PPC for e-commerce companies in Singapore: keeping unit econ

Checkout on a Singapore store does not fail in the auction. It fails when a shopper on mobile, comparing a Lazada voucher against your branded PDP, lands on a page that still quotes GST like an afterthought, hides shipping until step three, and takes 4.2 seconds to become interactive. That is the actual job of PPC for e-commerce companies in Singapore: buy high-intent demand, then keep the unit economics honest after GST, last-mile, and marketplace competition have taken their cut.

Google Ads still averages about 4.2x ROAS across industries in 2026 benchmarks. Meta sits nearer 2.8x. Those numbers are not a Singapore fashion, F&B, or electronics P&L. Search CPC has climbed (global average near $2.96, up 12% year on year), mobile already takes 63%+ of paid search clicks, and conversion on phone still lags desktop by 30-40%. If your Shopping and Search mix is judged on last-click ROAS in SGD without contribution after 9% GST, you will scale the wrong SKUs.

This piece is for marketing leads who own the paid number at Singapore e-commerce brands: D2C sites, marketplace-plus-own-site hybrids, and multi-brand operators running Google Shopping, Search, Performance Max, and a slice of Meta. The play is not “add more campaigns.” It is feed quality, query intent, PDP speed, and conversion signals that survive privacy loss.

Why Singapore shopping auctions punish last-click ROAS

Shoppers here bounce between Shopee, Lazada, Qoo10, and brand sites in the same session. They search “air purifier 40 sqm GST included” and “same day delivery east” in the same hour. If you bid like a US catalogue (headline ROAS, no GST, free-shipping folklore), Google will happily spend on SKUs that look cheap in the auction and thin after fulfilment.

Contribution, not ROAS, is the filter. A 5.1x last-click on a low-margin consumable that needs two-hour last-mile in CBD can lose money versus a 3.4x on a higher AOV bundle with pickup at a pop-up. Automated bidding does not know that unless you feed values that already net GST, typical discounting, and return rates. Plenty of teams still upload item prices as list, then wonder why smart bidding “inflates budgets without improving conversion quality.” It is doing what you asked: maximise conversion value on a lie.

Privacy makes the lie worse. You can lose 20%+ of conversion data to ad-blockers and consent gaps. Smart bidding then chases the remaining, easier conversions: existing brand search, remarketing, and cheap add-to-cart events that never become paid orders. Performance Max will drive the bulk of volume in many accounts. That is fine if the conversion you send it is a purchase with a realistic value, not a micro-event you turned on to “give the algorithm data.”

Mobile is the other tax. Most paid clicks in this market arrive on a crowded MRT or a HDB sofa. If Interaction to Next Paint is ugly, the bid you paid is already gone. Only 55.9% of origins pass all three Core Web Vitals in May 2026 CrUX data. Singapore retail PDPs that still ship unoptimised image carousels and third-party review widgets sit on the wrong side of that line. You do not need a study to see it: bounce after 8 seconds, then a branded search later that last-click credits to Search, not to the Shopping click that started the job.

If your landing experience is the leak, a dedicated Google Ads programme for e-commerce should start with the handoff, not another campaign clone.

How a Singapore catalogue should bid: GST, feed, and intent

Treat the Merchant Center feed as a pricing and availability system, not a CSV you refresh when someone remembers. Titles need the words people type in Singapore English: capacity, voltage, size, “official”, “local warranty”, “GST included” when that is true. Do not stuff every synonym. Do not leave GST exclusive on the feed while the PDP shows inclusive. Mismatch is how you buy clicks that bounce at the price block.

Split intent before you split budgets. Brand Search protects the name against resellers and marketplaces bidding on your trademark. Non-brand Search captures category and problem queries. Shopping and PMax cover SKU discovery. If the same hero SKU is in all three with identical creative and no negatives, you pay three times for one basket and then argue about cannibalisation in a monthly deck.

Value rules belong in the account, not in a spreadsheet the CFO opens after the month closes. Pass purchase value that already reflects typical promo depth. Exclude or down-weight SKUs that cannot clear contribution after 9% GST and your real shipping mix (self-collect, standard, express). New accounts under CPC pressure often snap back to manual CPC. That can be a short diagnostic, not a strategy. AI Max and well-signalled smart bidding can run 12-18% lower CPCs than sloppy standard Search, but only if Enhanced Conversions and first-party match quality are actually on.

Query hygiene still matters even in a “signals over keywords” year. Google still needs keywords and negatives. 20-30% of SEM budgets routinely die on irrelevant queries, neglected negatives, or PDPs that do not match the ad. For Singapore stores, the usual waste pile is wholesale, “free download”, DIY repair, job-seeker, and marketplace coupon queries. Review search terms weekly until the list is boring, then monthly. Constant structure reshuffles reset learning. Fewer campaigns, broader targeting, and a clean conversion setup beat a 2020 maze of exact-match ad groups.

Meta is optional, not mandatory, for most Singapore e-commerce brands until Search and Shopping contribution is stable. When you do run it, creative is your targeting. UGC and hooks die in days, not weeks. Rising CPMs show fatigue before volume drops. Do not use Meta’s in-platform CPA as the source of truth when it diverges from Shopify or your OMS. That objection is real: diagnostics like Event Match Quality do not reliably predict profitable orders.

A scorecard Singapore store marketers can run before they raise spend

Work through this in order. Do not jump to budget because last week’s ROAS held up.

• Confirm the conversion you optimise is a paid order (or a high-intent checkout start you have proven correlates), with value in SGD that nets typical discounting. Turn off junk micro-conversions from bidding.

• Turn on Enhanced Conversions with the unified toggle and a server-side or CAPI-style backup so you are not flying blind after 20%+ client-side loss.

• Audit Merchant Center: GTIN where required, availability, price parity with the PDP, GST treatment, and local shipping services that match what the site actually offers.

• Rewrite titles and custom labels for contribution: hero margin, clearance, marketplace-overlap, and “never scale.” Bid and tROAS by label, not by vanity category.

• Build negatives for wholesale, careers, tutorials, and competitor coupon language. Keep a shared list across Search and PMax brand exclusions where policy allows.

• Check PDPs on a mid-range Android on mobile data. If LCP or INP fails, fix images, third-party scripts, and sticky ATC before you add keywords.

• Align ad copy to the first screen: price inclusive of GST if that is how you sell, stock, warranty, and delivery promise. Do not send category queries to the homepage.

• Separate brand, non-brand, and Shopping reporting. If PMax is eating brand, constrain it. “PMax drives the bulk now” is not a reason to hide the mix.

• Reconcile Google, Meta, and the store daily for 14 days, then weekly. Last-click will double-count. Decide which number the bid uses and which number finance uses.

• Only then raise budgets on labels that clear contribution for two consecutive weeks, not on a 3-day learning spike.

Post-click, watch where thumbs stall. HeyLead Insights is built for that: session behaviour, scroll, and form abandon on the PDP and checkout, so you stop guessing whether the auction or the page is the problem.

Two Singapore account failures that looked like “the algorithm”

A beauty brand in Tai Seng scaled PMax after a 4.8x week. Finance later found 37% of attributed revenue sat on kits that were GST-exclusive in the feed and GST-inclusive on site. Shoppers clicked, saw a higher number, and bounced. Smart bidding kept buying the cheap feed price. The fix was not a new campaign type. It was price parity, a custom label that banned those kits from tROAS targets, and purchase values that matched what Stripe actually captured. Contribution recovered in 11 days. Volume dipped, then paid orders rose.

An electronics seller on a Jurong warehouse stack ran Search to the homepage because “everyone knows the catalogue.” Category queries for aircons and robot vacuums hit a slider, not a filtered collection. Mobile conversion sat 41% below desktop. They rebuilt two intent pages (capacity and room-size, not brand hero), cut third-party chat until after ATC, and added social proof above the fold. Checkout rate on paid traffic moved in a way CRO teams recognise: not a miracle, a leak closed. If someone tells you they just send traffic to the homepage, that is your cue to keep looking.

Realistic timelines sit at 3-6 months for a durable mix. No one can quote a ROAS before they have seen the account, the feed, and the unit economics.

What we hear from accounts we audit

These are composite observations from audits, not attributed testimonials: the same failure modes show up across D2C beauty and electronics retail.

“We were celebrating 4x PMax until we realised the conversion ping was add-to-cart, not paid order, and GST was still exclusive in Merchant Center. Contribution was negative on the SKUs we scaled.”

“Creative died in five days on Meta and we kept blaming the learning phase. Search was fine. The page was slow on Android and we had no negatives for wholesale queries.”

FAQs

Should Singapore e-commerce brands start with Performance Max or Search?

Start with clean conversion tracking, a truthful feed, and brand Search protection. Add Shopping, then PMax once purchase values and negatives exist. PMax can reach 4x-8x in optimised accounts. It will also hoover brand and junk queries if you let it.

Is automated bidding just going to inflate spend?

It will if values are wrong, micro-conversions are in the bid, and PDPs mismatch ads. With Enhanced Conversions, contribution-based values, and stable structure, it usually beats panic manual CPC after the first diagnostic week.

How do we handle GST in Google Ads?

Show the same GST treatment in the feed, RSA, and PDP. Bid on contribution after 9% GST, not on list price. Do not let “from” pricing in ads collide with inclusive checkout.

Do we still need negative keywords?

Yes. Signals help. They do not replace query control. Wholesale, jobs, tutorials, and coupon-only intent still burn 20-30% of lazy SEM budgets.

When should Meta enter the mix?

After Search and Shopping clear contribution and you can refresh creative on a days-not-weeks cycle. Broad targeting plus a testing engine beats 2020 lookalike stacks. Weak funnels plus broad targeting is how you buy cheap, useless traffic.

Putting it to work

Pull 30 days of Shopping, Search, and PMax by SKU custom label and net each row for GST, shipping, and returns before you touch bids. That is the near-term move that changes what you scale.

This week:

• Export purchase conversions and confirm the action used for bidding is a paid order in SGD.

• Spot-check 15 top-spend SKUs for feed vs PDP price and GST match.

• Run the Core Web Vitals checker on those PDPs on mobile.

• Add a first pass of negatives from the last 14 days of search terms.

• Tag every paid URL with the UTM builder so GA4 stops arguing with Google Ads.

Next 30 days: rebuild value rules by contribution label, constrain PMax off brand if it is stealing, and only then raise budget on labels that stay profitable for two weeks.

If the painful bit is the gap between auction ROAS and real contribution after GST, feed errors, and slow PDPs, HeyLead can own that loop: bidding on truthful values, feed and query control, and the page handoff so you are not scaling negative margin. Chat with us on WhatsApp