Client name withheld under NDA.
The client sells treadmills, benches, dumbbells and weight plates online and from a small showroom. Most carts are large, often over S$600, and shoppers take weeks to decide. Before we arrived, the only Facebook activity was boosted posts, around S$1,800 a month, aimed at a broad audience in Singapore. The posts collected likes and comments, but the owner could not say how many sales they produced.
Our first check showed why. The pixel was installed, but purchases were recorded only when a customer reached the final thank-you page, and many never did on mobile. Only 55% of store orders appeared in Meta, so the ads were learning from half the picture. We recorded purchases, revenue, spend and showroom bookings before changing anything, so the later numbers would mean something.
We rebuilt the measurement first. The pixel events were corrected, and we added the Conversions API so purchases from the website reached Meta even when a browser blocked the pixel. We tested a real order on a phone and a laptop, then matched Meta's count against the store's order list. We also moved the old boosted posts into a proper campaign structure with clear names.
Meta's system bids on what it can see. With tracking fixed, the ads could learn which shoppers actually bought, and the owner could trust the report. Without this step every later test would have been judged on half the data.
We connected the product feed to a Meta catalogue and cleaned up titles, prices and stock status. We built a catalogue campaign for shoppers who had looked at products and a prospecting campaign for new audiences, split by product type: cardio machines, strength gear and small accessories. Each had its own budget so a cheap accessory could not use up the money meant for a treadmill.
A treadmill and a pair of resistance bands are different purchases with different buying times. Separate campaigns let us judge each on its own cost per purchase and stopped low-ticket items from hiding the weak spots.
Shoppers could not feel the equipment through a screen, so we filmed short demos in the showroom: how a bench folds, how loud a treadmill runs, how plates store. We tested these against still catalogue images and a customer unboxing clip. Copy stayed on the product and delivery, with no claims about fitness, health or body results, in line with Meta's advertising policies.
Big purchases need proof the item is solid and fits the room. Short demos answered those questions before a visit or a message. They also gave us fresh creative each fortnight, which slowed ad fatigue.
We set up retargeting by stage: product viewers within seven days, add-to-cart visitors within three days, and past buyers shown accessories only. We added a showroom booking button to the website and ran a test using Meta instant forms to collect visit requests. The instant form brought cheap sign-ups, but few people showed up, so we stopped it after three weeks and sent traffic to the website booking page instead.
A shopper deciding on a S$1,000 machine often wants to try it first. Booking on our own page produced fewer but better visits. The failed form test is a good reminder that a low cost per lead means little if nobody arrives.
The new-year season lifted demand, and we raised the budget in steps of about 20% every few days, checking cost per purchase before each rise. We held back on the weakest prospecting audiences and moved money to catalogue and retargeting. A weekly one-page report covered spend, purchases, return on ad spend and showroom bookings, so the owner always knew where the money went.
Stepped increases protect the return while the system keeps learning. Month 5 is a peak season, so the owner should expect mid-year to be quieter and plan the budget with that in mind.
After 5 months, monthly Facebook Ads revenue rose from S$3,900 to S$21,000 on S$6,000 of spend, up from S$1,800. Purchases grew from 6 to 30, return on ad spend moved from 2.2x to 3.5x and cost per purchase fell from S$300 to S$200. Showroom visit bookings went from 5 to 21 a month, and Meta now recorded 92% of store orders instead of 55%. Part of the month 5 result comes from the new-year season.
Tracking corrected, Conversions API live, catalogue connected, campaigns split by product type.
Video demos tested, retargeting by stage running, showroom booking page live, instant form test stopped.
Budget raised in steps through the new-year season, weak audiences cut, weekly one-page report in place.
Only 55% of orders reached Meta at the start. After the Conversions API and pixel fixes it was 92%, and cost per purchase fell from S$300 to S$200. Bigger budgets on poor data just buy more of the wrong shopper.
Short showroom demos answered the questions shoppers had about size, noise and storage. Buyers who needed time to decide were reached again by retargeting, which kept the product in front of them during a long consideration period.
The instant form produced low-cost sign-ups but weak showroom attendance. Moving visit requests to the website lifted bookings from 5 to 21 a month. Judge each test on real visits and sales, not on the cheapest lead.
How we plan, run and report on Facebook campaigns for Singapore businesses.
Paid campaigns across Meta and other social platforms.
Bringing back shoppers who looked but did not buy.
Another Meta Ads story, this time for a skincare brand.
Tell us what you sell and what a sale is worth. We will tell you what a realistic five-month Meta Ads plan looks like.