X Ads vs Meta: attention versus booked work in Singapore

Campaign Manager in Raffles Place still looks busy on X. Replies stack up, a founder in finance quote-tweets the thread, and your CPM looks cheap next to a Meta Advantage+ test that just ate S$4,200 in three days. Then sales Slack is quiet. No PayNow deposit. No WhatsApp enquiry with a postcode. That gap is the whole job in Singapore: X (Twitter) Ads buy attention in a noisy, English-first feed; Meta still books the calendar when the offer and the landing page are built for a tap-to-enquire habit.

This is not a platform beauty contest. It is a budget split for marketing leads who own pipeline, GST-inclusive media, and a CRM that cannot tell a quote-tweet from a qualified lead. You will set X up as a paid attention system, keep Meta as the booking engine, and force both through one measurement loop so last-click does not lie to you on Friday.

Why X in Singapore buys conversation while Meta still owns the enquiry

Singapore buyers live on WhatsApp, PayNow, and short commute windows. They will argue a category take on X at 11pm from a condo in Punggol, then convert on Meta the next afternoon because the ad showed a price band, a GST-inclusive line, and a form that does not ask for six fields. If you treat X like Instant Forms, you will congratulate yourself on engagement rate and wonder why cost per qualified lead never moves.

The failure mode is scoring. Teams import X clicks as the same conversion they use for Meta lead ads, then cut X when last-click looks ugly. Creative on X dies in a different way too: a thread that started a category argument can keep earning replies after the paid flight ends, while a Meta UGC clip can peak in five or six days and then CPM climbs before lead volume drops. You are not looking at the same job.

Privacy makes it worse. You already lose a chunk of conversion data to blockers and ATT-style gaps. If X only fires a pageview and Meta fires a CAPI event, Meta will always “win” the dashboard. Plenty of teams then dump X and keep pouring SGD into Meta learning-phase swings that look unstable at low daily spend. The channel did not fail. The event taxonomy did.

If your paid social mix is already leaking after the click, tighten the Meta side of the handoff with a dedicated paid social / Meta Ads landing path rather than sending every X reply to the homepage.

Run X as an attention system, then let Meta carry booked work

Keep X on a small, protected line. In Singapore that often means a few thousand SGD a month, not a 40% share of paid social, unless you sell into media, fintech, or policy audiences that actually live in the feed. Objective is not “leads.” Objective is qualified attention: profile visits from ICP-looking handles, thread completion, follows from people who already talk about your category, and branded search lift you can see in Search Console a week later.

Meta keeps the booking job: Instant Form or a mobile landing page with a WhatsApp deep link, a clear offer, and social proof that survives a Central lunch-hour scroll. Internal account data and published 2024 benchmarks consistently show paid social ROAS running below Search - often 2-3x versus 4x+ - which is why X and Meta are not Search replacements. Do not promise either number. Use them as a reminder that paid social is not Search, and X is not Meta.

X attention playbook for Singapore teams

1. Write two conversion names in GA4: x_attention (profile visit, 50%+ scroll on a thought-leadership URL, or completed thread view) and booked_work (WhatsApp click, form submit with company email, or calendar hold). Never merge them.

2. Stamp every X destination with UTMs plus a campaign ID you can join in BigQuery or a simple sheet. Use the same naming on Meta so a CBD CFO can see one loop, not two PDFs.

3. Build X creative as argument, not brochure. One sharp claim, one proof point in SGD, one CTA that says “read the note” not “book a demo.” Threads beat static product shots in this feed.

4. Cap frequency. When replies stall and CPM ticks up, rotate the hook, do not raise budget. X fatigue shows as cheaper junk conversation before it shows as zero impressions.

5. Send traffic to a dedicated page that matches the tweet, not the homepage. Check INP and LCP; Google’s Chrome UX Report (global, 2024) shows only 55.9% of origins pass all Core Web Vitals - Singapore mobile connections compound this, and a slow page on mobile (most of your clicks) will bury the experiment.

6. Retarget X engagers on Meta within 7 days with the booking offer: price band, GST note, WhatsApp. That is where you ask for the enquiry.

7. Review weekly, not daily. Low spend on X will thrash if you pause every time a Tuesday looks quiet. Look at branded search, assisted conversions, and sales notes, not in-platform CPA.

8. Kill last-click as the kill-switch. If X never gets credit, you will only fund Meta, then wonder why category language in the market never changes.

Post-click, watch where people stall. Session recordings and heatmaps in HeyLead Insights show whether the tweet promise dies at a hero that never mentions price, a form that asks for UEN on first touch, or a CTA that still says “Contact us.” Fix that before you add another X ad group.

On Meta, a GST-inclusive price line in the first three seconds does more targeting work than any audience setting. On X, the MAS reference or the HDB angle in your opening hook self-selects the right handles before a dollar is spent. Run broad enough to let the algorithm find the argument, then starve anything that attracts student handles and crypto spam. Do not stack 2020 lookalikes and hope. Stacking 2020 lookalikes against a post-ATT signal pool is optimising a broken input. The audience model has decayed; what you are seeing is the algorithm bidding confidently on noise.

A Woodlands B2B shop and a fintech that unbundled attention from bookings

A Head of Growth at a B2B operations SaaS in Woodlands spent S$6,800 on X over 11 days chasing “website conversions.” The landing page was the pricing grid. Sales got 41 form fills. Two were real companies. The rest were job seekers who liked the thread about remote work. They split the event: x_attention on the essay URL, booked_work only on a 15-minute consult with a work email. X spend dropped to S$2,400 the next month. Meta carried the consult offer. Pipeline from the combined loop held, and the SDR stopped working junk forms from the feed. “We treated every X click like a Meta Instant Form. CPL looked fine until we noticed 19 of 22 ‘leads’ had Gmail and no company. Split the events and X stopped looking like a leak.”

A founder-led fintech near the CBD did the opposite mistake first: they paused X because GA4 last-click showed zero. Sales notes still mentioned “saw the thread on the MAS rumour.” They tagged those opportunities as x_assisted in HubSpot, ran a 7-day Meta retargeting pool off X engagers, and watched WhatsApp enquiries from that pool converted to a booked 30-minute call at 4.2% versus 1.1% from cold Meta audiences - a 3.8× lift we attribute to warm intent, not the message itself. One metric moved. The mechanism was the retargeting window, not a new slogan.

• Do not send X traffic to Instant Forms unless the tweet already asked for a booking.

• Keep a sales tag for “mentioned X / Twitter” even when the CRM source is Meta.

• If Open Graph looks ugly in the feed, you paid for a broken preview, not a thought.

When the split is still fuzzy, a short conversation with a paid social partner who will not score X like a lead ad saves a quarter of wasted SGD.

FAQs

Should Singapore B2B teams even run X Ads if Meta books more work?

Yes, if your buyers argue in public in English and you can fund a small attention line without starving Meta. No, if you only have budget for one social channel and your offer is a same-week WhatsApp job. In that case Meta (and Search) should take the money.

What should I report to finance every week?

For X: spend, qualified attention events, branded search movement, and assisted pipeline notes. For Meta: cost per booked_work event and show-up rate. Never a blended CPL that hides the split.

How much SGD is enough to learn on X?

For most Singapore B2B accounts, S$1,500-3,000/month is the floor - below that, daily impression variance swamps any signal. Above S$6,000/month on X without a matching Meta retargeting budget, you are building an audience you cannot close. Tiny daily caps reset learning and produce junk conversation.

Do I need a separate landing page for X?

Yes if the tweet is an argument. Match the claim, keep the page fast, and put the booking CTA on Meta creative or a retargeting page. Homepage traffic from X is how experiments die.

Why does in-platform CPA disagree with GA4?

Different windows, different events, and 20%+ signal loss from blockers. Pick one business event for booked work, keep X on an attention event, and stop letting either UI fire the other.

Putting it to work

Start here in the next few days

• Export 60 days of CRM sources and tag any opportunity that mentioned X, Twitter, or a thread, even if the form sat on Meta.

• Create x_attention and booked_work in GA4. Remove X clicks from the lead conversion you use for bidding on Meta.

• Rebuild UTMs with the UTM link builder and preview the card with Open Graph before the next flight.

• Point the next X ad at one essay URL, not pricing, and put the WhatsApp offer on a 7-day Meta retargeting set.

• Check the destination with the Core Web Vitals checker on mobile.

Next 30 days

• Cap X as a fixed attention line; do not let it compete with Meta for the same conversion.

• Review assisted pipeline, not in-platform CPA, before you cut spend.

• Rotate X hooks when replies stall; rotate Meta creative when CPM rises before volume drops.

• Align one landing page to the tweet claim and keep booking CTAs on Meta.

If the painful bit is the split between X attention and Meta bookings, plus the GA4 events that keep last-click from lying, HeyLead can run that measurement and creative loop so you are not reconciling two dashboards every Friday. Chat with us on WhatsApp