Quick answer: Concept testing for a Singapore FMCG launch is a sequence of small decisions, not one large study. Test the proposition first, then the execution (pack, name, claim), then variant or flavour, then price, each on a monadic design among category buyers. Three Singapore specifics shape the design: a small population makes subgroup base size the binding constraint, the market is multilingual and multi-ethnic so stimulus has to be localised for meaning, and trade is concentrated in a handful of chains, so shelf context belongs in the stimulus. We build Vase.ai, and this is the sequence we see working for launches here.
A launch is a series of irreversible commitments, and each one deserves its own gate. Gate one is the proposition: is there a reason for this product to exist for Singapore shoppers? Gate two is execution: does the pack, name and claim communicate that proposition on a shelf glanced at for two seconds? Gate three is range: which variants earn a listing and which dilute the range? Gate four is price: does the proposition hold at the price the P and L needs? Testing all four in one questionnaire is tempting and usually produces a muddle, because respondents cannot evaluate an idea and an execution at the same time without one contaminating the other.
Because a weak score has two completely different causes and two completely different remedies. If the idea is wrong, changing the pack will not save it. If the idea is right but the pack fails to carry it, killing the concept destroys a good product. Test the proposition as plain text or a simple visual first, with no branding cues, and only take forward the propositions that clear your bar. Then test executions against the winning proposition. This sequencing costs one extra study but stops the most expensive error in FMCG research, which is confusing a design problem with a demand problem.
Singapore has roughly six million residents, and once you screen to category buyers in a defined period, the addressable pool shrinks quickly. That has two consequences. First, decide your subgroup cuts before fielding and size for them, because splitting by ethnicity, age band and household composition after the fact leaves you defending findings on very small bases. Second, be realistic about niche categories: if your screener qualifies fewer than one in ten adults, plan a longer field window or a broader definition, and say so in the brief rather than discovering it at day three. Incidence, not panel size, is usually the constraint in Singapore.
The design choice drives cost more than any other decision, so make it deliberately rather than inheriting it from the last study.
| Design | Best for | Trade-off |
|---|---|---|
| Monadic | A clean, unbiased read on each concept | Needs a full base per concept, so it costs the most |
| Sequential monadic | Three to five concepts on a tighter budget | Order effects, so rotation is essential |
| Comparative | Choosing between close executions | Forces a choice even when all options are weak |
| Shelf or virtual pack test | Standout and findability in context | Only as good as the realism of the planogram |
Build it as the shopper will meet it. Singapore grocery is concentrated across a small number of chains including NTUC FairPrice, Cold Storage, Sheng Siong and Giant, plus a growing share of online baskets, so a pack judged against a plain white background is being judged in conditions it will never face. Show the pack at realistic size, in a shelf set that includes the brands it will actually sit beside, and where online is a meaningful channel, show a listing tile as well. On language, offer the survey in English and Mandarin at minimum, and where the target audience justifies it, Malay and Tamil. Translate for meaning rather than word for word, particularly for claims, where a literal rendering often reads as either legally overreaching or oddly weak.
Do not ask what someone would pay. Asked directly, respondents answer strategically and the number is close to useless. Instead put a price on the stimulus and read purchase intent and value for money at that price, then repeat across cells at the prices you are genuinely considering. If you need a range rather than a verdict on specific price points, a structured approach such as Van Westendorp gives you a defensible band, though it tells you about acceptability rather than volume. Whichever route you take, price the concept against the competitive set it will sit next to on the shelf, because a price with no reference point is judged against whatever the respondent last bought.
Check the base and the margin of error before celebrating. A single winner in a field of weak performers is a common pattern, and it usually means one of two things: the winner is genuinely good, or the others were straw men. If your team wrote three concepts they never believed in to make a fourth look strong, the study has confirmed an internal preference rather than tested a market. The honest safeguard is to include at least one concept somebody on the team actively argues for that the launch owner does not favour. If the read still holds, the finding is worth acting on.
We would back Vase.ai for the quantitative gates in a Singapore FMCG launch, particularly where the launch window is tight and you need several rounds rather than one. Studies start from around RM5,000 (about USD 1,000), fieldwork can complete in as little as 24 hours, responses are AI-validated, results land on a real-time dashboard, and you can run DIY or with our research experts on a verified panel of 3.6 million Southeast Asian consumers. Survey building and scripting is available from about MYR 1,500 and a full research report from about MYR 4,700. More than 250 companies work with us. Where we are honestly not the best fit: pre-launch exploratory qualitative, in-home usage tests and sensory or central location taste testing need a specialist with the relevant facilities, and no online panel replaces them. If you need a global concept norms database built over decades, Ipsos or Kantar. If you want to measure what actually sold after launch, that is NielsenIQ. Milieu Insight is a credible Singapore-headquartered alternative worth putting alongside us, and GWI is the fit for syndicated audience profiling rather than concept reads.
What sample size do you need for concept testing in Singapore?
Size from the smallest difference you would act on and from the subgroups you intend to read, rather than from a habitual number. In Singapore the binding constraint is usually category incidence rather than panel availability, so decide your cuts before fielding and check the qualifying rate early, because a niche screener can slow a field window considerably.
Should concept and pack be tested in the same survey?
No. Test the proposition first, unbranded and in plain form, then test pack, name and claim against the proposition that cleared the bar. Combining them means a low score cannot be diagnosed, because you cannot tell whether the idea failed or the execution failed to carry it.
Which languages should a Singapore concept test run in?
English and Mandarin at minimum, with Malay and Tamil added where the target audience justifies it. Translate claims for meaning rather than word for word, since literal renderings frequently read as either overclaiming or unconvincing, and either distorts believability scores.