branch-locator Example report
BL-2026-001

Ayam gepuk opens where young, Malay-majority neighbourhoods meet fast food

We mapped the 265 ayam gepuk outlets we could find (four chains’ own lists plus OpenStreetMap) and compared the 2 km around each one with the 1,113 KFC, McDonald’s and Marrybrown outlets in the same states. Population and income look the same as fast food. Age and ethnicity do not.

Outlets studied: Pak Gembus 110, Top Global 48, Pak Gendut 44, Artisan 36, independents 27. Built from public and aggregate data only; no sales data was used.

Figure 1. Every ayam gepuk outlet we found (Pak Gembus lime, Top Global orange, Pak Gendut pink, Artisan blue, independents yellow) and the fast-food benchmark (grey). Pak Gendut’s two Sabah outlets are off this view. 52 outlets are placed at a postcode or town centre; their pop-ups say so. The grid follows a street directory: columns A–L west to east, rows 1–12 north to south. Grid references in this report pan the map to their square.

It is a Klang Valley product

More than half of all outlets sit in Selangor, Kuala Lumpur and Putrajaya. Fast food, a mature category that has reached every state, puts only a third there. Ayam gepuk is under-represented in Johor (6% against 15%) and Penang (1.5% against 7%). East Malaysia is almost untouched: two Pak Gendut outlets in Sabah (0.8%, against 6.5% of fast food) and none in Sarawak or Labuan.

That gap is either untested ground or weaker demand. Public data cannot tell which; the chains’ own sales by state can.

Ayam gepuk, share of outletsFast food, share of outlets

Figure 2. Share of each group’s outlets by state. The line joins the two shares; a long line means ayam gepuk is placed very differently from fast food.

Show the data

Inside a state, the crowd is the same size and income, but younger and more Malay

Comparing outlets across states would mostly measure the Klang Valley effect above. So each outlet is ranked only against fast-food outlets in its own state. A score of 50 means “exactly like where fast food opens”.

Population and district income both land at the 48th percentile: ayam gepuk goes where fast food goes. (Income covers 225 outlets: the district join fails for Kuala Lumpur, Putrajaya and a few districts in Penang, Perlis and Perak.) The difference appears only at polling-district grain. Around the median outlet, 30.2% of voters are aged 18–29, against 26.8% around fast food, and 56% are Malay, against 40%.

District figures show no youth difference (50th percentile); polling-district figures do. The two also differ in source and age band (residents aged 15–24 against registered voters aged 18–29), so finer grain is the likely reason, not a proven one.

Median outlet, significant (p < 0.05)Not significantMiddle half of outlets

Figure 3. Where the median ayam gepuk outlet ranks among fast-food outlets in the same state (0–100). The bar spans the middle half of outlets. The dashed line at 50 means no difference.

Show the data

Young neighbourhoods set it apart; students barely do

Youth and Malay share move together (r = 0.43), and campuses sit in young areas. To separate them, one model weighs all the factors at once, within each state. The youth share of nearby polling districts has the strongest pull after fast-food density. Once youth is known, Malay share adds little the data can detect (p = 0.26), and so does enrolment within 2 km (public campuses p = 0.07, private colleges p = 0.35). Not significant is not zero: the intervals still allow modest effects.

Outlets are near campuses more often (17% have a public campus within 2 km, against 8% of fast food), but that gap largely disappears once youth and commercial density are taken into account. The UiTM survey that drew attention to ayam gepuk interviewed students (94% of 386 respondents). That tells us who answered the survey, not where the product sells.

Significant (p < 0.05)Not significant95% confidence interval

Figure 4. Log-odds, per standard deviation, that a restaurant site is ayam gepuk rather than fast food (youth +0.46 is about 1.6 times the odds). Logistic regression with state fixed effects and robust errors; enrolment is log-transformed before scaling; 255 ayam gepuk and 997 fast-food sites in the same states. It describes where outlets are, not whether a site will succeed.

Show the data

The chains sit on top of each other

118 of 265 outlets have another ayam gepuk outlet within 1 km (85 of 213 counting only outlets with an exact location). Across the four chains, 90 of 238 outlets have a rival chain within 1 km: Top Global most often (24 of 48), then Artisan (17 of 36), Pak Gendut (16 of 44) and Pak Gembus (33 of 110). In E8 Bukit Jelutong and F8 Setiawangsa a Top Global and a Pak Gembus are 47 m apart.

Clustering can mean a destination strip that feeds both, or two outlets splitting one crowd. Location data cannot separate the two. Outlet sales before and after a rival opens, set against outlets that gained no rival, can, and that is the first test we would run with a client.

Figure 5. Distance from each ayam gepuk outlet to the nearest other ayam gepuk outlet, any brand.

Show the data

What we would test next

With a chain’s own outlet sales, the same catchments become a model of what makes an outlet succeed, not only where outlets are:

  • Test whether polling-district youth share predicts outlet sales, not only where outlets are.
  • Score candidate sites against the outlets that sell best. Without sales, the same method already ranks 384 fast-food sites with no ayam gepuk nearby by resemblance: see the candidate sites.
  • Measure cannibalisation where outlets are under 1 km apart before adding another in E8 or F8.

Limits

  • Catchments are 2 km circles; drive-time catchments are the next version.
  • Voter figures count registered voters aged 18+ at their registered address (PRU15 roll, 2022). Students registered at home are counted there, not near campus, and today’s 18–24s were mostly under 18 in 2022; the direction of the resulting bias is unknown. Age and ethnicity are separate counts, so no “young Malay” segment can be isolated.
  • Student counts are allocated to campuses by rule; MOHE publishes no per-campus figures.
  • OSM under-maps independents and the fast-food benchmark in some states; two independents may duplicate a chain outlet.
  • Income is a district value, not a catchment value, and is missing for 40 outlets (all 30 in Kuala Lumpur).
  • 52 outlets are placed at a postcode or town centre, 1–3 km from the true spot. Without them the results hold: youth at the 63rd percentile, Malay share at the 65th, youth +0.50 per SD in the model.
  • Nearby outlets share catchments; robust errors do not correct for that spatial clustering, so p-values are optimistic.

Data and method

  • Outlets: Pak Gembus and Top Global store locators, Artisan’s outlet list and Pak Gendut’s branch posters (Sep 2026), plus OpenStreetMap. 146 carry the chain’s own coordinates and 31 OpenStreetMap’s; 35 are geocoded to the street, 50 to a postcode centre, 2 to a town centre, 1 placed by hand.
  • Benchmark: KFC, McDonald’s and Marrybrown from OpenStreetMap, deduplicated within 50 m.
  • Residents: WorldPop 2020 constrained population, 100 m grid.
  • Age and ethnicity: SPR electoral roll, PRU15, aggregated to 7,748 polling districts (counts only, cells under 10 suppressed), area-weighted into each 2 km circle.
  • Income: DOSM HIES 2024, district median. Students: MOHE Statistik Pendidikan Tinggi 2025 (Tables 1.3, 2.1, 4.1).
  • Tests: within-state percentile against the benchmark (Wilcoxon), and logistic regression with state fixed effects and robust errors.
  • Context: Mohamed Apandi et al., Jurnal Intelek 21(2) 2026, doi 10.24191/ji.v21i2.11394.