EggAI-v0.1 · Draft synthetic research · 26 October 2026
A synthetic study of how price, proximity, fresh food, promotions, senior schemes, housebrand, festive ranging, loyalty points, online delivery and opening hours might shape where Singapore households buy their groceries.
Recent switchers, split shoppers and loyal FairPrice shoppers; heartland families, seniors, Pioneer/Merdeka and CHAS holders; home cooks; Chinese, Malay and Indian households; shift workers and young families; shoppers by monthly grocery spend.
Thirteen question-led readouts: switching drivers, perceived price gap and tolerance, promotion mechanics, new-store risk, senior schemes, fresh, housebrand, festive, Linkpoints, online, hours and split shopping.
Decide which levers the rest of the POC tests, shape the CNY 2027 promotion mix and 2027 senior schemes, draft a store-defence playbook, then validate against real shoppers.
Price and proximity carry the pull. Recent switchers name lower regular prices first (31%) and a closer store second (26%); split shoppers reverse the two. Fresh food is the leading product reason among people who moved spending (28%, against 8% for packaged goods).
Small gaps move baskets. When Sheng Siong is 2% cheaper, the forecast share buying everything there rises from 13% to 29%; when FairPrice is 2% cheaper, buying everything at FairPrice rises from 26% to 42%. Distance matters as much: a new Sheng Siong a 5-minute walk away draws at least half the visits of 56% of residents, against 21% at 20 minutes.
Most of this cannot be checked yet. 20 of 29 items are grade C: hypothetical scenarios with no real benchmark, and 14 move noticeably when the answer order changes. On 60 published Singapore grocery survey results forecast blind, the harness averaged 11.8 TVD points: close on full answer distributions (9.7), no better than a coin-flip on several single shares. See the historical checks.
Recent switchers from FairPrice to Sheng Siong; Split shoppers who use both chains; Loyal FairPrice shoppers; HDB residents aged 55+; Private-housing residents aged 25-44
Ranked switching drivers by segment; sets which levers the rest of the POC tests
Reasons for switching
Scope of this run: Segment results are forecasts for described groups; how many shoppers are switchers or loyalists is not estimated.
Model forecasts, not shopper data: for the past-three-month choice, proximity leads overall (19%), followed by price excluding promotions (17.33%), promotions (7.33%), freshness (7%), range (3%) and hours (2%). Another 38.67% are forecast not to have chosen Sheng Siong over FairPrice.
Recent switchers rank price (31%) > proximity (25.5%) > promotions (16.5%) > freshness (11.5%) > range (6%) > hours (4%). Split shoppers reverse the top two: proximity (28.5%) > price (26%), then promotions (13%), freshness (10.5%), range (6%) and hours (3.5%).
Loyal FairPrice shoppers rank proximity (12.5%) > price (10.5%) > promotions (6.5%) > freshness (5%) > range (3%) > hours (1%). HDB residents aged 55+ rank proximity (23%) > price (21.5%) > promotions = freshness (8%) > range (3%) > hours (1.5%). Private-housing residents aged 25–44 rank proximity (16%) > price (14%) > promotions (7%) > freshness (6%) > range (3.5%) > hours (1.75%).
The largest past-choice segment gaps are price: recent switchers versus loyal FairPrice shoppers, 20.5 percentage points; and proximity: split versus loyal shoppers, 16 points. For hypothetical next-choice drivers, every segment ranks price > proximity > promotions > freshness > range > hours. Overall, price leads proximity by 10.33 points (34% versus 23.67%), supporting price and proximity as the first levers to test.
References for context and interpretation: R01 · Sheng Siong's rise: Value supermarket gains ground with affluent shoppers · R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R03 · FairPrice vs Sheng Siong vs Cold Storage: An Honest Side-by-Side. See the source roles and limits in the reference register below.
Loyal FairPrice shoppers; Split shoppers who use both chains; Mainly Sheng Siong shoppers; Seniors aged 65+
Perceived price gap and the items that drive it; shortlist of known-value items
Price perception
Scope of this run: Categories, not individual items. The gap question is asked only of people who name a cheaper chain, which can be either one.
Model forecasts favour Sheng Siong as cheaper: 43.33% versus 13% for FairPrice, a 30.33 percentage-point lead; 33.33% see prices as about the same.
Mainly Sheng Siong shoppers are most likely to call Sheng Siong cheaper (60.5%), versus 51.5% of split shoppers, 44% of seniors and 31% of loyal FairPrice shoppers—a largest segment gap of 29.5 points.
Among those saying one chain is cheaper, the most common perceived saving is 5% to less than 10% (33%), followed by 2% to less than 5% (26.33%). This measures the cheaper chain whichever it is, not Sheng Siong alone. The largest segment difference within a gap band is 8.5 points for 2% to less than 5%: split shoppers 29.5%, mainly Sheng Siong shoppers 21%.
The category-level shortlist for known-value items is eggs/dairy (14.67%), fruit/vegetables (13.33%), meat/fish/seafood (11%) and rice/noodles/pasta (10.67%); individual items are not identified. Seniors put most weight on fruit/vegetables (19.5%), 5.5 points above loyal FairPrice shoppers. The largest category difference is meat/fish/seafood: 17.5% among mainly Sheng Siong shoppers versus 12% among loyal FairPrice shoppers.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R05 · FairPrice Group to Lower and Hold Prices of Popular Protein Staples, Chicken and Eggs, Until 31 December 2026 · R06 · Sheng Siong offers 6% discount for blue and orange CHAS cards | The Straits Times. See the source roles and limits in the reference register below.
Heartland HDB families with children; Seniors aged 65+; Young couples without children; Price-sensitive singles
Switching curve by gap and segment; indicates how much price investment is enough
Price gap tolerance
Scope of this run: Stated basket split under one uniform, known price gap. Real gaps vary by item and are perceived imperfectly.
Splitting is forecast even at equal prices: 48.0% split between the stores. A 2% advantage already shifts choices: buying everything at FairPrice rises from 26.33% to 41.67% (+15.34 percentage points); buying everything at Sheng Siong rises from 13.0% to 29.0% (+16.0 points). The precise starting threshold is not established.
Full-shop concentration increases with the price advantage. At gaps of 2%, 5%, 10%, 15% and 20%, the share buying everything at the cheaper store is 41.67%, 50.0%, 55.0%, 63.0% and 66.33% for FairPrice, versus 29.0%, 41.0%, 50.33%, 51.33% and 56.33% for Sheng Siong. The largest difference between stores at matching gaps is 12.67 points, at 2%.
By segment, the first tested gap where more than half buy everything at FairPrice is 5% for young couples and price-sensitive singles, and 10% for families and seniors. For Sheng Siong, it is 10% for singles and 15% for the other three segments. These are forecast milestones, not proven minimum investments.
The largest segment difference in full-shop choice is at a 10% Sheng Siong advantage: 57.0% of singles versus 44.5% of seniors, a 12.5-point gap. For FairPrice, the largest is at a 20% advantage: 75.0% of singles versus 64.5% of seniors, a 10.5-point gap.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R03 · FairPrice vs Sheng Siong vs Cold Storage: An Honest Side-by-Side · R09 · Groceries in Singapore: FairPrice vs Sheng Siong vs Cold Storage, Where Do You Actually Save? | ShopBack Blog. See the source roles and limits in the reference register below.
Heartland HDB families with children; Seniors aged 65+; Working adults aged 25-44; Private-housing residents
Preference share for each mechanic; informs the CNY 2027 promotion mix
Promotion mechanics
Scope of this run: The four mechanics are described at equal budget but not tied to a dollar value per household.
In these model forecasts, weekly 50% member deals lead at 40%, followed by CDC voucher acceptance at 26%, spend-and-get vouchers at 16%, and price freeze at 12.33%. No preference accounts for 3.67% and don't know for 2%. Member deals lead CDC acceptance by 14 percentage points.
Member deals lead among Heartland HDB families with children (37.5%), working adults aged 25–44 (40.5%), and private-housing residents (41.5%). Seniors instead favour CDC acceptance (31.5%) over member deals (27.5%), a 4-point gap.
The largest segment difference is for CDC acceptance: 31.5% among seniors versus 15% among private-housing residents, a 16.5-point gap. Member deals vary by 14 points, spend-and-get vouchers by 8.5 points, and price freeze by 9 points across segments.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R12 · Your household just got $500 – here’s how to make it count · R05 · FairPrice Group to Lower and Hold Prices of Popular Protein Staples, Chicken and Eggs, Until 31 December 2026. See the source roles and limits in the reference register below.
Residents of mature HDB estates; Residents of new HDB estates
Share of local trade at risk by distance and best response; draft of a store-defence playbook
New-store pull
Scope of this run: Shares of shoppers diverting visits, not shares of local sales. The defence item kept a bracketed placeholder in its wording.
Distance matters: the model forecasts that 55.7% would divert at least half their otherwise-FairPrice visits to Sheng Siong at a 5-minute walk, versus 38.0% at 10 minutes, 27.3% at 15 minutes and 20.5% at 20 minutes—a 35.2 percentage-point drop from 5 to 20 minutes. These are shares of residents forecast to switch visits, not measured shares of local trade.
At 5 minutes, new-estate residents are more exposed: 54.0% would divert at least half their visits, versus 47.5% in mature estates—a 6.5-point gap. Switching risk remains at longer distances, but is lower.
Draft defence playbook: prioritise stores facing a nearby opening, then focus on lower regular prices on usual groceries. This leads retention preferences at 34.3%, ahead of frequent promotions at 17.3% by 17.0 points; better fresh-food quality follows at 10.0%. These preferences do not establish how many visits each change would save.
Tailor rather than replace that price-led response: senior discounts have greater forecast appeal in mature estates than new estates, at 9.0% versus 2.5%—a 6.5-point gap. Lower regular prices remain the leading choice in both segments.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R14 · What FairPrice Discounts Do Seniors and CHAS Cardholders Get in 2026? | ShopBack Blog · R06 · Sheng Siong offers 6% discount for blue and orange CHAS cards | The Straits Times. See the source roles and limits in the reference register below.
Seniors aged 60+; Pioneer and Merdeka Generation members; CHAS Blue card holders; CHAS Orange card holders
Preferred design and switching effect; informs the 2027 schemes, as current ones end 31 Dec 2026
Seniors and CHAS
Scope of this run: No item asks for the design preference directly; preference is read from main-store choice under each design (S$60 a year either way).
These model forecasts do not measure design preference directly. With annual savings held at S$60, everyday discounts generally predict a higher share choosing the offering supermarket than day-specific discounts, but not in every segment.
At FairPrice, everyday versus day-specific offers increase its predicted main-store share by 4.75 percentage points among Pioneer/Merdeka members, 2 points among CHAS Orange holders and 1 point among CHAS Blue holders. Seniors instead favour the day-specific arm by 2.5 points.
At Sheng Siong, everyday offers produce higher predicted main-store shares in all four segments: gains over day-specific offers range from 2 points for CHAS Blue holders to 5.5 points for CHAS Orange holders.
Both designs predict shifts towards the offering supermarket relative to usual-store forecasts. The largest increases are 11 percentage points for Sheng Siong’s everyday offer among Pioneer/Merdeka members and CHAS Orange holders. These are changes in predicted shares, not measured individual switching.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R16 · FairPrice extends discount schemes for vulnerable groups till end-2026, Singapore News · R17 · FairPrice doubles CHAS discounts to 6 per cent till 1 January 2027. See the source roles and limits in the reference register below.
Home cooks who cook most days; Larger households of 5 or more; Seniors aged 65+
Share of switching explained by fresh; indicates whether to invest in fresh counters
Fresh
Scope of this run: Share of switchers naming fresh as the main reason, not the share of switched spending.
Fresh is the leading forecast main reason for switching: 27.67% versus 8.33% for packaged goods, a 19.34 percentage-point gap. This is a share of switchers naming a reason, not the share of switched spending explained by fresh.
Fresh matters most to frequent home cooks: 31.5% name it as their main reason, versus 26% of larger households and 23% of seniors. The biggest segment gap is 8.5 points, between home cooks and seniors; fresh exceeds packaged goods by 23 points among home cooks.
In the fresh-parity scenario, 13% forecast moving no spending and 20% moving less than half; 24% would still move all of it. Home cooks are more fresh-sensitive: 17.5% would move none versus 9% of larger households, an 8.5-point gap. These categories do not establish an exact fresh-attributable spending share.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R18 · ANNUAL REPORT 2025 · R19 · ANNUAL REPORT 2024. See the source roles and limits in the reference register below.
Price-sensitive families with children; Seniors aged 65+
Willingness to trade down by category; informs housebrand range and pricing
Housebrand
Scope of this run: One price ladder per category (brand S$10 at FairPrice, S$9 at Sheng Siong, housebrand S$8); the housebrand discount is not varied.
Model forecasts favour switching to Sheng Siong for the usual brand rather than trading down to FairPrice housebrand: 50.33% versus 27% for rice, 47.67% versus 31.67% for cooking oil, and 49.67% versus 24.33% for detergent. Switching leads by 23.33, 16 and 25.34 percentage points, respectively.
Housebrand willingness is highest for cooking oil in both segments: families forecast at 34% for oil, 30% for rice and 25.5% for detergent; seniors at 29.5%, 25% and 20%, respectively.
Families are more willing than seniors to choose housebrand by 4.5 percentage points for oil, 5 points for rice and 5.5 points for detergent. The largest segment difference is staying with the usual detergent at FairPrice: 25.5% of seniors versus 13% of families, a 12.5-point gap.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R21 · NIQ: APAC Consumers Embrace Private Labels, While Name Brands Sustain Their Strength · R22 · How Much Does The Average Singaporean Household Spend Each Month On Groceries? – SSI. See the source roles and limits in the reference register below.
Chinese households (Chinese New Year); Malay households (Hari Raya Puasa); Indian households (Deepavali)
Festive chain preference and gap items; informs festive ranging for CNY 2027
Festive shopping
Scope of this run: One item covers three festivals at different times: Deepavali 2026, Chinese New Year and Hari Raya Puasa 2027.
Model forecasts put FairPrice first in all three communities: Chinese 41%, Malay 36.5%, Indian 34%, ahead of Sheng Siong by 15.5, 16 and 19 percentage points respectively. Indian households favour another supermarket more often than Chinese households: 21% versus 7%, a 14-point gap.
No range gap is expected by 51% of Chinese, 44% of Malay and 40% of Indian households—an 11-point Chinese–Indian difference. Leading gaps are festive snacks for Chinese households (10%), fresh meat/poultry for Malay households (12%), and festive snacks for Indian households (16.5%). Indian households’ cooking-ingredient gap is 11%, versus 3% for Chinese households, an 8-point difference.
With convenience held equal, lower total price is the strongest forecast pull to either chain: Chinese 44.5%, Malay 41%, Indian 40.5%. Dietary/religious range matters more to Malay (14.5%) and Indian households (13.5%) than Chinese households (1.5%)—gaps of 13 and 12 points. Better fresh-food quality is higher for Chinese than Indian households: 17.5% versus 11%, a 6.5-point gap.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R25 · Report: 77% of SG consumers turn to online ads for holiday shopping | Marketing-Interactive · R26 · Sheng Siong Deepavali Promotion 2026: Festive Grocery Deals from $1.60 (16 Oct–5 Nov) ~ All Singapore Deals. See the source roles and limits in the reference register below.
Monthly grocery spend under S$300; Monthly grocery spend S$300-699; Monthly grocery spend S$700 or more
Value placed on points vs price; informs Linkpoints redesign
Loyalty points
Scope of this run: Both arms give FairPrice a S$5 reward, so FairPrice wins either way; read the gap between the arms, not the level.
The model forecasts stronger FairPrice choice with an equal-value price cut than with loyalty points: 90.67% versus 84.33%, a 6.34 percentage-point gap. Points retain substantial appeal, but do not match the price cut in this scenario.
The price-cut advantage is largest for monthly grocery spend under S$300 (8.5 percentage points), followed by S$700 or more (8.0 points), and S$300–699 (5.0 points).
The S$300–699 segment has the highest forecast FairPrice choice with points (86.5%), 4.5 percentage points above the under-S$300 segment (82.0%). Segment differences under the price cut are smaller: a maximum of 1.0 point.
References for context and interpretation: R29 · YouGov:《价值购物者的崛起：亚太地区杂货购物报告》-发现报告 · R15 · Inside the rise of price-driven grocery shopping across APAC · R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive. See the source roles and limits in the reference register below.
Working adults aged 25-44; Families with children under 12
Online retention potential; informs the delivery offer
Online grocery
Scope of this run: The offer item is asked of shoppers who moved spending away; that condition reaches the forecast only through the wording.
Model forecasts suggest some online retention potential, not proven retention: 26% are very/somewhat likely under baseline terms, rising to 40% with free delivery. Forecast grocery drift from FairPrice stores is 23% overall—26.5% among working adults versus 23% among families, a 3.5 percentage-point gap.
Free delivery is the strongest offer overall: +14 percentage points versus baseline, compared with +10.7 for an order discount, +3 for a lower minimum and +1 for a narrower delivery window. Free delivery leads the order discount by 3.3 points.
Working adults respond more strongly to free delivery: likelihood rises from 25.5% to 48.5% (+23 points), versus 30.5% to 48% among families (+17.5 points). For families, an order discount also reaches 48%; for working adults it reaches 44.5%, 4 points below free delivery.
References for context and interpretation: R31 · What factors reduce trust amongst Singapore shoppers online? | Retail Asia · R32 · CHAS discounts: FairPrice doubles savings to 6% for Blue, Orange cardholders till Jan 1, Singapore News · R33 · Sheng Siong and foodpanda to roll out new on-demand grocery delivery | The Straits Times. See the source roles and limits in the reference register below.
Shift workers; Parents of children under 7
Effect of hours on chain choice; informs store hours
Opening hours
Scope of this run: Extended hours are a scenario; current opening hours of either chain are not modelled.
Model forecasts suggest longer hours shift chain choice toward the chain extending its hours. Overall, FairPrice’s share rises by 2.84 percentage points when it extends hours; Sheng Siong’s rises by 6.0 points when it extends hours, versus matched hours.
Shift workers show the larger response: FairPrice gains 12.5 points with extended hours, versus 2.0 points among parents of children under 7. Sheng Siong gains 16.0 points among shift workers, versus 8.5 points among parents.
The biggest difference between extended-hours arms is among shift workers: Sheng Siong’s share is 25.5 points higher when it extends hours than when FairPrice does; FairPrice’s share differs by 25.0 points in the opposite direction.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R36 · Retail in Singapore · R37 · 24 Hour Supermarket in Hougang, Sengkang & Punggol (2026). See the source roles and limits in the reference register below.
Multi-store shoppers; Deal-seekers who follow weekly promotions
Size of the split-shopper group and categories lost; full deal-cherry-picking analysis waits for phase 2
Split shopping
Scope of this run: Category destination is the most recent purchase, not share of spend. Deal cherry-picking waits for phase 2.
Model forecasts put shopping across two or more chains at 58.3%, versus 38.0% using one chain. Among weekly-promotion deal-seekers, 76.5% use multiple chains—18.2 percentage points above the overall forecast.
FairPrice leads as the main chain at 50.3%, ahead of Sheng Siong at 26.0%. Category destinations vary: FairPrice takes 42.7% of dairy/eggs but 24.3% of fresh seafood, an 18.3-point gap.
The clearest destinations outside supermarket chains are wet markets for seafood (23.7%), meat (20.7%) and produce (19.7%); independent shops for bakery (21.0%); and online-only retailers for cleaning/laundry (10.0%). Wet-market seafood is 20.7 points above dairy/eggs. These destination shares do not establish losses from a shopper’s main chain.
The largest category-destination gap between the two segments is FairPrice snacks: 38.0% among deal-seekers versus 32.0% among multi-store shoppers, a 6.0-point difference.
References for context and interpretation: R02 · Report: 63% of Singaporean grocery shoppers welcome new brands despite unfamiliarity | Marketing-Interactive · R29 · YouGov:《价值购物者的崛起：亚太地区杂货购物报告》-发现报告 · R41 · Supermarket unit pricing pilot to be expanded from Oct 1 to cover more operators, grocery items. See the source roles and limits in the reference register below.
This is an exploratory synthetic study, not fieldwork. It forecasts answer distributions for thirteen client questions to find hypotheses and sensitivities before a human study.
The audience is Singapore resident adults (citizens and permanent residents) who do some or all of their household's grocery shopping. A Nemotron-Personas-Singapore frame weighted to SingStat 2026 age-by-sex cells was built, but no topic needed per-person answers, so it was not used to generate respondents.
Segments such as “recent switchers” or “CHAS Orange card holders” are text descriptions given to the forecaster. Their size in the population is not estimated and they overlap.
The harness turned each objective into one to three single-choice items with neutral wording, a don't-know option where a real survey would have one, and explicit hypothetical scenarios. Offers, price gaps, distances and hours are scenario arms; each respondent would see one.
Items were rule-checked, repaired and frozen before any forecast (plan df0985b6c85b). Prices, distances and offers in scenarios are experimental assumptions, not current facts.
For each topic, three to six web searches limited to pages published before 25 October 2026 were condensed into dated evidence notes. A citation gate keeps only exact retrieved URLs.
gpt-6.1-sol forecast each item for all shoppers (three answer orders averaged) and for each segment (two orders averaged): 595 model calls and 63 searches. Forecasts are model runs, not respondents.
No calibration was fitted: no Singapore grocery survey with known results was available to fit and test against. Each item carries an expected error from a backtest on 402 published survey questions (UK/EU and SEA), its spread across answer orders, its evidence status and risk flags.
Grades describe process risk, not accuracy on this study. Historical checks against 60 published Singapore grocery results (below) are diagnostics, not validation. Next: a short preregistered human pulse with a reserved validation set before any calibration.
| Diagnostic | Items | Denominator |
|---|---|---|
| Grade A / B / C | 1 / 8 / 20 | 29 |
| Hypothetical scenario, no real-world benchmark | 14 | 29 |
| Unstable across option orders (spread > 6 pts) | 14 | 29 |
| No usable dated evidence; rests on model knowledge | 14 | 29 |
| Filtered base: check who answers | 10 | 29 |
| Wording rule failing at freeze (rules added after the run) | 1 | 29 |
We forecast 60 published results from 38 Singapore grocery survey items (Consumers Association of Singapore, Milieu Insight, National Environment Agency, NielsenIQ, Rakuten Insight, SMU Centre for Research on Successful Ageing, YouGov; 2019–2026) without access to the publishers' own pages and with search limited to pages published before each survey's fieldwork. Full answer distributions came out at 9.7 TVD points against 24.1 for equal shares; single published shares at 12.8, close to a 50/50 guess; multi-select options at 10.6. This is not accuracy on the FairPrice questions: none of these surveys asked them, most were published before the model's training cutoff (so it may have seen them), and some figures come from news reports of the survey rather than the publisher.
| Check type | Items | Equal-shares guess | EggAI-v0.1 | Difference (95% CI) |
|---|---|---|---|---|
| Full answer distributions | 7 | 24.1 | 9.7 | -14.5 [-20.6, -6.5] |
| Partly published distributions (+ unpublished remainder) | 4 | 22.9 | 21.7 | -1.2 [-11.2, +14.2] |
| Single published share (vs everything else) | 16 | 13.8 | 12.8 | -1.0 [-6.8, +4.8] |
| Multi-select options (chosen vs not) | 33 | 12.8 | 10.6 | -2.2 [-6.7, +2.2] |
| All checks | 60 | 15.1 | 11.8 | -3.2 [-6.6, +0.1] |
TVD points: half the summed absolute gap between forecast and published shares; lower is better. For two-option checks it equals the gap in the named share.
Each row is one published share. The line joins the forecast to the published figure; rows are grouped by topic.
| Survey | Option · published · forecast | TVD |
|---|---|---|
| YouGov 2024-06 Among those who used customer reward programs in the past three months: how many reward programmes did you regularly use |
| 3.0 |
| YouGov 2024-06 Among regular egg consumers (buy at least once a month): do you typically prefer locally sourced or imported eggs? |
| 5.0 |
| YouGov 2024-06 Among regular fish/seafood consumers: do you typically prefer locally farmed or imported fish and seafood? |
| 8.7 |
| YouGov 2023-06 How has your disposable income changed over the past 12 months? |
| 8.7 |
| YouGov 2024-06 Among regular vegetable consumers: do you typically prefer locally grown or imported vegetables? |
| 9.3 |
| YouGov 2021-10 Do you prefer buying groceries remotely (online) rather than in store? |
| 11.7 |
| Milieu Insight 2026-01 Compared with last year, do you expect your Lunar New Year spending to increase, stay about the same, or decrease? |
| 13.3 |
| Rakuten Insight 2020-04 Change in shopping frequency at Supermarkets (compared with last year) |
| 16.7 |
| Consumers Association of Singapore 2026-02 Beyond the Budget measures, what would provide the greatest relief? |
| 16.8 |
| Milieu Insight 2025-01 How do you approach discounts/promotions on CNY items? |
| 22.7 |
| Milieu Insight 2023-05 How has your household grocery budget changed? |
| 38.7 |
| Validation grade | Checks | Mean TVD |
|---|---|---|
| Grade A | 36 | 10.9 |
| Grade B | 12 | 13.7 |
| Grade C | 12 | 12.8 |
73% of checks fell inside the card's 80% bound. Grade A checks missed by less (10.9) than grade B and C (13.7, 12.8), but B and C are not separated; treat the grades as a coarse split, not a ranking. Grades use the card's Southeast Asia error model.
Not covered by any public survey we could find: which chain shoppers think is cheaper, why they choose a supermarket, and opening hours. Those topics remain unchecked.
Search the exact question text, base and response options used in this run. This is a transparent instrument record, not a recommended final human questionnaire.
| Question | Method | Base | Exact wording | Response options |
|---|
Public pages retrieved before fieldwork and cited by the forecasts. They give context; none is a survey of these questions, so none validates the results.
EggAI-v0.1-harness · forecaster gpt-6.1-sol · run fp_v3 · 26 October 2026 (Asia/Bangkok). Model outputs are for research and reporting only.