GA4 Reports Every Singapore Business Owner Should Actually Check
There are more than forty places to look in Google Analytics 4 and roughly six that will change a decision you make this month. Most business owners we meet have worked this out the hard way: they log in, land on a page showing a wiggly line labelled “Users”, stare at it for ninety seconds, conclude that things are broadly fine, and close the tab. Six months later nobody can say which channel is actually producing enquiries.
The problem is not that GA4 is hard. It is that GA4 is built for analysts and presented to owners, so the reports that answer an owner’s questions sit next to thirty that answer nobody’s. This guide is the short list — the reports worth a weekly twenty minutes, the ones worth a monthly hour, the settings that quietly corrupt everything you are looking at, and the metrics that are safe to ignore entirely.
It assumes your property is correctly installed. If you are not sure, start with setting up GA4 for a Singapore business first, because reading a broken property carefully is worse than not reading it at all. This post sits under our performance marketing guide for Singapore.
Ask the question first, then open the report
The single habit that separates owners who get value from GA4 from owners who do not: decide the question before you log in. GA4 will happily show you a number for anything. It will not tell you which number matters.
There are really only five questions a Singapore SME needs its analytics to answer each month.
| Your question | The report that answers it | The metric to read |
|---|---|---|
| Where did people come from? | Acquisition → Traffic acquisition | Sessions, by Session default channel group |
| Which pages brought them in? | Engagement → Landing page | Sessions and key events, by landing page |
| Did anything useful happen? | Engagement → Key events | Event count for your two or three real key events |
| Which channel produced the useful things? | Traffic acquisition, with key events as the column | Key events per channel, not sessions per channel |
| Is anyone finding us on Maps? | Business Profile collection | Calls, direction requests, website clicks |
Notice that four of the five are about where enquiries came from, not how many people visited. Traffic is an input. If you only have twenty minutes, spend eighteen of them on the columns to the right of “Sessions”.
1. Traffic acquisition — the one report to keep open
Acquisition → Traffic acquisition groups your sessions into channels: Organic Search, Paid Search, Direct, Referral, Organic Social, Paid Social, Email and so on. It is the closest thing GA4 has to a scoreboard.
Two things to do with it that most people skip. First, change the primary dimension from Session default channel group to Session source / medium when a channel looks strange — “Referral” is a bucket, and it matters enormously whether the referrals are from a directory listing or from your own payment gateway bouncing users back. Second, and more important, add key events as a column and sort by that instead of by sessions. A channel sending 4,000 sessions and two enquiries is worse than one sending 300 sessions and eleven.
The new AI Assistant channel, and why your Direct traffic may be shrinking
On 13 May 2026 Google added an AI Assistant channel to the GA4 default channel group. It identifies traffic from “chatbots like ChatGPT, Gemini, and Claude”, assigning the medium ai-assistant and the campaign name (ai-assistant).
This matters for two reasons. Before May, that traffic was landing in Referral or, more often, Direct — so a chunk of the mysterious Direct growth many sites saw through 2025 has now been given a name. And going forward it is a channel you can actually watch. For most Singapore SMEs the volume is still small, but it is the fastest-growing line in a lot of the accounts we look at, and it converts unusually well because someone who arrives via an assistant has typically already been given a recommendation.
Do not hand-tag your own links with utm_medium=ai-assistant to inflate it. You will corrupt a channel you want to trust. See UTM tags done properly for what to tag and what to leave alone.
Source Group: Facebook, Instagram and TikTok stop being six lines each
On 11 June 2026 Google added a Source Group dimension that consolidates the many source values those platforms generate — facebook, fb, m.facebook.com, l.facebook.com, facebook.com and friends collapse into one row. If you have ever tried to total up Facebook traffic by hand across six near-identical lines, switch your dimension to Source Group and stop.
2. Landing page — where the money actually enters
Engagement → Landing page tells you the first page of each session. It is the most under-used report in GA4 and usually the most actionable, because it maps directly onto work you can commission: this page pulls people in, that one does not.
Read it with key events as the second column. A pattern we see constantly in Singapore service businesses: the homepage takes the largest share of sessions and converts poorly, while a single specific service page — the one that matches a high-intent search — takes a tenth of the traffic and produces half the enquiries. That is not a reason to fix the homepage. It is a reason to build three more pages like the one that works.
If a landing page shows sessions but zero key events over a full month, it is either the wrong audience or a broken form. Both are worth twenty minutes of checking.
3. Key events — but only two or three of them
GA4 renamed conversions to key events, and the rename was overdue: it stopped people treating “scrolled 90% of a page” as a sale. The report lists every event you have flagged as important, with counts.
The failure mode here is not too few key events. It is too many. When twelve things are key events, the total is meaningless and every channel looks like it converts. Mark the two or three actions that genuinely represent a person raising their hand — a submitted enquiry form, a completed booking, a purchase — and demote the rest to ordinary events. You can still see them; you just stop pretending they are revenue.
A Singapore-specific wrinkle: clicks on a WhatsApp link or a tel: number are the most common enquiry route for local SMEs and are only a proxy for an enquiry, because GA4 sees the click leaving your site and nothing after it. Count them, label them honestly, and read how to track phone calls and WhatsApp leads for what can and cannot be known.
4. The Business Profile collection — new in June 2026, and a big deal locally
On 8 June 2026 Google added native Google Business Profile linking to GA4. Once the profile is connected in Admin, a Business Profile collection appears in the Reports menu carrying seven metrics: interactions, calls, bookings, directions, website clicks, messages and menu views.
For a Singapore business with a physical location — a clinic, a restaurant, a showroom, a service counter — this closes a gap that existed for years. Plenty of your demand never touches your website at all: someone searches, sees the Maps panel, and taps Call or Directions. That activity used to live in a separate Business Profile dashboard that nobody opened. Now it sits beside your web data.
It requires no tags or tracking-code changes; the link is made entirely in GA4 settings. Two caveats worth knowing before you build a report on it. GA4 retains these metrics on a rolling six-month window regardless of the date range you select, so it is not a source of year-on-year comparison. And the metrics are profile actions, not website sessions — do not add them to your session totals.
If those numbers are thin, the fix is upstream of analytics: see Google Business Profile optimisation and local SEO in Singapore.
5. Pages and screens — for content, not for decisions
Engagement → Pages and screens ranks every page by views. It is useful for one narrow job: finding out which of your blog posts or service pages hold attention, using average engagement time rather than views.
It is not a decision report, because it counts every page anyone touched at any point in a session, so your thank-you page and your contact page will always rank highly without telling you anything. Use Landing page for acquisition questions and this one for content questions.
6. User acquisition vs traffic acquisition — the difference that trips everyone
GA4 offers both and they are not interchangeable. User acquisition attributes to the source that first brought a person to your site, ever. Traffic acquisition attributes each session to the source that brought that session.
The practical consequence: a customer who found you through organic search in March, then returned in July via an email link and bought, appears under Organic Search in user acquisition and under Email in traffic acquisition. Neither is wrong. But if you are asking “what should I spend next month on?”, traffic acquisition is the honest one, because it reflects the touch that actually preceded the sale. If you are asking “what builds our audience?”, user acquisition is more useful.
Pick one for your monthly reporting and stay with it. Switching between them mid-argument is how marketing meetings go sideways. Our guide to attribution models covers why both numbers can be true at once.
The settings that quietly corrupt every report above
You can read all six reports perfectly and still be wrong, because four configuration settings change what the reports contain. These are worth checking once a quarter.
| Setting | What it silently does | What to do |
|---|---|---|
| Data retention | Event data is kept 2 or 14 months (2 is the default on many properties). Beyond that, explorations and funnel reports have nothing to draw on. | Set to 14 months. Note it does not affect the standard aggregated reports — those keep working. |
| Reporting attribution model | Changing it “applies to historical and future data” — it rewrites the past. Last month’s report will no longer match the PDF you sent. | Change it deliberately, once, and note the date you did. |
| Key event lookback window | 30 days by default for acquisition events, 90 days for everything else. Changes apply going forward only. | Leave it unless you have a genuinely long sales cycle. |
| Internal traffic filter | Unfiltered office and agency traffic inflates sessions and, if anyone tests a form, key events too. | Filter your office IP and your developer’s. Check it is set to Active, not Testing. |
The attribution and lookback pair catches people out more than anything else on this list, because the two settings behave in opposite directions: one rewrites history, the other does not. If a report you already circulated no longer reproduces, someone changed the attribution model.
What you can safely ignore
Owners waste more time on these than on anything in the list above.
- Real-time. Genuinely useful for exactly one thing: confirming a tag fires after you change it. It tells you nothing about the business.
- Bounce rate. GA4 defines it as the inverse of engagement rate, so it is the same number twice. Read engagement rate and ignore its mirror image.
- Users, to two decimal places. GA4’s user counts are modelled and de-duplicated imperfectly across devices and browsers. Treat the trend as real and the precise figure as approximate.
- Average session duration, in isolation. A long session can mean engrossed or lost.
- Comparing GA4 to the ad platforms. They use different attribution rules and different windows, so they will never agree. That is expected, not a bug — conversion tracking in Singapore explains why one enquiry can legitimately be counted four times.
A twenty-minute monthly review that actually works
Here is the routine we would give an owner who wants to stay close to the numbers without becoming an analyst. Set the date range to last month, with the comparison set to the previous period.
- Traffic acquisition, sorted by key events. Which three channels produced enquiries? Are they the three you are paying for? (4 minutes)
- Landing page, sorted by key events. Which pages did the work? Any page with traffic and zero key events? (4 minutes)
- Key events. Did total enquiries go up or down, and by roughly how much? (2 minutes)
- Business Profile collection. Calls and direction requests — up or down? (2 minutes)
- One thing that looks wrong. Chase exactly one anomaly to its cause. Not three. (8 minutes)
That last step is the one that compounds. Most GA4 skill is accumulated by investigating single strange numbers — a referral source you do not recognise, a page that suddenly stopped converting, a spike on one Tuesday. Do it once a month for a year and you will know your own data better than most agencies do.
What this routine deliberately does not include is a judgement about whether the money is working. Sessions and key events tell you what happened; they do not tell you whether it was worth it. For that you need cost alongside outcome, which is the job of ROAS versus ROI and of a proper monthly report — see how to read your marketing report. You can see how we present this to clients in our case studies.
One more report, once a quarter
Everything above is a standard report. Once a quarter it is worth opening Explore and building a single funnel exploration: landing page → key page → enquiry. It takes about fifteen minutes the first time and shows you the step where people leave, which no standard report will.
This is also the moment the data retention setting bites. If retention is on the two-month default, your quarterly exploration has nothing to look at. Fix the setting first, then wait — it is not retroactive either.
Frequently asked questions
How often should I actually look at GA4?
Monthly for decisions, weekly at most for monitoring. Traffic is noisy at SME volumes, and looking daily mostly produces false alarms — a quiet Tuesday is a quiet Tuesday, not a crisis. The exception is the fortnight after you launch a campaign or change a key page, when a weekly check catches breakage early.
Why don’t my GA4 numbers match Google Ads or Meta?
Because they are counting different things by different rules. Google Ads and Meta each attribute conversions using their own model and their own lookback window, and each can only see its own clicks; GA4 attributes sessions using its model across all channels. Meta’s reported window is now shorter than it once was, and Google Ads defaults to a 30-day click window for Search. Expect disagreement of tens of percent and reconcile trends rather than totals.
Is the AI Assistant channel worth watching yet?
Watch it, do not budget against it. In the Singapore accounts we see it is still a small share of sessions, but it is growing quickly and it tends to arrive with high intent. The useful thing right now is that it is finally separated out, so you can tell whether it is real for your business rather than guessing.
Do I need the Google Business Profile link if I already check the Business Profile dashboard?
You do not need it, but the link costs nothing and puts calls and direction requests next to your web data, which is where you will actually see them. Remember the six-month rolling retention — if you want longer history, keep exporting it or store it in your own dashboard.
Should I use standard reports or build custom ones?
Standard reports, for at least the first year. Custom reports and explorations are powerful and are also the most common way to produce a confident-looking number that is subtly wrong. Learn what the standard reports say first; build custom views when a specific recurring question genuinely is not answered.
My data looks wrong since a change was made. Can I recover it?
It depends which change. A reporting attribution model change is retroactive, so switching back restores the old view. A time zone change, a lookback window change and data retention are all forward-only — the historical data cannot be regenerated. This is why the settings table above is worth a quarterly glance.
Where to start
If you take one thing from this: add key events as a column in Traffic acquisition and sort by it. That single change reframes analytics from a traffic report into a business report, and it takes about four seconds.
If your reports are not telling you anything useful, the cause is usually upstream — key events that are not really key, a missing filter, or measurement that never captured the WhatsApp and phone enquiries that make up most of your pipeline. We work on that with clients as part of performance marketing: getting the measurement honest first, then making decisions from it. If you would like a second pair of eyes on your property, get in touch and we will tell you what we would fix first.
Written by Adrian Tan, SDM. Last updated 31 July 2026. Sources: Google Analytics 4 release notes (support.google.com/analytics/answer/9164320), GA4 default channel group documentation (answer 9756891), GA4 data retention (answer 7667196), GA4 attribution settings (answers 10596866 and 10597962).
Related measurement guides
- Build a marketing dashboard in Looker Studio (now Data Studio)
- Work out your customer acquisition cost
Related measurement guides
- Vanity metrics in Singapore marketing — which of these reports belong in front of a decision-maker, and which belong in an appendix.
- PDPA-compliant marketing tracking — the Singapore consent rules behind your GA4 configuration.
- AI SEO tools in 2026 — what the paid trackers add once GA4 has told you all it can about AI traffic.
