Beyond ROAS: The D2C Metrics That Actually Predict Profit in 2026

The ROAS trap

ROAS (return on ad spend) tells you revenue per rupee of ad spend on a single platform, but it ignores product cost, shipping, payment fees, discounts and returns. A 4× ROAS on a product with thin margins and high returns can still lose money on every order. Worse, platform-reported ROAS double-counts and over-attributes, flattering numbers that don’t reconcile with your bank balance.

CM2: the metric that tells the truth

Almost no D2C brand tracks CM2, yet it’s the single metric that tells you whether your business model is viable or just a cash-burning machine with good marketing.upGrowth, D2C Performance Marketing India 2026

Contribution Margin 2 (CM2) is revenue minus cost of goods, shipping, payment and fulfilment, and marketing cost. It’s the money left to cover overheads and profit. If CM2 is positive and growing, scaling makes sense. If it’s negative, every extra rupee of ad spend digs the hole deeper, no matter how good ROAS looks.

LTV: the real return on a customer

Acquisition only makes sense in the context of lifetime value. A customer who reorders three times is worth far more than a one-time buyer at the same CAC. Track LTV by cohort and channel, and judge acquisition against LTV, not against the first order alone.

Repeat purchase rate: the growth flywheel

Repeat rate is the quiet engine of profitable D2C. High repeat rates lower your blended CAC (because returning customers cost little to nothing to re-acquire) and lift LTV at the same time. If repeat rate is falling, no amount of top-of-funnel spend will fix the economics.

Blended CAC, not platform CAC

Platform-reported CAC flatters every channel because each one claims the same conversions. Blended CAC — total sales and marketing spend divided by all new customers, is the honest number, and it’s the one that reconciles with reality. Manage to blended CAC against LTV, and the platform dashboards become inputs, not the verdict.

Build the dashboard that ties it together

These metrics only change behaviour when they’re in one place, updated automatically, and tied back to revenue. That’s the job of proper attribution and analytics, exactly what our AI & Data Analytics service builds: one source of truth that forecasts the next decision instead of just reporting the last one.

Ratios worth checking your own numbers against

No published benchmark will tell you whether your brand is healthy. Category, price point, repeat frequency and channel mix move the absolute numbers so far apart that an average across all of Indian D2C means almost nothing. Ratios, not absolutes. What does travel between brands is the shape of the relationships, and if you check the ratios below against your own last quarter you will usually find the weak joint inside an hour, because a struggling brand is rarely failing on all of them at once.

Start with lifetime contribution divided by blended acquisition cost. Not revenue over cost. Contribution, meaning what is left after product, packing, shipping, payment fees and returns, measured across every order that customer has placed. A brand where that ratio sits comfortably above three has room to spend harder. A brand hovering near one is buying customers at roughly what they are worth, which is a business that grows in revenue and never in bank balance.

Then look at payback. Count the months it takes for a cohort’s accumulated contribution to cross what you paid to acquire it. Shorter payback is the single most useful thing a bootstrapped brand can own, because payback is really a question about cash rather than profit, and a brand with a four-month payback can recycle the same rupee three times a year while a brand at fourteen months is financing its own growth out of somebody’s savings.

Third, contribution as a share of gross revenue, tracked monthly. The absolute percentage will differ wildly between a skincare brand and a furniture brand. The direction will not lie. If that share slid four points over six months while your top line grew, you have bought growth by discounting and you have not noticed yet.

Fourth, the share of a month’s revenue that came from people who had already bought once. Watch that one closely. A brand where that share climbs quarter after quarter is compounding, and a brand where it stays flat while spend rises is renting its revenue from Meta at an escalating rent.

One month, worked end to end

Ratios stay abstract until somebody does the arithmetic. So here is an invented month for an invented brand. The numbers are illustrative and yours will differ, but the sequence of steps is the part to copy. Steal the method.

Say the brand did three thousand orders at an average order value of ₹1,200. Gross revenue, ₹36 lakh. Discounts and coupon codes took ₹3.6 lakh off that, so net revenue is ₹32.4 lakh. Cost of goods at, say, thirty percent of net is ₹9.72 lakh. Shipping across three thousand orders at ₹70 a shipment is ₹2.1 lakh. Payment gateway charges, packing material, and the cost of the returns that came back add another ₹2.5 lakh between them. That leaves roughly ₹18 lakh of contribution before any marketing at all.

Now the spend. The brand put ₹9 lakh into ads that month and the platform dashboards claimed 4x. Take the ₹9 lakh off the ₹18 lakh and you have ₹9 lakh left, out of which come salaries, rent, software, agency fees and the founder’s own time, none of which the ad platform has ever heard of. If those add up to ₹7 lakh, the brand made ₹2 lakh. On ₹36 lakh of revenue. That is the whole story, and no ad account report contains it.

Run one more step. Of those three thousand orders, suppose eleven hundred came from customers who had bought before and nineteen hundred were new. The ₹9 lakh of media bought nineteen hundred new customers, so the blended acquisition cost is about ₹474. Contribution on a first order sits near ₹600. So the first order roughly covers acquisition and nothing more, which means every rupee of profit in this business lives in the second order onwards, and the correct response is to move money towards retention rather than to keep asking the media buyer for a better return.

Do this every month. It takes twenty minutes once the sheet exists.

What cash on delivery does to every number here

Most Indian D2C measurement goes wrong at the same place. The brand counts an order when the order is placed. Cash on delivery does not permit that, because a COD order is a request rather than a sale, and a meaningful share of those requests never turn into money while every one of them costs you picking, packing, forward freight and return freight regardless. Count the money, not the intent.

The practical fix is to keep two order counts and never mix them. Orders placed. Orders delivered and paid for. Every ratio in the section above should be built on the second number. If your contribution margin is calculated on placed orders you are reporting a business that does not exist, and the gap between the two versions widens exactly when you scale, since aggressive prospecting tends to bring in the least committed buyers and those are the parcels that come back.

Return to origin also has a cost that sits outside the shipping line. A returned parcel ties up stock for a fortnight, arrives with damaged packaging often enough to matter, and occupies a slot in your warehouse that a paying order could have used. Count it properly. Add forward freight, reverse freight, the packing material written off and a realistic allowance for unsellable stock, then divide the total by delivered orders to get the real per-order drag.

Prepaid share is therefore a growth metric, not an operations metric, and it deserves a line on the same dashboard as revenue. Nudging buyers towards UPI at checkout with a small incentive, tightening address validation, confirming high-value COD orders over WhatsApp before dispatch and quietly declining COD on pin codes that have burned you repeatedly all move that share, and each point of movement drops straight into contribution because you have removed a cost rather than added a sale. That is margin found, not earned.

Track it weekly. Prepaid share is the cheapest margin available to most Indian brands, and it costs nothing but attention.

A marketplace rupee and an own-store rupee are not the same rupee

Plenty of brands report one revenue figure that adds Amazon, Flipkart, quick commerce and their own Shopify store together. That single number hides the only comparison that matters, which is what each channel leaves behind after everything it costs and what each one gives you the right to do next. Split the number.

The economics diverge sharply. Marketplaces charge commission, fulfilment fees, storage and often an advertising toll on top just to be visible on your own brand term, so the contribution on a marketplace order is structurally thinner even when the top line looks identical. Your own store carries the acquisition cost instead, which is usually larger upfront and behaves completely differently, because a customer you acquired is a customer you can email, message and sell to again at close to zero cost while a marketplace buyer belongs to the marketplace.

None of that argues for abandoning marketplaces. They are genuinely good at producing trial from people who would never have found you, and for a young brand that discovery is hard to buy any other way. The argument is for separate reporting. Keep a contribution figure per channel, a repeat rate per channel, and an honest view of what share of your total contribution depends on a platform that can change its fee card without asking you.

Then set a target for the split and manage towards it. Decide it deliberately. Most brands find the answer is not either extreme.

The measurement mistakes that cost the most

Four errors show up in nearly every brand that asks us to look at its numbers, and all four are fixable in a week. Look for these.

The first is judging channels on platform-reported conversions. Meta optimises delivery against the event you selected and reports on the attribution window you left at default, which means the same sale can appear in two dashboards and your total attributed revenue can exceed what actually landed in your bank. Use platform numbers to decide what to change inside the platform. Use bank-verified revenue to decide what to spend overall. Two numbers, two jobs.

The second is measuring lifetime value on a cohort that has not lived long enough to have one. A three-month-old brand does not know its repeat rate, and a projection built on two months of data will flatter itself badly. Report what you actually have. Say ‘contribution per customer at ninety days’ and let the number grow up.

The third is leaving discounts out of the margin calculation, which sounds too basic to happen and happens constantly, usually because the coupon sits in a different report from the product cost. Add a single column to your order export that carries the discount value per order, then recompute the last six months, and be prepared for a genuinely unpleasant afternoon if you have been running a permanent site-wide offer that everybody stopped questioning. Check the coupon column.

The fourth is a dashboard nobody argues with. If your weekly review has no moment where two people disagree about what a number means, the dashboard is decoration. Pick a fight monthly.

A short plan fixes all of it. Week one, export twelve months of orders with discount, shipping, payment mode and delivery status on every row. Week two, build the contribution calculation. Split it by channel and by payment mode. Week three, cut cohorts by acquisition month and plot contribution per customer at thirty, sixty and ninety days. Week four, pick the two ratios that look worst, decide one change for each, and put a date on when you will check whether it worked.