Retail business model

4.3.1 The physics: profit = margin × turns

A retailer buys at wholesale, sells at retail, and pays rent and wages out of the spread. Its ROCE is DuPont in its purest form, ROCE = EBIT margin × capital turns. Format sets both terms, so there are two sound ways to run the machine:

• Low margin, high turns (grocery/value). Gross margin is 12 to 18% and inventory turns 10 to 15× a year. You earn a sliver on each sale, thousands of times.
• High margin, low turns (apparel/lifestyle/jewellery). Gross margin is 35 to 55% and inventory turns 3 to 4×. You earn a wide spread on each sale, but sales come rarely. You carry fashion and markdown risk in between.

Neither one is "better." The test is whether margin × turns clears the cost of capital, and whether the format's flywheel is spinning. Scale gives you buying power. Buying power pays for lower prices or a better product, and that brings traffic, which brings more scale. The metric that tells you whether that flywheel is spinning is comps.

4.3.2 SSSG / comps: THE metric (worked)

Same-store sales growth (India), or comparable-store sales (US), is revenue growth from stores open at least 12 months. Some Indian reporters use 24, so check the definition. It matters because total revenue growth is two very different things glued together:

Total growth = SSSG contribution + new-store contribution

New stores are bought with capex. Comps are earned with the customer. Comps also carry the operating leverage. A mature store's rent, wages, and electricity are nearly fixed, so most of the gross margin on an extra rupee of comp sales drops straight to EBITDA.

How do you split same-store growth from new-store growth?

Worked example: Meridian Mart (synthetic). Meridian has 100 mature stores averaging ₹30 cr each, a base of ₹3,000 cr. This year SSSG is +7%, and 20 new stores opened evenly through the year, so on average each was open for half of it. They ramp at 60% of mature productivity.

• Comp base: 3,000 × 1.07 = ₹3,210 cr
• New stores: 20 × 30 × 0.5 × 0.6 = ₹180 cr
• Total revenue is ₹3,390 cr, so total growth is +13.0%, with 7.0 points earned from comps and 6.0 points bought with capex.

Now rerun it with SSSG = 0%. Total growth is still +6%, and the press release still says "record revenue." But every point of that growth consumed capital while the existing machine sat flat. That's the new-store treadmill. Keep opening stores with zero comps and ROCE grinds down toward the cost of capital, while headline growth looks fine. This one decomposition takes 60 seconds with the investor presentation open, and it's the fastest quality test in retail.

You can decompose comps one level further: (1 + SSSG) = (1 + transaction growth) × (1 + ticket growth). Ticket is the average bill, which e-commerce calls AOV (average order value). It splits again into items per basket × price per item, and transactions split into footfall × conversion. Meridian's +7% is transactions +3.0% × ticket +3.9% (1.030 × 1.039 = 1.070), which is balanced and healthy. In an inflationary year, comps driven only by ticket can hide negative real traffic. That's the red flag of "negative comps ex-inflation." A 6% comp against 7% CPI means fewer real baskets. The customer is drifting away behind a nominal curtain.

So what's a healthy level? India value formats in expansion mode print high-single to low-double-digit comps, and mature US retail prints 3 to 5%. The floor test is always comps ≥ inflation, meaning real traffic is at least flat. The quality test is whether transactions are contributing as well as ticket.

4.3.3 The productivity stack: sales/sq ft, footfall × conversion, private label

• Sales per square foot is the format's speedometer: revenue ÷ average trading area. DMart runs around ₹33,000 to 36,000/sq ft/year (illustrative, ~FY24/FY25 scale). Indian apparel value formats do ₹8,000 to 13,000. A Costco warehouse does roughly $280m+ a year, among the highest sales densities on earth (illustrative, 10-K). Compare only within a format, and watch the trend. Falling density while stores multiply means the new stores are worse or the old ones are leaking. Either way the incremental ROCE is below the average, and that's how great retail stories fade out.
• Footfall × conversion × basket is the store's demand equation. Grocery converts nearly all its footfall, because people come in to buy. Apparel converts 20 to 35%, because people come in to browse. So apparel operators manage conversion and grocery operators manage basket. Listen for which lever management talks about. It tells you what they can actually control.
• Private-label mix trades brand pull for margin. On its own label the retailer captures the FMCG company's margin layer, typically a 300 to 800bps pickup in gross margin, if its store brand has earned trust. Costco's Kirkland Signature is on the order of a quarter of sales, bigger by itself than most CPG companies (illustrative). Trent's Zudio is effectively 100% private label. That's how a value-fashion retailer can print ~40% gross margins at ₹500 price points. Private label is also a moat transfer. Every point of Kirkland share is a point of P&G/Unilever share that stopped paying the brand tax. Put that next to the D2C flag from the FMCG red flags and you'll see the brand machine getting squeezed from both ends.

4.3.4 Inventory economics: turns by format and GMROI (computed)

Inventory turns = COGS / average inventory. Use COGS and leave sales out of it, or high-margin formats look artificially slow. The format norms are grocery 10 to 15× (~25 to 35 days) and apparel 3 to 4× (~90 to 120 days). The most useful normaliser across formats is this one:

GMROI = annual gross margin ₹ / average inventory at cost = GM% / (1 − GM%) × inventory turns

Here it is computed for two formats (synthetic, ₹ cr):

So every rupee tied up in grocery inventory generates ₹2.1 of gross margin a year, and in apparel it generates ₹3.5. That's before you remember that apparel's number carries fashion risk, since a bad season means markdowns that vaporise the margin. Grocery's is metronomic. As rules of thumb, GMROI below ~2 strains most cost structures and 3+ is strong, but the trend matters more than the level. A retailer whose GMROI is sliding is either buying badly or discounting to move stock. If it's buying badly, markdowns are coming. Either way, go straight to the retail red flags below.

4.3.5 Working capital and the cash conversion cycle: DMart, worked

CCC = inventory days + receivable days − payable days, as you built it in the returns-on-capital work. Retail's structural gift is that customers pay instantly while suppliers wait. So a well-run retailer can hold CCC ≤ 0, and best-in-class retail runs on its suppliers' money. Amazon runs roughly −20 to −30 days. Costco is near zero, with inventory ~30 days and payables ~31, so the shelf is financed almost entirely by the people who stocked it (illustrative, 10-K scale).

Worked example: DMart's cash conversion cycle

Worked example: Avenue Supermarts (DMart), illustrative, modelled on ~FY24/FY25 scale (pull the annual report). Take revenue of ~₹51,000 cr and COGS of ~₹43,600 cr, a GM of ~14.5%. Inventory is ~₹3,900 cr, receivables ~₹200 cr and trade payables ~₹850 cr.

• Inventory days = 3,900 / 43,600 × 365 ≈ 32.7
• Receivable days = 200 / 51,000 × 365 ≈ 1.4
• Payable days = 850 / 43,600 × 365 ≈ 7.1
• CCC ≈ 32.7 + 1.4 − 7.1 ≈ +27 days, tying up ~₹3,250 cr.

Wait. Positive? The best retailer in India runs a positive cycle? Yes, and on purpose. DMart pays suppliers in about a week to harvest cash discounts. Stretch payables to a conventional 45 days and they become 43,600 × 45/365 ≈ ₹5,375 cr. The CCC swings to ≈ −11 days and ~₹4,500 cr of cash is freed. DMart turns that trade down because paying early buys roughly 1.5 to 2% off purchases. That's ~₹650 to 870 cr a year of extra gross margin, against perhaps ₹300 to 320 cr of value from investing the float at ~7%. The discount wins. It gets passed into shelf prices, and the price gap drives the traffic that drives the turns.

Costco's membership model, coming up below, teaches the same lesson: best-in-class retailers choose where to take their profit, whether float, fees, or price, and that choice is the strategy. Your job is to check the arithmetic of the choice. You don't need every great retailer to print a negative CCC. But when a retailer has neither a negative CCC nor a visible advantage that makes up for it, that's just capital sleeping on shelves.

Why can't you compare Indian and US retail EBITDA directly?

The India-US accounting landmine (from M1): leases. Ind AS 116 and IFRS 16 take rent out of EBITDA and turn it into depreciation + interest. US GAAP's ASC 842 keeps operating-lease rent inside opex. So an Indian leased-store retailer's EBITDA margin is structurally inflated by ~300 to 600bps against a US comp. The fix is to compare EBIT, or the pre-Ind AS 116 EBITDA that companies disclose. EBITDAR also works if you treat rent the same way on both sides. Never quote an India-vs-US retail EBITDA comparison without this adjustment.

4.3.6 Store economics: the unit P&L (bridge to M4.02)

You already know how to find the smallest repeatable unit. In retail it's one store, and the chain is just that unit times a rollout. Build the 4-wall P&L, which is store revenue minus store-level costs, before corporate overhead:

Zephyr Value Fashion, one store (synthetic). The store is 8,000 sq ft and sells ₹11,000/sq ft, so revenue is ₹8.8 cr. At a 38% gross margin that's ₹3.34 cr of gross profit.

How do you work out a store's payback and extrapolate it?

Opening the store takes fitout ₹1.6 cr + security deposit ₹0.3 cr + opening inventory net of payables ₹0.9 cr ≈ ₹2.8 cr of cash. Payback = 2.8 / 1.5 ≈ 1.9 years, which is excellent. Under ~3 years is the strong zone for value formats, and India value-fashion operators have claimed sub-2-year paybacks in this range. Store-level ROCE is EBIT (1.50 − 0.25 depreciation) / 2.8 ≈ 45%.

Make three professional adjustments before you extrapolate. (1) Maturation comes first. A store takes 2 to 4 years to reach that ₹8.8 cr, so a fast-expanding chain's consolidated margin gets dragged down by its youngest cohort. Judge the mature cohort and set the blend aside. Management usually discloses productivity by vintage, or for "stores older than 2 years." (2) Then corporate overhead. The chain's EBITDA = 4-wall EBITDA − head office, warehousing, tech. At 300 stores a ₹250 cr overhead eats ~950bps, and scale dilutes it, which is retail's operating leverage. (3) And cannibalisation. Store #40 in a city nibbles at the comps of stores #1 to 39. Densification strategies, like DMart's deliberate one, trade comps for logistics density and market control.

For valuation, the rollout maths is total opportunity ≈ (stores possible) × (mature store economics) − (overhead), discounted for the years it takes. It's the same reverse-multiple exercise you'll meet in the consumer multiples map and the 60-minute consumer teardown.

How do DMart and Costco make money?

Anchors. India: DMart is an everyday-low-price grocer with ~380 to 420 owned stores. Owning them means higher capex and no rent. It also means a slower rollout and a fortress balance sheet. Add GM ~14 to 15%, EBITDA ~8%, sales/sq ft ~₹33,000+, and the pay-early gross-margin harvest above (all illustrative, ~FY24/FY25 scale). US: Costco is the theoretically pure version of the same idea. Merchandise gross margin is ~11%, capped by policy, and essentially all operating profit comes from membership fees (~$4.8bn of fees vs ~$9.3bn operating income, roughly half, illustrative ~FY24 scale). Renewal rates run ~90%+, turns ~12× and CCC ≈ 0. Costco sells goods at cost and monetises trust as a subscription. Read Nick Sleep's Nomad letters on it, under the name "scale economies shared." The moat literally is the refusal to take margin. Both trade at premium P/Es (DMart historically 80 to 100×, Costco ~50×, as of mid-2026). Both are standing rebukes to the idea that premium multiples are always wrong. And both are standing tests of how carefully you reverse a multiple.

4.3.7 Retail red flags