Paid social is the easiest channel to start and the easiest to waste money on, which is why so many Singapore businesses have tried it, got an unclear result, and shelved it. When you hire a social media advertising agency, the fee is not really buying access to the platforms. You can open those accounts yourself in an afternoon. It is buying a process for deciding what to make, who to show it to, how fast to spend, and what to believe about the results. Those four things are the whole job, and three of them are usually done badly.
This piece is about what that process should look like from the inside, so you can tell whether the one you are paying for exists.
Creative Is the Variable That Matters Most, and It Gets the Least Attention
If you strip paid social down, most of the performance difference between two accounts with similar budgets comes from the creative. Targeting has become progressively less controllable as platforms have moved to automated delivery. Bidding is largely automated. What you can still control completely is what the ad says and shows.
Despite that, the typical arrangement we inherit is an agency running six carefully structured campaigns off three pieces of creative that were made eight months ago, because creative production sits with a different supplier or with the client's internal team and moves at a different speed.
A working creative process has a few properties.
It tests one thing at a time, at the level where the difference matters. The hierarchy of impact, in our experience, runs roughly: the offer, then the hook in the first two seconds, then the format, then the visual style, then the copy details. Testing three different button colours while the offer stays unclear is theatre. Testing two genuinely different offers against each other will tell you more in a week than a quarter of minor variations.
It produces enough volume to learn from. A single creative per month cannot generate a signal. Our rule of thumb is a minimum of four to six new concepts per month for an account spending meaningfully, with variants within each concept. That is a production requirement, not a media requirement, and it is the reason a paid social programme without a video content pipeline behind it tends to stall after the first quarter.
It retires winners before they collapse. Every creative decays. Frequency climbs, the addressable pool gets exhausted, and performance falls. The mistake is waiting for the decline to show up in cost per acquisition, by which point you have spent several weeks of budget learning something you could have anticipated. Watch frequency and click-through rate for the leading indicators.
It reuses what worked, in new forms. A concept that performs well is not one ad, it is a template. Same hook, different demonstration. Same structure, different customer. Most accounts throw away winning ideas because the ad itself got tired, and then start from a blank page.
It respects format. A film shot for a landscape website hero and cropped to vertical is visibly a crop, and it underperforms material framed for the feed it sits in. Each placement has its own conventions, and the differences between what works on Instagram ads, a Facebook feed, TikTok and LinkedIn are large enough that shared creative is a compromise on all four.
Audience Structure: Fewer Boxes Than You Think
The second thing an agency controls is how the account is divided up. This is where a lot of unnecessary complexity enters.
The instinct is to segment heavily. Separate ad sets for each interest category, each age band, each location, each product line, on the reasoning that granularity gives control. On modern paid social platforms that instinct is mostly counterproductive, because every segment you create splits the budget and the conversion data, and delivery systems need a certain volume of conversions per segment to optimise at all. Slice a modest budget eight ways and no slice gets enough data to learn from, so every one of them stays in a permanent learning state.
The structure we tend to use for a Singapore SME looks closer to this.
| Layer | Typical share of budget | Purpose | What success looks like |
|---|---|---|---|
| Broad prospecting | 55 to 70 percent | Let the platform find buyers, creative does the targeting | Volume at acceptable cost, stable over weeks |
| Focused prospecting | 10 to 20 percent | One or two distinct segments worth separating | Beats broad on cost per acquisition consistently |
| Mid-funnel retargeting | 10 to 20 percent | Engaged viewers, site visitors, video watchers | Higher conversion rate, low frequency |
| Existing customers | 5 to 10 percent | Repeat purchase, cross-sell, exclusion management | Incremental revenue, not cannibalised |
The key discipline is exclusions. If your retargeting audience is also being reached by prospecting, you are bidding against yourself and paying twice for the same person. When we audit accounts, missing or contradictory exclusions is one of the most common structural faults we find, and it costs real money silently.
The second discipline is separating by economics, not by demographics. If two products have different margins or different sales cycles, they should not share an ad set optimising to the same event, because the platform will chase whichever converts more easily regardless of what it is worth to you. Segment by what the conversion is worth. Do not segment by what is easy to describe.
Budget Pacing Is Where Quiet Money Disappears
Pacing is unglamorous and it is where a large share of avoidable waste lives.
Changing budgets too often resets learning. Every significant edit to an ad set can send it back into a learning phase, during which performance is unstable and costs are typically higher. An account where somebody adjusts budgets daily in response to yesterday's numbers can spend much of the month in that unstable state. We generally hold changes to a defined cadence, adjust by meaningful steps rather than small ones, and resist reacting to a single bad day.
Month-end rushes are the other classic. An agency that underspends for three weeks and then dumps the remainder in the final days will show you a full spend report and a poor result, because the last few days were bought at whatever price was available. Pacing should be roughly even unless there is a commercial reason for it not to be.
Day of week and time of day matter here more than people expect, particularly for B2B, where enquiries that land at 11pm on a Saturday convert at a different rate from the ones that land on a Tuesday morning. That is not a reason to switch off the weekend automatically. It is a reason to look.
Scaling needs a rule. Increasing a winning ad set's budget by a large multiple in one move usually breaks it. Incremental increases at intervals, or duplicating into a new ad set at a higher budget, are the two usual approaches, and which works better varies by account. What matters is that somebody has a stated method rather than reacting to enthusiasm.
There is also the floor question. Below a certain daily spend, an ad set cannot generate enough conversions to optimise and the whole exercise becomes noise. For most Singapore SMEs running lead generation, we would rather run two properly funded ad sets than six starved ones. If the total budget cannot support the structure, the structure is wrong, not the budget.
Your attribution is wrong, and you should plan around it
In paid social reporting, the numbers in the platform dashboard are not measurements. They are estimates produced by a system with a commercial interest in the answer, working with progressively less data than it used to have.
Privacy changes, cookie restrictions, browser defaults and app tracking rules have all reduced what the platforms can observe. They fill the gap with modelling. Modelled conversions are not fabricated, but they are inferences, and they tend to be generous to the channel doing the inferring. If you add up the conversions claimed by your paid social platform, your search platform and your analytics tool, you will almost always get a number larger than your actual order count. Everyone has noticed this. Very few reporting packs acknowledge it.
A common mistake we see is an agency reporting platform-attributed return on ad spend as though it were a financial figure, then a client making budget decisions on it. The figure is directionally useful for comparing one ad against another inside the same platform over the same period. It is close to useless for deciding whether the channel as a whole is worth funding.
What we recommend instead is boring and more reliable.
Watch blended figures. Total marketing spend against total revenue in the same period. It is crude, it does not tell you which channel did what, and it is the only number that cannot be inflated by double counting.
Run occasional holdouts. Turn paid social off in a defined period, or off for a defined geography, and watch what happens to blended results. This is uncomfortable and it costs something. It is also the only causal evidence available to a business at SME scale. We have run these for clients and had the answer come back in both directions, which is exactly why they are worth running.
Ask buyers directly. A single "how did you hear about us" field on the enquiry form, free text rather than a dropdown, is noisy and biased and still frequently more informative than the dashboard, particularly for high consideration purchases where the customer saw six things over three weeks.
Agree the reporting metric before the campaign starts. If the agency picks the metric after the results are in, you will always be shown the flattering one.
None of this means paid social does not work. It often works very well. It means anyone presenting you a precise return figure to two decimal places is presenting a model, and you should ask what happens to the recommendation if the model is 30 percent out.
A Singapore Case Study Worth Learning From
One Singapore consumer services business we worked with came to us after roughly nine months with a previous arrangement that had reported a return on ad spend of just over 4. Their finance lead had raised an awkward question: if the ads were returning four times spend, why had revenue been flat all year?
The audit answer was unglamorous. The account had eleven ad sets running on a media budget of about SGD 9,000 a month, which meant most of them never accumulated enough conversions to leave the learning phase. Retargeting was not excluded from prospecting, so a meaningful share of the reported conversions were existing customers who were already going to buy, counted as new acquisitions. The same three creatives had run since the second month. And spend was consistently back-loaded, with close to a third of the monthly budget going out in the final five days.
We rebuilt it around four ad sets, put proper exclusions in, moved a fixed portion into remarketing with its own separate reporting so it could never be confused with new acquisition, and set a production schedule of five new creative concepts a month.
Then we ran a two week holdout, which the client agreed to reluctantly. Blended revenue fell by noticeably less than the platform-reported contribution implied it should have. The honest read was that the true incremental return was somewhere around half what had been reported. Not zero, not fraudulent, just a channel doing less than the dashboard claimed.
Knowing that changed the decisions. They cut the budget modestly, put the difference into the creative pipeline, and over the following two quarters blended revenue rose while total media spend was slightly lower than the year before. The reported return on ad spend in the platform went down, because we had stopped counting existing customers as conversions. That was the point.
The Offer Does More Work Than the Targeting
One thing worth separating out, because it sits awkwardly between the agency's job and the client's: paid social performance is bounded by what you are actually offering, and no amount of media skill lifts a weak offer past that ceiling.
Cold audiences on a social feed were not looking for you. They were looking at something else and you interrupted. That is a harder starting position than search, where somebody typed a problem into a box. It means the thing you ask for has to be proportionate to how little the person currently cares. Asking a cold viewer to book a paid consultation is usually too large a step. Asking them to watch thirty seconds, read a useful page, or take a small low-commitment action is proportionate, and it builds the pool your retargeting layer then works on.
This is why accounts with nothing between the ad and the sale tend to plateau. There is no mid-funnel to retarget, so every month starts from cold and the cost per acquisition never improves. Adding one sensible intermediate step frequently does more for blended performance than any structural change we could make to the campaigns.
The corollary is that the destination page is part of the media buy, not an afterthought. A page that loads slowly on mobile data, buries the relevant information below three scrolls, or asks for nine form fields will waste a meaningful share of everything you spend getting people to it. We have taken over accounts where the fastest available improvement was not in the ad platform at all, it was rewriting the landing page the traffic arrived on. That change costs once and improves every click you buy afterwards.
Pricing transparency belongs in this conversation too. For considered purchases, a page that gives an honest range converts better with the right people and filters out the wrong ones, which shows up as better lead quality even when the raw conversion rate falls. A lower conversion rate with a higher close rate is a better business outcome, and it will look like a decline in every dashboard you own. That is worth agreeing before it happens, because otherwise somebody will panic in week three and reverse a change that was working.
What to Ask Before You Sign
If you are choosing a paid social partner, the questions that separate the useful from the plausible are fairly specific.
How many new creative concepts will you produce per month, and who makes them? If the answer is that creative is the client's responsibility and there is no volume commitment, the account will stall by month three.
What is your account structure, and why that many ad sets for my budget? A partner who cannot explain why the structure matches the spend level is applying a template.
How do you handle attribution, and what will you show me? Listen for whether they volunteer the limitations before you raise them.
What is your change cadence and your scaling rule? Any coherent answer is better than none. No answer means somebody is adjusting things on instinct.
What happens in month one? A sensible first month is mostly measurement setup, exclusion hygiene, and a first creative batch. An agency promising results in week two is either lucky or managing your expectations badly.
And the one people forget: what does the landing experience look like? Paid social sends cold traffic to a page, and the page converts it or does not. Plenty of accounts we inherit have a media problem that is actually a page problem. Fixing the destination is usually cheaper than buying more traffic to a page that leaks.
Field Notes
Across the paid social accounts we took over in the past year, the median number of active ad sets at handover was nine, against a median monthly media budget of around SGD 7,500. On our rebuilds, the median dropped to four. In most of those accounts, cost per acquisition improved inside two months without any increase in spend, purely from consolidating the data into fewer, better funded ad sets.
The other figure that keeps repeating is creative age. At handover, the median age of the top spending creative was over four months. Once we hold a production cadence, the top spender is usually under six weeks old, and the accounts that keep the cadence show far less of the sawtooth pattern where performance collapses and has to be rescued.
On attribution, the pattern across the holdout tests we have run is consistent enough to mention: platform-reported contribution has exceeded the measured blended effect every time. The size of the gap varies a lot by business, and the direction has not yet varied at all.
The Short Version
A social media advertising agency is worth its fee if it is producing enough new creative to keep learning, structuring the account to match the budget rather than to look sophisticated, pacing spend deliberately, and telling you honestly what the numbers can and cannot prove. If it is doing three of those four, you will probably do fine. If it is only doing the reporting, you are paying for a dashboard.
If you want a second opinion on an account you are already running, including the uncomfortable parts, get in touch and we will look at the structure, the creative cadence and the attribution set-up. There is more on how we think about paid channels in our piece on performance marketing agencies, and on the organic side in our notes on influencer marketing agencies. The full picture of where paid social sits within a wider programme is covered under digital marketing and the campaign work itself under social media ads.
Natasha Tan is the founder of Digital Marketing Singapore, a full-service SEO and digital marketing agency based in Singapore. With hands-on experience across SEO, paid media, and content strategy, she works directly with Singapore businesses to build organic visibility and generate consistent leads. Natasha specialises in the Singapore market — including local search behaviour, PDPA compliance, and government grant navigation for SMEs.
