Inside MarginMind™: How KhanaOS Found 61 Loss-Making Dishes on a 130-Dish Bengali Restaurant Menu

Ranjit-da had been running Hangla Heshel for eleven years.

The restaurant is a proper Bengali adda — long wooden tables, handwritten specials on a blackboard, a menu that runs to 130 dishes across rice, fish, meat, vegetarian, snacks, and sweets. It does 85–110 covers a day on weekdays and closer to 160 on weekends. Ranjit-da knows his kitchen. He can taste a dal and tell you it's under-tempered. He can feel, by the weight of a lunch rush, whether the day will finish in the green.

What he could not do — what almost no restaurateur can do, because the tools have never existed to make it possible — was tell you, at the dish level, which items on his 130-dish menu were making money and which were quietly destroying it.

The MarginMind™ onboarding session took four hours and a shared Google Sheet. By the time Ranjit-da had finished his morning tea, 61 dishes were flagged. Three were named as critical.

This article walks through exactly what happened — and what the numbers mean for any restaurant with a menu longer than 40 dishes.

What Is MarginMind™ — And Why We Built It

MarginMind™ is the menu intelligence layer inside KhanaOS. It answers a question that every restaurant owner asks but almost none can actually answer: which dishes are making me money, and which ones am I subsidising with my profits from the dishes that are?

Most billing software — and we include the expensive ones in this — stops at revenue. You can see which dishes sell the most. You cannot see which dishes earn the most. Those are completely different lists, and confusing them is one of the most expensive mistakes in food service.

MarginMind™ works by building a fully-loaded cost for every dish on your menu: ingredient cost (with yield adjustments), direct overheads (staff, utilities, packaging), and indirect overheads (rent, EMIs, insurance). It then maps every dish onto a profitability-and-volume matrix, identifies which quadrant each dish belongs to, and gives you a set of specific, actionable recommendations for each.

It is not a self-serve calculator. MarginMind™ is a managed service. The KhanaOS team runs your onboarding session, builds your recipe master, applies your overhead structure, and delivers a live dashboard. You do not need to know what a "yield factor" is. You need to know what to do with the output.

The Dashboard — What You See on Day One

When Ranjit-da opened his MarginMind™ dashboard for the first time, the first screen he saw was the Profitability Summary: a single number (his blended food cost %), a trend line showing the last 30 days, and a red badge that said 61 dishes flagged.

Below that: three cards, each one a dish name in red.

• Prawn Malai Curry — food cost 71%. Fully-loaded loss: ₹830 per serving.
• Special Chingri Bhapa — food cost 68%. Fully-loaded loss: ₹620 per serving.
• Chef's Ilish Platter — food cost 59%. Fully-loaded loss: ₹410 per serving.

Ranjit-da's first reaction was that the numbers must be wrong. His second reaction — after we walked through the recipe card together — was silence. Then: "But this dish always gets ordered."

That is the most common response. High volume does not imply high profit. A dish that sells 40 portions a day at a loss of ₹830 each is not a popular dish — it is a ₹33,200-per-day liability.

The dashboard had surfaced this fact in four hours. It had been hiding in plain sight for eleven years.

Understanding the Four Quadrants

The core of MarginMind™ is the Menu Engineering Matrix — a 2×2 grid that plots every dish by two axes: contribution margin (how much profit the dish generates per serving) and popularity (how many portions you sell).

This framework has existed in hospitality academic literature since the 1980s, developed by Kasavana and Smith at Michigan State. What MarginMind™ adds is the automation, the yield-adjusted ingredient costing, and the overhead loading that most implementations skip entirely.

⭐ Stars — High Margin, High Volume

Stars are the engine of your restaurant. At Hangla Heshel, the Stars were concentrated in a perhaps unexpected place: the rice dishes.

Mutton Biryani sold 28 portions on a typical weekday at a net contribution of ₹148 per portion. Prawn Fried Rice sold 34 portions at ₹72 net. Neither dish was Ranjit-da's most celebrated offering — the fish dishes carry more prestige on the menu — but they were quietly funding the operation.

The prescription for Stars is simple: protect them, promote them, and do not mess with them. Do not change the recipe to save cost. Do not drop the portion size. Put them where the eye goes first on the menu — upper-left or centre, above the fold on a printed card.

🔷 Puzzles — High Margin, Low Volume

Puzzles are the opportunity dishes. They make good money per portion but are not selling enough. At Hangla Heshel, the Puzzles included Mochar Ghonto (banana flower curry), Cholar Dal (Bengal gram dal with coconut), and three other traditional Bengali preparations.

These dishes have low volume for one of a few reasons: they are buried on the menu, they are not described compellingly, they are unfamiliar to newer customers, or they are priced at a point that signals "expensive" without enough context.

The prescription for Puzzles: surface them. Feature them as chef's specials. Give them a story on the menu. Price them with context — a note about preparation time or ingredient sourcing helps customers understand value.

🐎 Plowhorses — Low Margin, High Volume

Plowhorses are the workhorses of the menu. They sell well but don't make much money per portion. At Hangla Heshel, the Plowhorses included Chicken Curry (food cost 41%, contribution ₹28), Dal Tadka (food cost 38%, contribution ₹22), and most of the bread items.

These dishes are psychologically difficult to reprice — they are often the items customers feel they "know" the price of, and a visible increase creates friction. The solution is not always repricing. It is often bundling.

Pairing a Plowhorse with a high-margin item — a combo that includes Dal Tadka with a Puzzle-category Cholar Dal and a bread — can lift the average check per cover without the customer feeling a price increase.

🐕 Dogs — Low Margin, Low Volume

Dogs are the dishes that need a decision. They neither sell well nor make money. At Hangla Heshel, 23 dishes were classified as Dogs after overhead loading.

The prescription for Dogs is not automatically "remove them." Some Dog dishes carry menu completeness value — a menu without any sweet dishes, for example, is incomplete regardless of their margin profile. Some Dog dishes have loyal customers who would notice their absence.

But most Dogs on most menus exist because they were added at some point and never evaluated. They increase prep complexity, extend cook time variance, and dilute kitchen focus. The default prescription is to remove or consolidate, then monitor whether anyone notices.

At Hangla Heshel, 11 Dog dishes were removed from the menu over a six-week period. Ranjit-da had a list of five he was certain would generate complaints. None of them did.

The Overhead Module — The Numbers Nobody Talks About

Most recipe costing tools calculate food cost. MarginMind™ calculates fully-loaded cost — and the difference is where the real insights live.

Consider Hangla Heshel's cost structure:

That ₹173.20 is the overhead cost that lands on every single cover the restaurant serves, before a single ingredient is purchased. It is not a fixed monthly expense that sits somewhere in the P&L and disappears. It is a per-cover cost, and it must be recovered from every table.

Most restaurateurs who have never done this calculation are shocked by the overhead-per-cover number. The temptation is to think of overheads as separate from dish profitability — the rent gets paid, the staff gets paid, and then whatever food sales generate on top is the result. MarginMind™ forces a more accurate framing: you need to recover ₹173 in overhead contribution from every cover before you make a single rupee of profit.

The Full Dish Cost Card

Here is the Mochar Chop — a Bengali deep-fried banana flower cutlet, Hangla Heshel's signature starter — as MarginMind™ sees it.

Selling price: ₹180

At a food-cost-only view, Mochar Chop looks excellent — 24.4% is well within target. It is exactly the kind of dish that gets ignored in a conventional costing exercise. It looks fine. It feels fine.

But fully loaded, the dish is contributing ₹35 per portion. That is not bad — until you run the Ilish Paturi next to it.

Ilish Paturi (Hilsa wrapped in banana leaf, steamed):

The dish was priced at ₹680 when it was added to the menu three years ago. Hilsa prices have increased by approximately 40% in that period. The price was never revised.

The dish felt fine. It was not fine. It was losing ₹254.60 on every serving, before overhead.

The What-If Price Simulator

After the diagnosis comes the question every owner asks: what do I price it at?

MarginMind™ includes a what-if price simulator that runs two parallel calculations:

Method A — Food cost target method: What selling price achieves your target food cost % (say, 32%)?

Method B — Full contribution method: What selling price recovers all overhead and achieves a target net contribution (say, ₹80 per portion)?

For Ilish Paturi:

• Method A (32% food cost target): ₹2,380 — clearly not viable for a menu where the current price is ₹680.
• Method B (₹80 contribution at current overhead): ₹1,095

Move the slider. The simulator shows you the contribution at every price point between ₹680 and ₹1,500 in real time. At ₹890 — a 31% increase — the dish breaks even on a fully-loaded basis. At ₹1,050, it contributes ₹80.

The decision for Ranjit-da was not a pricing decision. It was a dish decision. Ilish Paturi at ₹1,050 would still sell — Hilsa has no acceptable substitute, and his customers know it. But it would sell fewer portions, which might actually be acceptable: fewer loss-making portions is better than more.

The dish is now priced at ₹980 and sells 6–8 portions per day instead of the previous 11–13. Revenue from this dish dropped by ₹2,800 per day. Losses dropped by ₹3,300 per day.

The Ingredient Master — Yield Adjustments

One of the least visible but most consequential features of MarginMind™ is the yield-adjusted ingredient cost calculation.

Most costing systems use purchase price. If you buy onions at ₹40/kg, they cost ₹40/kg in your recipes. This is wrong — not in theory but in practice — because not all of what you purchase ends up in the dish.

At Hangla Heshel, the yield factors that most affected the cost calculations:

The Hilsa example is the starkest. At ₹2,200/kg purchase price, a system that ignores yield would say the fish costs ₹2.20/g. With a 65% yield — bones, head, gills, and trimming account for 35% of the purchased weight — the effective cost is ₹3.38/g. That 54% understatement flows through into every recipe that uses Hilsa, and into every food cost calculation that is built on those recipes.

What Happens After the Session

The onboarding session ends. The dashboard stays live.

This is the part that most restaurant profitability exercises miss. A one-time costing exercise is a photograph. MarginMind™ is a live feed.

What the KhanaOS team does after the onboarding:

Weekly: Wholesale price indices for key ingredient categories (fish, vegetables, dairy, oil) are monitored. When a significant price movement occurs — Hilsa prices shifting with season, mustard oil spiking, vegetable prices varying post-monsoon — the team updates the ingredient master and the dashboard recalculates automatically.

Monthly: Each restaurant on MarginMind™ receives a Monday Margin Report — a one-page summary of the previous month's quadrant shifts (which dishes moved, in which direction, and why), flagged cost alerts, and three specific recommendations for the coming month.

Quarterly: A review call with the restaurant owner to walk through any menu changes, pricing decisions taken, and a comparison of the current vs. original quadrant map.

Ranjit-da's Monday reports have become part of his weekly routine. He reads them before the lunch service. The weekly cost alerts — a text message when any ingredient's effective cost moves more than 8% — mean he knows about price changes before they affect his margins, not six weeks later when he notices the P&L is thinner than expected.

Is MarginMind™ Right for Your Restaurant?

MarginMind™ is not for every restaurant. Here is an honest assessment.

MarginMind™ is ideal for:

Restaurants with 50+ dishes: The value of the quadrant analysis multiplies with menu complexity. A 20-dish menu can be costed manually in an afternoon. A 130-dish menu cannot.

Multi-category menus: Restaurants that serve starters, mains, breads, desserts, and beverages — where overhead allocation across category types creates meaningful cost differences.

Chain operators and multi-outlet businesses: At two or more outlets, the ability to compare menu performance across locations, standardise recipes, and identify outlet-specific margin issues becomes extremely valuable.

Restaurants where the chef is also the owner: If you are running the kitchen yourself, you do not have a costing team. MarginMind™ is the costing team.

Restaurants that have grown: If you started with 30 dishes and are now at 90, your original price points are probably based on intuition built in a smaller, simpler business. A full profitability audit typically reveals that expansion has introduced cost creep that was never explicitly priced in.

MarginMind™ is not ideal for:

Single-item stalls and very small kiosks: If you sell three things, you can cost them in an hour. You do not need a managed service.

Restaurants that change their entire menu weekly: If your menu is completely fluid, the recipe master requires constant rebuilding that does not make sense as a managed service. MarginMind™ works best when there is a stable core menu, even if specials rotate.

Businesses that cannot or will not adjust prices: The tool surfaces opportunities. If the output cannot be acted upon — because of competitive pricing constraints or lease agreements tied to specific price points — the insight has limited operational value.

FAQ

How long does the MarginMind™ setup take?

The typical setup takes 48–72 hours from the point at which you share your menu, recipe information, and cost data. The onboarding session itself is 3–4 hours with a member of the KhanaOS team. Most of that time is spent building and validating the recipe master, not collecting data.

Do I need to provide all my recipes in a specific format?

No. We work with whatever format you have — a WhatsApp photo of a handwritten recipe book is fine. The KhanaOS team structures the data. You validate the quantities and methods.

What happens when ingredient prices change?

You can notify the team directly, or we monitor commodity price indices for key categories and proactively update when we see significant movements. The dashboard recalculates the moment the ingredient price is updated — you do not need to re-run anything.

Is this the same as the food cost calculator on your resources page?

No. The food cost calculator is a simple self-serve tool for rough estimates — you enter a cost and a price and it tells you the food cost percentage. MarginMind™ is a fully-loaded profitability engine that accounts for yield, overhead allocation, and menu-wide portfolio analysis. They serve completely different purposes.

Can I use MarginMind™ if I already have KhanaOS POS?

Yes — and this is the optimal configuration. When KhanaOS POS tracks which dishes are being sold, MarginMind™ combines that with the profitability data to update the quadrant positions in real time based on actual sales volumes, not estimates. The system becomes more accurate the longer you use it.

Hangla Heshel is a real KhanaOS customer. Numbers and dish names used with permission. Revenue figures are approximate and based on session data from Q1 2026.