ZANE™. EST2016.
waiting for you Book a meeting →

Blended CAC Is Lying to You: The Case for Channel-Level Attribution

Most founders we meet can quote one number for their cost of acquisition. One figure, blended across every channel, trending in roughly the right direction. It feels like control. It is the opposite. A single blended CAC is an average, and an average is exactly where a failing channel goes to hide. As long as the blend looks healthy, nobody asks the harder question. Which rupee of spend actually brought the customer, and which one just rode along for free.

This is the most common way we see good Indian brands waste real money. Not on an obviously broken channel, but on a broken channel propped up by a great one, where the blend stays calm and the budget keeps flowing to the part that does not work.

What blended CAC actually averages away

Blended CAC is total acquisition spend divided by total new customers. That is its whole definition, and its whole problem. It does not ask where a customer came from. It treats the brand-search click that was always going to convert and the cold prospecting impression that built genuine new demand as the same unit of work, judged by the same average.

So picture two channels. One is brand search and retargeting, cheap, converting buyers who already wanted you, posting a tiny CAC. The other is cold prospecting, expensive, missing badly, posting a CAC that would horrify you alone. Blend them and the cheap channel’s efficiency drags the ugly one’s waste into a number that looks fine. You are not running two channels. You are running one channel funding another’s failure, and the blend is the trick that stops you noticing.

Blended CAC tells you what you paid on average. It never tells you what the next rupee will buy, and the next rupee is the only one you still control.

The honest number is marginal, and it is per channel

The number that matters is not the average. It is the marginal. What does one more rupee into this specific channel, today, at this spend level, actually return. Average CAC looks backward at money already spent. Marginal CAC looks at the decision in front of you, which is the only money you can still move.

This distinction is where most media-mix thinking quietly falls apart. Channels do not hold a flat efficiency as you scale them. The first rupees into a channel are usually cheap, harvesting the warmest, most ready demand. Push harder and you reach colder audiences, and the cost to convert each new buyer climbs. A channel can post a beautiful average CAC and a brutal marginal one at the same time, which means you are already past the point where adding budget makes sense, even though the average still flatters you.

The questions a blended number cannot answer

  • If I add one lakh to this channel, how many incremental customers do I get? Not total, incremental. The ones who would not have bought anyway.
  • Which channel is at its efficient ceiling and which still has cheap room to grow? Two channels at the same average CAC can be in completely different places on this curve.
  • How much of my best channel’s volume is demand other channels created? Brand search converts cheaply because something upstream made people search.
  • If I cut my worst channel tomorrow, what actually happens to total sales? Sometimes very little. Sometimes the whole funnel sags. The blend hides which.

Why the comfortable average survives so long

Be fair about why this persists. Blended CAC is easy. It needs no attribution model, no incrementality testing, no arguments about which touch deserves credit. It is one clean line for a board slide. And on a good month it tells a flattering story, so there is little pressure to look underneath.

Whoever owns the spend has every reason to keep the conversation on the blend. Channel-level truth implicates specific decisions and budgets. The blend implicates nobody. This is the same incentive trap we keep returning to. The metric that protects the people reporting it is rarely the metric that protects the business. Our Performance Marketing work starts by refusing the comfortable average, because you cannot fix a channel you have agreed not to look at.

The retention blind spot inside the blend

There is a second lie folded into the first. Blended CAC treats every acquired customer as one customer, full stop. But channels do not just differ on what they cost to acquire. They differ on what they bring. The cheap channel that posts a flattering CAC often buys discount-driven, one-and-done buyers who never return. The expensive channel you keep wanting to cut sometimes brings the patient, high-retention customers who actually build the business.

Judge those two on CAC alone and you will defund the channel that builds your future to protect the one that flatters your present. The fix is to stop reading acquisition in isolation. A channel is only as good as the cohort it brings, which is why we argue that retention cohorts are the one growth metric that survives a budget cut. Cut your CAC against the value a channel actually delivers over time, not against a headcount of first orders, and the rankings often flip.

This matters even more in India, where the cheapest acquisition is frequently the most disloyal. A channel that fills the funnel on deep discounts can post a wonderful blended contribution and a terrible repeat rate. The channels that earn durable relationships, including owned ones like WhatsApp used properly as a retention channel, change the real economics in ways a one-touch average never shows.

How to find the truth without rebuilding everything

You do not need a perfect attribution model. Perfect attribution is a multi-year project and a religious war. You need to break the average apart far enough to act, and a few disciplines get you most of the way.

  1. Report CAC by channel, every month, never just the blend. One line per channel, side by side. The first time you do this, the subsidy usually becomes obvious in a single glance.
  2. Watch the marginal, not the average. As you add budget to a channel, track what each new tranche of spend returns. When marginal CAC climbs sharply, that channel is near its ceiling regardless of how pretty its average looks.
  3. Test incrementality by switching things off. The cleanest read you can get cheaply is a holdout. Turn a channel down or off in a region for a fixed window and watch total sales, not just that channel’s attributed sales. If nothing moves, you were paying for customers you already had.
  4. Carry retention into the CAC view. Tag each channel’s cohorts and follow repeat behaviour, so an expensive channel that brings loyal buyers is not cut to protect a cheap one that brings churn.

None of this requires heroic tooling. It requires the willingness to look at the part of the picture the blend was hiding, and to let the spend follow that truth.

Where this bites hardest

The cost of the comfortable average is highest exactly when the stakes are. Early in a brand’s life, when every rupee is scarce, a blended number can mask the fact that your entire growth is one cheap channel harvesting demand while three expensive ones quietly fail. We have written about this in the context of the first ninety days of launching a D2C brand in India, because the habits you set in that window decide whether you spend the next year scaling a channel that works or defending an average that does not.

What changed recently

The case for channel-level truth got sharper in 2025, because the channels themselves stopped behaving like each other. Meta CPMs in India have risen sharply since 2023 while a new channel matured fast, and the gap between them is exactly the kind of thing a blend erases.

Quick commerce has become a genuine performance channel, and a different-shaped one. Reporting from Inc42 describes ads on Blinkit, Zepto and Instamart converting at roughly 3 to 8 percent against 1.5 to 3 percent on Meta and Google, with early campaigns posting about 1.5 to 2 times higher ROAS, measured as direct cart conversion at the point of intent rather than impressions. That same report is candid that the edge fades once the newness wears off and in-app keyword competition rises, which is the marginal-versus-average distinction playing out in real time. A channel can look spectacular on its first month’s average and ordinary on this month’s next rupee.

The money following this is not small. Per a Datum Intelligence forecast cited by Storyboard18, Blinkit, Zepto and Instamart together are projected to draw roughly Rs 4,900 crore in ad revenue in 2026. When a channel that books revenue at SKU and city level sits inside the same blend as a Meta line whose unit costs are climbing, the average is no longer averaging like-for-like. It is hiding which channel is buying incremental demand and which is harvesting demand the other one created. The brands that win the quick-commerce shift are the ones reading those channels apart, the same way you would read quick-commerce unit economics after platform fees rather than trusting a tidy contribution line.

The short version

Blended CAC is not wrong. It is incomplete in a way that happens to comfort the people reporting it. It tells you what you paid on average and stays silent on the only two things that change a decision. What the next rupee will buy, and which channel is paying for which. Break the average into channel-level, marginal truth, weigh it against the customers each channel actually keeps, and the loser stops hiding behind the winner.

If you can quote one CAC for your whole business and have never once seen it split by channel, that is the report to demand this week. Our Performance Marketing teams are measured on marginal efficiency and the retained value behind it, not on a blended figure that flatters a slide. The number your media mix is not showing you almost always explains more than the one it is.

Marketplace vs D2C: The Margin Tradeoff Indian Brands Get Wrong

Walk into almost any Indian brand’s strategy deck and you will find the same quiet contradiction. They are building a D2C website to protect their margin, and they are scaling on Amazon and Flipkart to chase volume. Both moves are defensible. The problem is that they are usually made separately, by different instincts, at different times, and nobody has reconciled them. The brand ends up half-committed to two channels and fully optimized in neither.

This is the single most common channel mistake we see. It is not that a brand picked wrong. It is that the brand never actually made a choice. It drifted into a channel mix driven by what felt safe, what a competitor did, or what an investor asked about last quarter. The result is a margin profile nobody designed and a discovery engine nobody tuned.

The two channels are not competitors, they are different jobs

Start by being honest about what each channel actually does. A marketplace is a discovery and trust machine. Customers are already there, already searching, already conditioned to buy. You rent that demand. The rent is steep: referral fees, fulfilment fees, advertising to stay visible, and the slow erosion of owning the customer relationship. What you get back is volume and velocity that a young brand cannot manufacture on its own.

D2C is the opposite trade. You own the margin, the data, the customer, and the brand experience. But you also own the entire cost of demand generation. Nobody lands on your site by accident. Every visitor is paid for, directly or indirectly, and your blended acquisition cost is the real price of that margin you were so proud of.

Marketplaces sell you discovery and tax your margin. D2C protects your margin and taxes you for discovery. Neither is cheaper. The bill just arrives in a different currency.

Once you see it this way, the ideological war dissolves. The question is never marketplace or D2C in the abstract. It is which job you need done at this stage of the brand, and at what cost.

Why the margin math fools founders

The trap is that D2C margin looks gorgeous on a per-order basis and terrible at the blended level. A founder sees a 65 percent contribution margin on their own checkout and assumes D2C is the high-margin channel. Then they layer in performance marketing, the agency retainer, the offer discounts, the return logistics, and the repeat rate that never quite materialized. The true margin after acquisition is often thinner than the marketplace channel they were trying to escape.

Marketplaces flatter you in the opposite direction. The headline margin looks brutal after platform fees. But the customer arrived without a single rupee of your own acquisition spend, which means the comparison founders almost never make is the honest one: marketplace margin after fees versus D2C margin after acquisition. Run that comparison properly and the verdict is rarely what the gut predicted. We push every brand we work with through this calculation before touching a channel plan, and the same discipline applies to newer formats. The economics of quick commerce after platform fees and returns humble even more founders, because the fee stack there is deeper than it looks.

The discount loop nobody budgets for

There is a second hidden cost that breaks both channels when it leaks across them. Marketplaces train customers to expect deals, and brands respond by discounting to defend rank during sale events. Then the same customer visits the D2C site, sees full price, and bounces. The brand quietly cannibalizes its own premium channel to feed its volume channel. Channel mix without price coherence is not a strategy. It is a slow margin leak with a dashboard.

The right mix is a function of lifecycle, not preference

Here is the part most channel debates miss entirely. The correct answer changes as the brand grows. A channel mix that is right at launch is wrong at scale, and a brand that does not re-balance deliberately will be carrying a stale allocation for years.

  • Pre-product-market-fit. Lean into marketplaces. You need volume, reviews, and live demand signal to learn what actually sells. Spending on D2C acquisition before you know your hero product is paying tuition twice. The first reads on price sensitivity and assortment come fastest where the buyers already are.
  • Early traction. Start the D2C engine in parallel, but treat it as a brand and data asset, not a volume play yet. Use it to capture the customers marketplaces hide from you, and to test bundles and pricing you cannot run on a marketplace listing. This is also when the first 90 days of a D2C launch in India either build a real foundation or quietly waste your runway.
  • Scaling. Now the mix becomes a deliberate margin lever. Push repeat customers toward D2C where you own the economics, while keeping marketplaces as the top-of-funnel discovery layer for new buyers. The marketplace becomes paid acquisition with a built-in storefront.
  • Mature. Defend margin aggressively. Quick commerce and marketplaces handle impulse and convenience, D2C handles loyalty, subscription, and your highest-margin SKUs. The portfolio is fully intentional, and every channel has a job it is uniquely good at.

The brands that win are not the ones who picked the right channel. They are the ones who kept re-picking as they grew.

Within marketplaces, the choice is not binary either

Even the marketplace half of the portfolio gets flattened into a false single decision. Brands default to whichever platform they personally shop on, then wonder why the category does not respond. Amazon and Flipkart behave differently by category, by buyer intent, and by fee structure, and the right call is rarely both at full intensity from day one. Choosing a marketplace by category rather than habit is a research question, not a coin flip, because the platform that fits your margin and your buyer is rarely the one you shop on yourself. Sound Channel Strategy treats each marketplace as a distinct economic environment, not interchangeable shelf space.

What changed recently

The fee map this whole tradeoff sits on has moved, and it has moved in opposite directions on the two sides. On the marketplace side, the headline cost is actually falling at the low end. Amazon India is expanding zero referral fees to products priced under Rs 1,000 across more than 1,800 categories from March 2026, a move it says can cut total selling fees by up to 70 percent for eligible items, per About Amazon India. This followed Flipkart taking its own commission to zero on sub-Rs 1,000 products, a competitive race to zero on low-ticket items reported by YourStory. If your hero SKU sits under Rs 1,000, the marketplace side of your math just got materially friendlier, and any channel plan written before this should be re-run.

Quick commerce is moving the other way. The platforms that once subsidised brands to win shelf are now extracting revenue, and the squeeze lands hardest on small D2C names. Mandatory listing-cum-ad wallets, minimum monthly ad spends, and bundled onboarding packages now run into several lakh per quarter, while founders report return on ad spend rarely clearing 1.2x to 1.5x, according to Storyboard18. The honest read is that the cheapest discovery for a low-ticket brand may now be a low-fee marketplace listing, while the most expensive may be the quick-commerce shelf everyone is chasing. That inverts the assumption a lot of 2024 decks were built on, and it is exactly the kind of input that should force a rebalance rather than a defence of last year’s allocation.

How to actually decide

Stop asking which channel is better. Ask three sharper questions instead, and answer them with your own numbers rather than the industry’s anecdotes.

  • What does discovery cost me in each channel, fully loaded, after fees on one side and acquisition on the other, using this year’s fee table and not last year’s?
  • Which channel teaches me something I cannot learn anywhere else right now, whether that is demand signal or customer data?
  • Where does my next rupee of margin actually come from at this stage, and is my current allocation aimed at it or at last year’s priority?

Answer those honestly and the mix designs itself. It will also keep changing, which is the point. A channel plan is not a one-time decision you defend. It is a portfolio you rebalance. Treat it like an operator, not an ideologue, and the margin-versus-discovery tradeoff stops being a trap and starts being a lever you control.

The brands that get this wrong are not reckless. They are usually careful, just careful about the wrong thing. They optimize each channel in isolation and never optimize the relationship between them. Get the relationship right, tune it to your lifecycle stage, and you stop choosing between margin and volume. You start choosing how much of each you want this quarter.

Book a meeting →