Methodology

How the Forkcast tools are calibrated

Every estimate on Forkcast is grounded in a specific dataset and refresh cadence. This page documents the sources, limitations, and how we keep the numbers honest.

Data sources

  • Agmarknet (Govt. of India)
    Daily mandi prices for 12+ commodities (onion, tomato, potato, ginger, garlic, oil, etc.). Refreshed every 24 hours. Used in the menu pricing tool's mandi-priced ingredient line.
  • Capital benchmarks dataset
    Curated data from 200+ Indian restaurant openings (2023-2026), cross-checked against Petpooja hardware partner quotes. Refreshed quarterly. The capital estimator uses per-format drivers (sqft, rent band, fit-out tier, headcount) calibrated to published format bands — not fixed percentage splits.
  • Cuisine listing density
    Aggregated Zomato + Swiggy listings by pincode, normalised to city population. Refreshed monthly.
  • State licence portals
    FoSCoS (FSSAI), state excise departments, MCG/PMC/BMC trade licence schedules, fire-services NOC fee schedules. Refreshed when policy changes are notified.
  • IMD weather
    India Meteorological Department forecast data, used in the in-product daily forecast (not exposed in free tools).

Calibration cadence

  • Daily: Agmarknet mandi prices.
  • Weekly: Tool usage anonymous QA — outlier outputs flagged and reviewed.
  • Monthly: Cuisine listing density refresh; viability and demand heatmap recalibration.
  • Quarterly: Capital benchmark dataset refresh; format and tier coefficients re-fit.
  • On policy change: Licence cost, validity, and authority updates within 7 working days.

Limitations we're upfront about

  • Capital estimates are driver-based defaults you can edit line-by-line. Real openings can still fall outside expectations when the format is unusual or location is exceptional.
  • The location scorer falls back to a heuristic when GOOGLE_PLACES_API_KEY isn't configured. The result page tells you which source was used.
  • Cuisine demand is a directional read — region-specific cultural patterns (e.g. Konkan coastal seafood, Saoji in Nagpur) aren't fully captured at the slug level. Use it alongside in-person catchment validation.
  • Mandi prices are reference rates, not your supplier's quoted rate. A 5-10% delta is normal; flag >15% deltas for re-quote.
  • Licence fees are state-issued; some PMC/BMC sub-classifications can swing the number 30-50% within the published band.

Capital estimator formulas

  • Interiors: sqft × fit-out cost per sqft (lean / mid / premium) × city tier multiplier.
  • Deposit: sqft × rent per sqft × deposit months (format-specific, typically 4–12).
  • Equipment: format base + per-sqft adder, scaled by city tier.
  • Licences: sum of state + format checklist from official fee schedules; fire-install add-on when applicable.
  • Working capital: 1.5× monthly fixed OPEX (editable).
  • Total funding: total CAPEX + pre-opening OPEX buffer (default 3 months of fixed costs).

How we keep ourselves honest

Each tool publishes its formula and source list inside the FAQ on the tool page. When we change a coefficient or add a new dataset, we note it in the next blog issue. Pilot operator feedback drives the weekly QA loop — if a tool's output disagrees with their lived experience, we investigate.

Comments, corrections, and dataset suggestions are welcome — write to hello@forkcast.in. We acknowledge every email and credit substantive corrections in the next monthly digest.

Open the free tools, read the complete opening guide, or browse the blog for the latest applied research.

Forkcast methodology — how we calibrate the free tools | Forkcast