A service firm's website has two jobs, to be found and to get the enquiry. Four numbers describe them. Search Console clicks tell you how many people arrived from a search, the query list tells you what they typed, GA4 key events tell you how many of them submitted the form or clicked the phone number, and the channel report tells you where everyone else came from. Every other figure in those two interfaces is instrumentation. It is real, and it does not belong at the top of your reading order. Once a month, ten minutes, in that order.
The four numbers that describe the two jobs
Google's help pages define these terms precisely, and the definitions are narrower than they look. Read the vendor definition, then ask what it can tell a firm of ten people selling a service.
- Search Console clicks. Times someone came through to your site from Google results. Source: Search Console.
- Queries. The searches your pages appeared for, which is the demand side of the same report. Source: Search Console.
- Key events. Actions you marked as important, the form submission and the phone number click among them. Source: GA4.
- Channels. How GA4 grouped arrivals by where they came from, with Direct usually the largest single line. Source: GA4.
Exhibit 1
Four numbers cover a service firm's website, and each has one source
Two of these come from Google's search index and two come from the tag on your own site, so they answer different questions.
Three ways GA4 counts the same visit
GA4 reports three counts for one stream of traffic, and they are easy to swap in a report. Total users is the number of unique users who triggered any event in the date range you picked. Active users is the narrower number: the unique users who had at least one engaged session, and it is the only user count GA4 shows in reporting. A session is a period during which a user is engaged with your site, and it ends after 30 minutes of inactivity by default. Views counts page views, and GA4 adds to it each time the same page loads.
Take one person who reads your homepage and the pricing page, then leaves. That is counted once as a user. It is counted once as a session. The two pages are counted as two views.
Pick a date range before you compare any of them. GA4 also estimates session counts, so ignore small gaps between two reports.
Exhibit 2
One afternoon of browsing fills all three GA4 counts at once
Engaged sessions and why GA4 bounce rate is not the old number
A session counts as engaged if it lasts longer than 10 seconds. It also counts as engaged if it includes a key event, and again if it includes two or more page views. Any one of those is enough. Google's own pages are not identical at the boundary, since the engagement page says longer than 10 seconds while the user metrics page says 10 seconds or longer.
Engagement rate is the percentage of sessions that were engaged, and bounce rate is the rest of them. The two add up to 100%, so pick one and stop reading the other. We read engagement rate, since it states the positive.
GA4's bounce rate measures something different from the number your old reports meant. A long read of a single page can be an engaged session. A person who reads your homepage top to bottom and then phones you never viewed a second page. No page Google publishes gives a target value for engagement rate on a small site, so do not chase one.
Key events are the number closest to a finished enquiry
In GA4 the metric used to be called conversions. Google's help pages now say that events that matter to your business are called "key events." A key event is any action you mark as important, such as a form submission or a phone number click, and it is the figure that means "this website did its job". Today Google reserves the word conversion for Google Ads reporting, so a report labelled "conversions" is the Ads view of your key events, and it measures nothing new.
Nothing in GA4 matters until you mark at least one key event. An unmarked site records visits with no outcome attached, and the reports fill up without telling you anything about enquiries. Mark the form submission first. If people phone, mark the phone number click. Mark a booking link too if you use one. Whether your homepage earns its form placement is a separate question, covered in what a small business homepage needs to say.
Where traffic came from, and what Direct really collects
GA4 sorts arrivals into channel categories. Organic Search means non-ad links in search results, and Google now counts AI Overviews and AI Mode traffic in that channel, one reason your organic line can move with your site unchanged (how a small business shows up in AI answers). Referral is a non-ad link on someone else's site, so a directory listing or a local news mention shows up there. Direct is a mixed bucket. It collects people who arrive from a saved link or by typing your URL, and it also absorbs visits whose source tags went missing.
A large Direct line on a small firm points to unmeasured arrivals more often than to typed addresses. Links in email lose their tags, and so do links inside a PDF attachment you send out.
What Search Console counts: clicks, impressions, CTR and position
Impressions counts times your link appeared in Google results, and clicks counts times someone came through to your site. Click-through rate is clicks divided by impressions. Position is a relative ranking where 1 is the topmost spot, reported as an average over the searches where your site appeared.
The average hides a spread of positions. It is the topmost spot your site held, averaged into a summary, and searching for yourself will disagree with it. Google's own page says so, because search history and location change what you see. Watch the direction over months. A dash in the position column means no recorded position.
The query list is the most valuable page in either tool for a small firm, since it prints the words people used when your page appeared. What to do about it, and which advice to ignore, is in SEO for a small service business.
Why GA4 and Search Console never agree
Search Console and GA4 will not agree on how many visits came from Google, and Google documents the causes. The two use different URL definitions. They use different time zones. Analytics needs JavaScript on the page, and Search Console does not, so a page with no Analytics tag still counts as a click in Search Console while GA4 never sees it. Google lists the causes on its own page and publishes no percentage, so nobody can tell you how wide your gap should be.
Read search demand in Search Console. Read what people did after they arrived in GA4. Do not average the two.
Missing rows and the setting that erases last year
On a small site GA4 sometimes withholds a row entirely, because the count is low enough that a person could be identified. A blank row is not a zero. It means GA4 chose not to show the number. Widening the date range is the first fix, because more users in the window can bring the row back. Google publishes no minimum count that triggers this withholding.
Retention is the setting that decides how far back you can compare. GA4 deletes user-level and event-level data on a monthly cycle once it passes the retention period you set. The free property gives you a choice of 2 months or 14 months, and 14 is the setting that lets you compare this month to the same month last year in Explorations. Open Admin and find the Data Retention control under Data Settings. Google's page lists the choices without naming a default, so your property is set to whatever it was set to. The setting changes Explorations and funnel reports. Your monthly totals in the standard aggregated reports survive, and the detailed comparison does not.
Consent banners and what a cookieless tool gives up
Consent mode is the plumbing that tells Google tags what each visitor chose in your cookie banner, and Google's page states that it does not provide the banner itself. Where a visitor declines, GA4 estimates part of what happened through modeling, so some of what you read is a model's best guess. Google's guidance is to set consent defaults by region, because privacy laws are region-specific, so many firms show a banner to visitors in regions with those laws and skip it for everyone else.
A cookieless tool is the other route. Cloudflare Web Analytics is free, and its own page says it does not use client-side state such as cookies or localStorage to collect usage metrics. Cloudflare's own summary of what it shows is top pages, views, and top referrers. Plausible lists a Starter plan at $9 per month billed monthly for up to 10,000 pageviews a month, with a 30-day free trial and no credit card required, as of October 2026. Both tools trade depth for simplicity, and on a firm that needs the form count, the depth is the whole point.
Three speed numbers with their thresholds
Core Web Vitals are three measurements of how a page feels, and web.dev gives each a target. LCP covers loading, targeted at 2.5 seconds or less. INP covers responsiveness, 200 milliseconds or less. CLS covers layout stability, 0.1 or less.
Google scores these at the 75th percentile of real page loads, so one fast test run is not a passing grade. If your report still shows First Input Delay, it is showing a retired measurement, since INP is the current interactivity metric. These thresholds are about user experience on web.dev's own page, and that page attaches no ranking claim to them, so read a slow LCP as a reason to fix the page.
The monthly ten-minute read
Set the date range first, since every count is defined for the range you picked. Read key events for the month. Read engagement rate and skip bounce rate. Open Search Console and check clicks and impressions for direction. Then skim the query list for anything new, and look at channels last to see whether one source moved.
Exhibit 3
A monthly read takes ten minutes and most of what the tools offer stays out
The monthly read leaves out some things on purpose. It leaves out today, because day to day counts on a small site are noise. It leaves out your own Google search, and any row that came back blank. If key events stay flat while clicks and impressions climb, the problem is on the page, and the website redesign trap covers when a new site will not fix it. When clicks and impressions are flat, the problem is being found, and that is the part our work on SEO for small firms covers.