Nok Roast: Kiosk Economics and Cannibalization
Nok Roast is a Chiang Mai specialty coffee roaster with 38 company-owned cafés in Thailand. FY2025 revenue was THB 640 million at an average ticket of THB 178; gross margin is 66% and store-level EBITDA margin 11%. New cafés cost THB 4.5 million and must pay back within 24 months. Same-store sales fell 4.2% while six new stores grew total revenue 3%, and app data show 210,000 members but only 31% active in the last 90 days. Management proposes a 12-square-metre 'Nok Express' kiosk format in Bangkok office towers: capex THB 1.8 million, forecast 140 cups/day at THB 95, 62% product margin, monthly rent THB 45,000, labour THB 55,000 and other fixed costs THB 25,000. The CFO likes the 18-month payback; the COO worries the kiosks will steal sales from two nearby cafés that average THB 178 per ticket. The board wants a decision within six weeks, before lease commitments.
1. Calculate the monthly break-even volume and standalone payback for Nok Express. Then estimate the cannibalization rate at which the format stops creating value, given the nearby cafés' economics. State your assumptions.
Read the worked answer
Break-even. Contribution per cup = THB 95 × 62% = THB 58.90. Monthly fixed = 45,000 + 55,000 + 25,000 = THB 125,000. Break-even = 125,000 ÷ 58.90 = 2,122 cups/month, or about 71 cups/day. Forecast 140/day = 4,200/month, contribution THB 247,380, store EBITDA THB 122,380/month. Standalone capex payback = 1,800,000 ÷ 122,380 = 14.7 months, inside the 24-month hurdle.
Cannibalization. Each cannibalized café ticket loses THB 178 × 66% = THB 117.48 contribution. If x share of 4,200 kiosk cups comes from Nok cafés, lost contribution = 4,200 × x × 117.48 = 493,416x. Setting 493,416x = 122,380 gives x = 24.8%. Above roughly 25% cannibalization, the kiosk adds no store-level EBITDA.
Because two cafés are nearby and the kiosk price is about 47% lower, 25% is plausible. The standalone payback is therefore not the decision metric; use incremental contribution after cannibalization. The CFO's 18-month payback is an upper-bound, best case.
2. Diagnose the same-store sales decline. Distinguish between market-level demand, competitive, and execution causes, and name the two data cuts you would request first.
Read the worked answer
Decompose same-store sales as traffic × conversion × ticket. Ticket is THB 178; if ticket is stable and transactions fell 4.2%, the problem is traffic or conversion, not price.
Request two cuts first: (1) store-level transactions by daypart and week, split by trade-area type (office, mall, residential), to see whether the decline is broad or localized; (2) app/member frequency and new-versus-returning customer counts, to separate market shrinkage from retention failure.
Then test three hypotheses. Market demand: category traffic down, for example lower office occupancy. Competitive: new specialty entrants within 500 metres. Execution: service speed, stockouts, staff turnover, review scores. A broad decline across all trade areas points to category or brand weakness; a localized decline points to competition or store execution. The 31% active member rate suggests a retention problem independent of new-store growth.
Recommendation: fix retention and CRM before adding kiosk capex, because the same demand weakness would depress kiosk forecasts.
3. Design a six-week test that would let the board approve a rollout without guessing the cannibalization rate. Specify unit of test, metrics, and decision rule.
Read the worked answer
Run a 12-week matched-pair test rather than six weeks if leases allow; six weeks is enough only for a rough read. Use eight kiosk sites: four inside trade areas with an existing café within 500 metres and four in areas without one. Match on office-worker density, daypart traffic and income.
Metrics: kiosk cups/day and AOV; café transactions and ticket in exposed and matched control areas; gross contribution; app sign-ups; and customer overlap identified through the loyalty app.
Decision rule: roll out only if incremental contribution (kiosk contribution minus measured café loss) covers capex at a 24-month payback in at least six of eight sites, and if the 95% confidence interval on cannibalization is below 25%. Use differences-in-differences against matched control cafés not exposed to a kiosk. Pre-register the threshold before the test to avoid post-hoc rationalization.