds heatmaps and session recordings: you can literally watch where people click, scroll, and give up. This is how you find broken UX that statistics alone hide.

Similarweb / Semrush: “What are competitors getting?”

These are estimates, not measurements. They scrape, sample, and model. Use them to compare magnitudes and trends — “this competitor is roughly 5x my size and growing” — and never to compare exact numbers across tools. If GSC says you got 1,000 clicks and Semrush says a competitor got 800 visits, you cannot conclude you’re bigger. Different instruments, different units.

The Weekly GSC Routine

Open GSC once a week. Not daily — daily fluctuations are noise. Here’s the drill:

1. Find queries with high impressions but low CTR, ranked 5–15.

This is the highest-leverage thing a small site can do. Google is already showing you for these queries — the demand and the ranking are half-built. If your title and description don’t win the click, fix them: match the search intent more precisely, add the key detail the top results show, make the promise clearer. You don’t need new pages; you need better packaging for pages Google already trusts.

2. Look at new queries you didn’t target.

GSC’s query report constantly surfaces phrases you never planned for. Each one is a question from a real person. Decide: does an existing page deserve to rank for this? Strengthen it — add a section that answers exactly that query. Or is it a genuinely separate topic? Then it earns its own page. Either way, don’t let free demand go unanswered.

3. Compare time periods.

Use the date comparison view. Traffic fell last week? Was it a ranking drop on your top pages, or fewer impressions on one query? Traffic rose? Which page drove it, and can you replicate that? Without before/after, you’re guessing. The comparison tells you whether something changed — and changes are what you manage.

Reading On-Site Behavior

Traffic without engagement is vanity. Watch three metrics:

• Engagement/bounce rate. If 80% of visitors leave within seconds, your page didn’t answer their question — or it loaded slowly, or it looks like spam. Go watch a Clarity recording and see exactly where they leave.
• Dwell time. Short visits on a long guide mean they didn’t find what they needed (or the page is hard to read). Long visits on a tool page mean people actually use it. Context matters: a landing page and a blog post shouldn’t share the same bar.
• Pages per session. More than one page per visit means your internal links are doing their job. If everyone reads one page and leaves, add “what next” paths: related tools, deeper guides, the next step in the workflow.

With AI search and AI coding, you can go further: pipe these numbers into your AI assistant and ask it to spot anomalies, or use AI to A/B generate new titles and meta descriptions to test in the weekly cycle.

Competitor Analysis, Done Right

When you research competitors with Similarweb or Semrush, filter out brand keywords first. A competitor’s traffic to searches for their own brand name is not traffic you can compete for. What you want is their non-brand queries: those reveal the real demand they serve and the pages that earn it. Then check GSC for your own coverage of the same topics.

Always Split by Country

This is the lesson people learn too late. One example from GeFei’s community: a site with 100,000 monthly visitors earned under $300 — because 95% of the traffic came from India, where advertising rates are a fraction of US or European rates. Traffic is not revenue. Traffic from the right countries is revenue.

Split everything — queries, pages, revenue — by country. A page with half the traffic but ten times the value per visitor is the better page. This country-level thinking will matter again when you monetize (see the next article).

End the Week With a Review Table

GeFei’s discipline: every weekly review ends with a short table and two answers:

This week	Last week	Change	Why
Clicks, top queries, top pages, key countries	…	…	…

Then the honest sentence: “What does the data say?” followed by three concrete actions for next week — no more than three. Examples: “rewrite the title of page X for query Y,” “build a page for the new query Z,” “fix the slow load on the tool page.”

Data doesn’t decide for you. It tells you where to look, and your three actions say what you’ll do about it. Repeat weekly, and the compounding is real: each week the decisions get sharper because the baseline gets clearer.

---

Series: Build Overseas Websites with GeFei | Previous: Promotion and Backlinks | Next: Monetizing Traffic: Ads or Paid Subscriptions

Based on GeFei’s methodology (哥飞): https://github.com/qiayue/opengefei