Why your Meta ROAS fell in 2026, and what actually fixes it

Almost every Indian D2C founder we speak to this year opens with the same sentence: the Meta account used to work and now it does not. Same product, same team, roughly the same creative cadence, and a return on ad spend that has quietly halved. The instinct is to blame creative, because creative is the part you can see. Usually creative is not the first problem, and starting there costs you a quarter.

Three separate things moved between 2023 and 2026, and they interact. Here is what each one is, how to confirm it in your own account, and the order to fix them in.

1. The auction genuinely got more expensive

This is not folklore, it is in Meta's own quarterly reporting. In Q1 2026 Meta said average price per ad rose 12% year on year while ad impressions delivered grew 19%. In Q2 2026 it reported the same 12% price rise with impressions up 14%. Meta is showing more ads and charging more for each one, two quarters running.

Those are global figures, not India figures, and Meta does not break price per ad out by country. But a rising global clearing price is the backdrop every Indian advertiser is bidding into, and small-ticket consumer categories feel it first. If you sell a ₹499 jar, a 12% CPM rise eats a visible share of your contribution margin. If you sell ₹40,000 furniture, it barely registers. FMCG, F&B and nutrition brands sit at the sensitive end of that curve.

What this does not mean: that you should stop spending. What it does mean: a plan built on 2023 media prices needs rebuilding on 2026 media prices, and the rebuild has to come from margin, order value and conversion rate, not from wishing the auction back.

2. The platform can see less than it used to

Apple's App Tracking Transparency framework made cross-app tracking an explicit user permission. Browser-side pixels lose more events every year to tracking prevention, blockers and consent tooling. The practical result is that a browser-only Meta Pixel now reports a partial version of what actually happened on your store.

Meta's own answer to this is the Conversions API, a server-to-server connection that sends website, app and offline events from your systems to Meta's, where they are processed like pixel events for optimisation, reporting and measurement. If you have not implemented it, or you implemented it and never verified deduplication, you are asking a delivery system to learn from an incomplete record of who bought. It will still deliver. It will just deliver worse, and the number it reports back will be wrong in a direction you cannot see.

3. The dashboard number drifted away from the bank number

The third change is the one founders find hardest to accept, because it means some of the good years were partly an accounting effect. Speaking to Inc42 in July 2026, Sunitha Viswanathan of Kae Capital made the point bluntly: a brand can proudly show a 3X ROAS on a campaign and that number can still be hiding a real profit margin of just 5%, once you subtract cost of goods, discounts and returns. In the same piece, ROAS is described as having become a diagnostic for whether a specific ad works, rather than the number that decides whether to scale a channel.

The agency iMark Infotech, writing about Indian D2C accounts in 2026, puts the gap between platform-reported revenue and actual revenue at 30 to 60%. We would not present that range as a measured fact, because it is an agency estimate rather than an audited dataset. But the direction matches what we find every time we reconcile a new account against its Shopify orders and its bank statement, and the size of the gap is rarely small.

Fix them in this order: measurement, creative, budget

The order is not a preference. It is mechanical. Creative tests are judged on conversion data, so a broken measurement layer means you cannot tell a winning ad from a losing one. Budget and structure decisions are judged on creative results, so they inherit both errors. Start at the top and each layer is read correctly by the one below it. Start at the bottom and you spend three months learning things that are not true. iMark's write-up lands on the same three-layer sequence, and it is the order we run recoveries in.

Layer one: measurement, roughly two weeks

• Server-side Conversions API live for Purchase, InitiateCheckout and AddToCart, with an event_id on both the browser and server event so Meta deduplicates them instead of counting twice.
• Hashed customer data on every event so match rates hold up.
• One attribution setting across the whole account. Mixed windows make campaigns incomparable to each other.
• A weekly reconciliation: platform-reported revenue, store-reported revenue, money actually settled. Write all three in one row. The gap is your inflation factor, and you now know it instead of guessing it.
• Marketing efficiency ratio, which is total revenue divided by total marketing spend, tracked alongside campaign ROAS. Inc42 reports Indian brands moving to exactly this set of metrics along with blended CAC and contribution margin after marketing.

Layer two: creative, roughly four weeks

Inc42 also reports that audiences now tire of the same ad within 30 to 45 days, so brands pay a constant refresh tax simply to hold position. Volume is necessary. It is not sufficient, because ten variations of a weak idea is still a weak idea.

Barosi is the clearest example we have. The account sat at 0.6× ROAS, which means it returned less than it spent, so every extra rupee made the loss bigger rather than the business bigger. The ads were selling ghee. In ghee and honey, every label claims purity, and almost none of those claims are checkable at the moment of scroll. The correction was to stop selling ghee and start selling Barosi: a farm-to-table story with a face, a place and a reason to trust it. Three months later ROAS was 3.8×, CAC was halved, on-site conversion had moved from 0.8% to 6.8%, and monthly revenue had gone from ₹25k to ₹21 lakhs.

Pro Nature shows the same principle in FMCG. At 1.2× ROAS there was no room to fund testing, which is the trap: the account cannot afford to find out what would work, so it keeps doing what does not. The creative was rebuilt around everyday usability rather than the organic badge, because organic is a category table stake in Indian FMCG and does no persuading on its own. Ten months later the account was running at a steady 8×.

Layer three: budget and structure, from week seven

Aazol had a structure problem dressed up as a performance problem. Spend leaned hard on Meta Advantage+, which is efficient at finding buyers and opaque about how it found them, so ROAS swung and nobody could say which cohort, creative or SKU was carrying the result. Breaking that dependency with a 60:40 split, 60% of budget on manual targeting for readable data and 40% left on Advantage+ for scale, plus hero-SKU identification, a creative overhaul and switching Google on, brought CAC down nearly 70% and ROAS to 2.7× within four months, with revenue past ₹22.5L a month.

Kroslo is the fast version. Twenty-five days, no new budget, no rebrand: fresh creative angles in tight cycles, audiences cleaned up so ad sets stopped competing with each other in the same auction, and the funnel step immediately before payment tightened. ROAS doubled on the same spend. Fixes that remove waste pay immediately, which is why that result arrived in under a month.

The one recovery mistake to avoid

Cutting budget while ROAS is bad feels prudent and usually makes things worse. Lower spend means fewer conversions per ad set, which means slower learning, which means delivery gets less accurate, which means ROAS falls further. If the economics genuinely do not work, consolidate into fewer, better-funded ad sets rather than shrinking everything proportionally. Concentration buys you readable data. Thinly spreading a reduced budget buys you noise.

A ROAS number is an output of measurement, creative and structure. Change the output without changing an input and all you have done is re-roll the dice.

If you want the same sequence applied to your account rather than described, that is what our performance marketing work is, and the category-specific version of it lives on our FMCG page.