Conversion Tracking in Singapore: Setting Up GA4, Google Ads and Meta So the Numbers Mean Something

A Singapore aircon servicing company came to us with three dashboards open. Google Ads reported 74 conversions last month; GA4 reported 61; Meta reported 39. The owner had closed 44 jobs. Four numbers, and only the last one paid the staff.

Nothing was broken in the sense of a tag failing to fire. Every platform was measuring exactly what it says it measures. The problem was that nobody had decided what a conversion was, which system was the source of truth, or how long each platform got to claim credit. The business had run for two years on numbers that could not be added, subtracted or compared.

This guide is the fix: what counts as a conversion, how to set up Google Ads conversion tracking, GA4 key events and the Meta dataset properly, the traps that only bite Singapore businesses, and how to stop the same sale being counted three times. It sits under our performance marketing guide for Singapore, and it is the plumbing that makes your monthly marketing report worth reading at all.

Decide what a conversion is before you touch a tag

This is the step almost everyone skips, and it is the reason most tracking setups are quietly useless. A conversion is not “a form submission”. It is the cheapest event you can measure that reliably predicts revenue — and the gap between those two things is where budget goes to die.

An e-commerce brand selling S$80 skincare has a purchase event that is the revenue. Easy. A B2B software firm selling S$60,000 contracts has a “Download the whitepaper” form that correlates with revenue about as well as the weather does. Tell Google’s bidding algorithm to optimise for whitepaper downloads and you will get a magnificent quantity of whitepaper downloads and no pipeline.

Before configuring anything, write down three things:

• The revenue event. The moment money is committed — purchase, signed quote, deposit paid.
• The proxy event. The on-site action closest to it that happens often enough to bid on: a qualified enquiry, a booking, a call over 60 seconds.
• The ratio between them. If 100 enquiries produce 22 jobs at S$1,400 gross profit, one raw enquiry is worth S$308. That single number turns every cost-per-lead argument into arithmetic; our guide to ROAS versus ROI works it through in full.

Everything else — PDF downloads, newsletter signups, scroll depth — gets measured but is not a conversion. It is a diagnostic. Confusing the two is how an account ends up with 74 “conversions” and 44 jobs.

The four systems that will each give you a different number

A single Singapore lead passes through several counting systems, each with its own definition, its own window and its own incentive. Here is what each one is actually for.

That table explains the aircon company’s problem entirely. Google Ads gets 30 days to claim a conversion; Meta gets 7 for a click; GA4 will attribute a key event to a touchpoint 90 days earlier. Nobody is lying — they are answering different questions over different periods, and the answers were never designed to reconcile.

That last point causes a monthly panic in a lot of Singapore agencies. Google Ads Help is explicit that a conversion is credited to the date of the click, not the sale. Last month’s total therefore keeps rising for weeks after month-end: the report you sent on the 3rd is not the report the client sees on the 20th.

Setting up Google Ads conversion tracking properly

This is the tag that matters most, because it is the one Smart Bidding learns from. Get it wrong and the algorithm optimises confidently towards the wrong thing. Our guides to Smart Bidding in Singapore and why Google Ads campaigns stop converting trace most failures back here.

Primary versus secondary: the trap that catches almost everyone

Google Ads splits conversion actions into primary and secondary. Primary actions are “reported in the ‘Conversions’ column” and used for bidding. Secondary actions are tracked “for observation only”, appear in “All conversions”, and are not used for bidding — with one exception, where a secondary action sits inside a custom goal.

Here is the part that catches people. Google’s own documentation states that conversions created from Google Analytics “are secondary by default when created from Google Analytics and can only be updated to primary in Google Ads.” So the very common sequence — set up key events in GA4, import them into Google Ads, switch on Maximise Conversions, wonder why performance is flat — produces an account whose bidding strategy is optimising against nothing you configured. The conversions are visible. They are simply not biddable.

Check this first on any account you inherit: Tools and Settings, then Conversions, then the Goal column. Anything you want bid on must be Primary.

Conversion windows: set them deliberately

Windows are configurable per conversion action, and most people leave the defaults forever:

Two practical rules. Match the window to your real sales cycle: a Singapore renovation firm with a six-week consideration period is throwing away signal on 30 days, while a food delivery promo does not need more than seven. And know that changes are not retroactive — Google states that when you change a conversion window, “any changes apply to all conversions going forward”. Change it once, deliberately, and note the date.

Enhanced conversions

Enhanced conversions supplement the standard tag with hashed first-party data. Google applies “a secure one-way hashing algorithm called SHA256 on your first party customer data, such as email addresses, before sending to Google”, then matches those hashes against its signed-in users to recover conversions cookies would have lost.

Two things worth knowing. If you upload data yourself via API or file, hash only the genuinely personal fields — email, phone, name, street address — because Google’s guidance is explicitly not to hash country, region, city or postcode. And be sceptical of the uplift figures in circulation: a “5% to 15% increase in reported conversions” is quoted widely by vendors, but Google does not publish that number in its own enhanced conversions documentation. Note the word reported — recovering attribution is not creating sales.

The lead-generation variant, enhanced conversions for leads, is the right choice when the sale closes off the website. For most Singapore service companies — law firms, clinics, contractors, B2B consultancies — that is the relevant one, because the on-site event is an enquiry and the revenue happens weeks later in a meeting.

Setting up GA4 key events

GA4 is not your bidding source. It is your referee: the one place that sees organic search, direct, referral, email and both ad platforms in the same table. Its job is to answer “where did this actually come from” — a question no ad platform can answer honestly about itself. The minimum competent setup, in order:

1. One property, correctly configured. Reporting time zone Singapore, currency SGD. Google warns that “changing the time zone only affects data going forward” and that you “may see a flat spot or a spike in your data” if you change it later, so get it right at creation. Our GA4 setup guide for Singapore businesses covers every one of these decisions.
2. One web data stream, with the Measurement ID beginning “G-” installed site-wide.
3. Enhanced measurement on, so scrolls, outbound clicks, site search and file downloads arrive without custom work.
4. Events for the things that matter — enquiry submitted, booking confirmed, call tapped, WhatsApp tapped — named consistently in lower case with underscores.
5. Only the real conversions marked as key events. Marking twelve things as key events is the same as marking none.
6. Google Ads linked, then check on the Google Ads side whether the imported actions need promoting to Primary.
7. Internal traffic excluded. A five-person office reloading its own thank-you page produces a beautiful conversion rate and no money.

One setting worth changing on day one: the key event lookback window. Google’s defaults are 30 days for acquisition key events (first_open and first_visit, switchable to 7) and 90 days for all other key events, with 30 and 60 available. Crucially, “the key event lookback window you choose also applies to session attribution.” Unlike the reporting attribution model, which “applies to historical and future data”, lookback changes “apply going forward” only.

Setting up the Meta dataset and Conversions API

Meta’s browser pixel alone is no longer a serious measurement setup in 2026. Safari’s tracking protection deletes JavaScript-written storage after seven days, and much Singapore traffic arrives on iOS. The pixel is half of a pair; the Conversions API sends the same events server-side. The mechanics, briefly — our Meta pixel setup guide covers the full implementation:

• What used to be called a pixel is now a dataset in Events Manager. Your old pixel ID is the dataset ID; nothing needs renumbering.
• Send each event both ways — browser and server — with a shared event_id so Meta can deduplicate. Without it you double-count your own events.
• Watch Event Match Quality, which scores how much identifying information you attach. Passing an email and phone number on a form submission moves it materially; passing nothing leaves server events barely attributable.
• Meta publishes a 75% coverage target for server events relative to browser events. Below that, the CAPI implementation is missing events.

On the widely-quoted claim that the Conversions API delivers “32.9% additional conversions”: read Meta’s own wording, which is about additional conversions reported. That is attribution recovery, not incremental revenue. Your bank balance does not change because a conversion became visible.

Meta’s 2026 attribution redefinition

If your reported Meta conversions fell in early 2026 and nobody could explain it, this is why. In January 2026 the longer view-through windows were removed. Then on 3 March 2026 Meta redefined click-through attribution to require an actual link click — one that sends the user to a website, app or lead form. Likes, shares, saves, comments, profile taps and short video views, all previously counted as clicks, moved into a new engage-through bucket with a fixed 1-day window that cannot be extended. The engaged-view threshold also dropped from 10 seconds of video to 5.

The default is now 7-day click, 1-day engage-through, 1-day view. Reported conversions fell as a definitional artefact, not a performance decline, and Meta confirmed the change applies to how conversions are classified in reporting. These dates come from practitioner and vendor accounts rather than a fetchable Meta support document, so treat them as well-corroborated rather than officially quoted.

The Singapore-specific problems no global guide covers

This is where imported tracking advice quietly fails, because the standard playbook assumes a customer who buys on a website. A lot of Singapore commerce does not work that way.

Leads that leave the website immediately

The largest measurement hole for Singapore SMEs is WhatsApp. A prospect taps the floating button, the conversation happens in an app, the quote goes out as a PDF and the deal closes on a voice call. Your tag saw a button click. Your bank saw S$4,000. Nothing connects them.

The fixes, in increasing order of effort: fire a distinct event on the tap and treat it as a proxy conversion with a known close rate; append a UTM-derived reference into the pre-filled message so the source survives the jump; or use enhanced conversions for leads to upload the closed deal against hashed contact details. The same logic applies to phone calls.

PDPA and consent

Singapore’s position is not Europe’s, and precision matters. There is no prior-blocking requirement here — you are not legally obliged to withhold all tags until a banner is clicked. But the PDPC’s Advisory Guidelines on Selected Topics (revised May 2024) do require consent for cookies used to target advertising, and browser inaction is not consent. The Legitimate Interests Exception does not cover direct marketing. And DNC obligations apply to marketing messages sent to a Singapore phone number — which is the awkward part of the WhatsApp workaround above: the ad is unregulated, the follow-up message is not. Practically: publish a real cookie notice, get consent for advertising cookies, and make sure whoever built your banner wired it to your tags rather than just displaying it.

One further change deserves a careful flag. Many consent-management vendors have published that from 15 June 2026 the Consent Mode ad_storage parameter becomes the single control over whether GA4 passes advertising identifiers to a linked Google Ads account, with Google Signals narrowed to analytics-only reporting. We could not find that statement on Google’s own consent mode or Google Signals help pages, or in the GA4 release notes, as at 30 July 2026. It is plausible and consistently reported, but a vendor consensus is not documentation. Check the behaviour in your own property — and implement your consent signal correctly either way.

GST distorts the comparison, in both directions

Two Singapore accounting quirks bend tracking numbers. On the revenue side, e-commerce platforms usually pass the gross order value to the pixel, GST included. At 9% that inflates every ROAS figure by roughly 8.3% before you have done anything wrong.

On the cost side the platforms differ. Google bills Singapore accounts through Google Asia Pacific Pte. Ltd., a Singapore entity, and charges 9% GST — reclaimable as input tax if you are GST-registered. Meta bills under the Overseas Vendor Registration regime and charges no GST once you supply your GST number. A S$10,000 media month therefore costs a non-registered business S$10,900 on Google and S$10,000 on Meta, and neither dashboard mentions it. Our Google Ads cost guide and Meta ads cost guide cover the local billing detail.

How to stop double-counting

The commonest corruption in an ad account is the same conversion arriving twice. Three causes, all fixable in an afternoon.

Then accept the structural double-count you cannot fix. Google and Meta will both claim the same sale, because each is only ever asked about its own contribution. Adding platform-reported conversions across channels always overstates reality. The only cure is to compare one blended number — total leads or revenue from your own system, divided by total media spend — against the platform reports, and treat the gap as information rather than error. Our guide to attribution models explained takes that argument further, and measuring social media ROI applies it to organic channels.

A quarterly tracking audit you can actually run

Tracking rots. Someone rebuilds a thank-you page, a plugin updates, a developer removes a tag they did not recognise. Run this quarterly and date it.

What a good setup actually looks like

Strip away the tooling and a competent setup has four properties. One conversion definition the whole business agrees predicts revenue. One system nominated as the source of truth for money, which is almost never an ad platform. Windows that match the real sales cycle and are written down. And a known, explainable gap between what the platforms claim and what the bank received.

The aircon company’s fix took a day and a half: one conversion action promoted to primary, the GA4 import demoted to secondary, the WhatsApp tap given its own event and a measured close rate, and one spreadsheet reconciling invoices to media spend monthly. The reported numbers went down. The decisions got better immediately, because they were finally being made against something real.

If you would rather not audit your own plumbing, that is what we do — see our performance marketing services and the results we have produced for Singapore businesses. If you are choosing between channels rather than fixing measurement, start with Meta versus Google Ads.

Frequently asked questions

Why do Google Ads, GA4 and Meta all report different conversion numbers?

Because they measure different things over different periods. Google Ads counts conversions attributable to a Google ad click within a 30-day default window, credited to the click date rather than the sale date. GA4 counts all conversions from all sources with a 90-day default lookback for most key events. Meta counts conversions attributable to a Meta impression or click on a 7-day click, 1-day engage-through, 1-day view default. Different scope plus different windows equals different totals, permanently.

Should I use the Google Ads tag or import conversions from GA4?

Use the native Google Ads tag as the bidding source and GA4 for cross-channel analysis. If you import from GA4, note that Google states conversions created from Google Analytics are secondary by default and can only be promoted to primary in Google Ads — so an imported action will not drive Smart Bidding until you change it. Never run both a native tag and an imported key event for the same action as primary.

Do I need consent for conversion tracking in Singapore?

Singapore has no prior-blocking requirement, so you are not obliged to suppress all tags until a banner is clicked. However the PDPC’s Advisory Guidelines on Selected Topics require consent for cookies used to target advertising, and browser inaction does not count as consent. The Legitimate Interests Exception does not cover direct marketing. Publish a clear cookie notice, obtain consent for advertising cookies, and confirm your banner actually controls the tags. This is not legal advice.

How do I track WhatsApp enquiries as conversions?

You cannot track the conversation, so track the handoff and its economics. Fire a distinct event when the WhatsApp button is tapped, treat it as a proxy conversion, and measure how many taps become paying jobs so you know what a tap is worth. To close the loop, pass a source reference into the pre-filled message, or use enhanced conversions for leads to upload the closed deal against hashed contact details. DNC rules apply to marketing messages sent to a Singapore phone number.

Is conversion tracking setup claimable under PSG or EDG?

Only narrowly. The Productivity Solutions Grant supports up to 50% of eligible costs capped at S$30,000, but only for pre-approved solutions — typically software, not implementation retainers or ad spend. The Enterprise Development Grant explicitly excludes advertising and media buys. EnterpriseSG also requires that the company “must directly apply for and manage the grant. Third-party applications or management are not permitted.” SDM is a pre-approved PSG vendor, but you apply through the Business Grants Portal yourself. EDGE, which streamlines MRA, PSG and EDG into a single scheme available to all Singapore businesses including non-SMEs, arrives in the second half of 2026 with no published percentages or caps — so nobody should be quoting you any.

How often should conversion tracking be audited?

Quarterly as a baseline, and immediately after any website change, plugin update, checkout change, consent banner change or developer handover. Tracking rarely breaks loudly. It breaks silently, and you find out when a quarter of reporting turns out to be fiction.

Related measurement guides

• How to track phone calls and WhatsApp leads in Singapore — the enquiry routes this guide flags as hard to measure, covered properly.
• UTM tags done properly — labelling the traffic before it ever reaches a conversion.

Related measurement guides

• PDPA-compliant marketing tracking in Singapore — the consent rules that sit underneath every tag in this guide.
• Vanity metrics in Singapore marketing — once the tracking is right, deciding which numbers are worth reporting.