Menu engineering matrix for Indian restaurants — Stars, Plowhorses, Puzzles, Dogs

About this piece. Menu engineering is the most underused unit-economics tool in Indian independent restaurants. The framework — Stars, Plowhorses, Puzzles, Dogs — is decades old and still works. The trick is running it on your menu with your numbers, on a regular cadence, and acting on what the matrix shows. This piece gives you the template, a worked example for an Indian casual-dine menu, and the action playbook for each quadrant.

The matrix in one paragraph

For each menu item, plot two numbers: contribution margin per item (₹ contribution per plate, not %) and popularity (units sold as a share of menu sales). Average each across the menu. The four quadrants:

Each quadrant has a different action. Most operator damage comes from treating all items the same way — running a 10% discount across a menu pulls Plowhorses harder (already high volume, already low margin) and barely moves Puzzles (the items that needed the push).

The two numbers — defined precisely

Get these definitions wrong and the matrix is misleading.

Contribution margin per item (₹) = Menu price (pre-GST) – Plate cost (food + bev cost for that item)

Popularity index (%) = Units sold of item / Total units sold across menu × 100

Average popularity (%) = 100 / Number of menu items × 0.7   (the 0.7 factor is the convention)

The 0.7 factor for the popularity threshold is convention — items above 70% of mean popularity are considered "high popularity". Be consistent.

Plate cost includes:

• All raw ingredients at vendor cost
• Garnish + accompaniments (the chutney, the salad, the achaar)
• Packaging (for delivery/takeaway items)
• Allowance for in-recipe waste (typically 5–8%)

Plate cost excludes:

• Labour to prepare
• Aggregator commission (treated as a sales reduction, not a plate cost)
• Overhead allocations

Contribution margin uses ₹ per plate, not %. A dish with 75% margin selling at ₹100 has lower contribution than a dish with 50% margin selling at ₹400.

Worked example — a 30-item NCR casual-dine menu

A composite NCR casual-dine outlet with a 30-item menu sees these top 10 items by units sold last month. (The remaining 20 items follow the same calculation.)

Mean contribution across the menu = ₹212. Mean popularity threshold = (100 / 30) × 0.7 = 2.33%.

Apply the matrix:

So this snapshot has 4 Stars, 4 Plowhorses, 1 Puzzle, 1 Dog among the top 10. The remaining 20 items would be slotted similarly.

The action playbook — by quadrant

Stars — protect them

These are the items that fund the business. The instinct is to "leave them alone" — wrong. Stars need active protection.

1. Lock the recipe. A Star with portion drift becomes a Plowhorse silently. Standard portion scales, weighed first plate per shift.
2. Lock the supplier. Star ingredient contracts get reviewed first when supplier pricing moves.
3. Position prominently on the menu (top right of each section, dedicated photo on aggregator listings).
4. Watch the price ceiling. Test a ₹15–₹25 price increase at low downside risk — the price elasticity on Stars is usually low.

Plowhorses — re-engineer the recipe or the price

Plowhorses are popular, low-margin. They drive footfall but don't fund the business. Three options, in order of preference:

1. Re-engineer the recipe. A 5–8% reduction in plate cost via portion re-spec, garnish trim, or supplier swap can move a Plowhorse to a Star without changing the customer's experience.
2. Bundle with a high-margin item. Garlic naan with a butter chicken combo at ₹525 (vs ₹500 a la carte) lifts contribution per cover.
3. Test a small price increase. A ₹10 price bump on a high-volume Plowhorse is the highest-yield single change in the matrix. Run it for 30 days, measure unit-sales drop, calculate net contribution change.

Puzzles — promote, don't kill

Puzzles are the operator's quiet opportunity. They're high-margin items the kitchen already knows how to make — they just don't sell. Three actions:

1. Reposition on the menu. Move from the bottom of a section to the top right of the section. Add a "chef's recommendation" tag.
2. Train the service team. Most Puzzle underperformance is service staff defaulting to recommending Stars. A 30-minute training and a Puzzles-of-the-week rotation typically lifts unit sales 30–50%.
3. Photograph for aggregator listings. A great photo on aggregator app listings can lift a Puzzle's units by 50–100% with no operational change.

Dogs — kill them, eventually

A Dog is low popularity, low margin. It contributes neither to revenue nor to footfall. The instinct is to remove it immediately — don't.

1. Confirm it's a real Dog over 60 days, not 30. Seasonal items and recently-launched items can look like Dogs in their first month.
2. Check menu balance. A vegetarian thali option with low velocity might still belong on the menu for non-veg-led tables that include a vegetarian guest.
3. If it's truly a Dog, remove it. Each Dog removed simplifies inventory, reduces wastage from low-velocity ingredients, and frees menu space for a new Puzzle test.

A typical menu-engineering pass removes 2–4 Dogs and tests 2–3 new items every quarter.

The cadence — quarterly is right

Monthly menu engineering is over-engineering — POS noise dominates the signal at one-month windows. Annual is too slow — supplier prices, customer tastes, and aggregator dynamics shift faster than that. Quarterly is the right cadence for most independent operators:

The template — what to build in a sheet

A 30-row, 8-column sheet will run a 30-item menu. Lock the formula columns; update only the inputs.

Column A: Item name
Column B: Menu price (pre-GST)
Column C: Plate cost
Column D: Contribution / plate     = B – C
Column E: Units sold (last month)
Column F: % of menu units          = E / SUM(E)
Column G: Quadrant flag            = IF + AND combo of D vs avg(D), F vs threshold
Column H: Action (free text from playbook)

Add a summary panel below the table:

Summary:
  Mean contribution/plate        = AVERAGE(D)
  Mean popularity threshold      = (100 / COUNT(A)) × 0.7 / 100
  # Stars / Plowhorses / Puzzles / Dogs
  Total contribution (₹/month)   = SUMPRODUCT(D, E)

The total contribution number is the bottom-line health check — it should rise quarter-over-quarter as the matrix actions kick in. If it's falling, the actions aren't working and the underlying assumptions need revisiting.

What goes wrong in practice

• Plate costs not refreshed. Vendor prices move; plate costs in the sheet don't. After 6 months a previously-Star item can be a Plowhorse and the matrix doesn't show it.
• Aggregator commission added to plate cost. It isn't a plate cost — it's a sales reduction. Mixing the two distorts both contribution and popularity.
• Beverages mixed with food. Run two matrices, one for food and one for bar/beverage. The dynamics are different.
• Reading the matrix without action. Every quadrant has actions. Doing the analysis without acting on it is the most common failure mode.

Menu engineering is a discipline, not a one-off project. The first run takes 4 hours. Every subsequent quarterly run takes 90 minutes. The compounding gain is 1.5–3 points of contribution margin over a year.

What to do this quarter

1. Pull last 90 days of POS data — units sold per menu item.
2. Cost each menu item from current vendor prices (don't reuse last quarter's plate costs).
3. Run the matrix in a fresh sheet.
4. Tag each item by quadrant.
5. Pick one action per quadrant (one Star to protect, one Plowhorse to re-engineer, one Puzzle to promote, one Dog to remove).
6. Implement and measure for 60 days.

Six implementations a year and the menu is structurally healthier than it was at the start. Sales rise modestly, contribution rises substantially.

Related on Restaurant Daily

• Restaurant unit economics in India — full guide
• Restaurant prime cost calculator India — formula + worksheet
• Food cost percentage in Indian restaurants — benchmarks by format