Skip to content

The Cloud Kitchen That Grows and Bleeds

Orders double every year. So do the losses.

moderate
9 min read
unit-economicsfood-delivery

The Prompt

Your client runs 25 cloud kitchens across three metros, selling four house brands exclusively through food-delivery apps. Order volume has doubled for two consecutive years, yet the company has never made money and losses are widening. The founder wants a path to profitability within 12 months.

Opening exchange

candidate

Three clarifications. Is the loss widening in absolute terms only, or is the loss per order also widening? Are all 25 kitchens loss-making, or is there a spread? And are we free to change channels — for example, our own app or dine-in — or must we stay on the aggregators?

interviewer

Loss per order is roughly flat, around ₹18. About a third of kitchens are contribution-positive. Channels are open to discussion, but 95% of demand currently comes from two aggregators.

candidate

Flat loss per order with doubling volume means we are scaling a broken unit economic — fixed costs are not the story. I'll focus on the per-order P&L first, then use the spread between good and bad kitchens as a natural experiment.

Restates what the data implies before structuring — this is what separates a hypothesis-led candidate.

Structure & Hypothesis

Build the order-level P&L, then split the network: what do the contribution-positive kitchens do differently?

The order P&L waterfall, then the density band — ₹87 pre-opex contribution vs ₹19k/day opex makes the whole case arithmetic.

Analysis & Data

interviewer

Here's the twist: the contribution-positive kitchens have the same commission, COGS, and discounts. What could they be doing differently?

candidate

If the per-order lines are identical, the difference must be the divisor — order density. Kitchen opex is semi-fixed, so kitchens doing more orders per day spread rent and staff thinner. I'd check orders per kitchen per day across the network.

Works the only remaining free variable instead of guessing.

interviewer

Correct. Positive kitchens average 210 orders/day; negative ones average 90. Staffing is nearly identical across both.

candidate

So the model works at ~200 orders/day and fails at 90. The question becomes: can the 90-order kitchens get to 200, or should they close? I'd segment them — some are young and ramping, some are in micro-markets that will never support four brands, and some may have an operational issue like poor app ratings.

Recommendation

Recommend to the founder

  • Triage the 17 sub-scale kitchens: ramp (marketing push, brand mix change) the ones in dense micro-markets; convert thin markets to a 2-brand, 4-staff format; exit the bottom ~5.
  • Attack the ₹140 aggregator take: negotiate volume-tiered commissions and shift the top 10% repeat customers to a direct-order channel with a loyalty hook.
  • Cut discounts from blanket to surgical — fund them only on first orders and dead hours, not on repeat orders that would come anyway.
  • Make orders/kitchen/day the operating metric the network is run on; review weekly.

Key Takeaway

What this case teaches

When loss per unit is flat while volume scales, fixed-cost dilution is not coming to save you. Find the density or take-rate lever — and use the spread between good and bad units as your fastest diagnostic.