Our Meta CPMs tripled in three years. Where should a small Indian D2C brand actually put its money in 2026?

Ad costs in India have gone up faster than most small brands’ margins, and the organic traffic that used to cushion that is now getting intercepted by AI answers. This is what the numbers actually look like right now, what still works at a ₹15,000–30,000 monthly budget, and where an agency is and isn’t worth paying for.

Full disclosure: I work with GrowthLife, a growth agency in Chandigarh. I’ve kept the pricing below honest, including the parts that argue against hiring anyone.

Quick answer if you’re skimming

• Meta CPMs in India have gone from roughly ₹80–150 in 2023 to ₹200–350 in 2026. It’s not your account. It’s the auction.
• Your platform ROAS is inflated by 30–60% against what actually hits your bank. Reconcile to bank deposits monthly or you are optimising a fiction.
• Reels placement CPM runs about ₹40–80 versus ₹120–200 on Feed. Placement choice moves your cost more than bid tuning does.
• Creative fatigue is now 3–5 days at scale, down from 2–3 weeks. Volume of creative beats cleverness of targeting.
• AI search influences 12–18% of referral traffic globally and 25–35% of it is untracked. Don’t build a channel strategy on it yet — but do check your analytics can see it.

Why did my CPMs go up so much when I didn’t change anything?

• Indian Meta CPMs were roughly ₹80–150 in 2023 across most categories. In 2026 they run ₹200–350 (iMark Infotech, 25 May 2026).
• More advertisers are bidding for the same impressions. Every D2C brand that raised a round in 2021–22 is still spending, and a lot of offline retail moved budget online.
• Your CAC can rise while your campaign settings stay identical. Blaming the media buyer for auction inflation is the most common wasted argument I see.
• What this actually changes: at ₹300 CPM you need roughly double the click-through rate you needed in 2023 to hold the same CPC. That is a creative problem, not a bidding problem.

The 2023 to 2026 jump is auction inflation, not an account problem. The Reels bar is the lever you still control.

Is my reported ROAS lying to me?

• Yes, by about 30–60%. A platform showing 3.0x commonly reconciles to around 1.6x blended when you check the bank (iMark Infotech, May 2026).
• Both Meta and Google claim the same conversion. Add a third tool and all three will claim it.
• In India there’s a second leak most dashboards never see: COD returns. Typical RTO/return rates run 15–25% in fashion, 8–12% in beauty, 5–8% in food (Clicknify, 12 May 2026).
• A 3.2x reported ROAS on a fashion SKU with 25% returns is closer to 2.4x before you’ve paid for anything else.
• The fix is boring and free: one spreadsheet, monthly, total ad spend against actual bank deposits net of returns. That single number is worth more than any attribution tool you can buy.

A worked example, not a measurement of any one account. The gap between the top bar and the bottom one is what most brands never calculate.

What’s the smallest budget that can actually tell me anything?

• Around ₹15,000–30,000 per month, with a floor of about ₹500 a day per ad set (Clicknify, May 2026). Below that you’re collecting noise, not data.
• Under ₹15,000 a month, the honest advice is to not run paid at all yet. Put it into product photography, your landing page, and WhatsApp follow-up on the orders you already get.
• Don’t split ₹20,000 across six ad sets. One or two, enough daily volume to exit the learning phase.
• Give a test 7–14 days before you judge it. Killing on day three is the most expensive habit in small-budget accounts.

Where’s the cheapest inventory right now?

• Reels: roughly ₹40–80 CPM. Feed: roughly ₹120–200 (Clicknify, May 2026). That’s a 2–4x difference in what you pay to be seen.
• The catch is that Reels needs vertical video that survives the first second without sound. Your Feed static reformatted to 9:16 will underperform and you’ll wrongly conclude Reels doesn’t work for you.
• Hindi and regional-language creative gets 15–30% higher CTR than English-only in Tier-2 and Tier-3 (Clicknify, May 2026). Most D2C brands run English everywhere by default.
• Cheap impressions on bad creative is still money burnt, just slower.

How often do I really need new creative?

• At meaningful spend, 3–5 days. In 2023 it was 2–3 weeks (iMark Infotech, May 2026).
• That means roughly 8–12 new assets a month per active campaign to hold performance, not one big shoot a quarter.
• The practical response is a repeatable low-cost format — UGC-style phone video, founder pieces to camera, customer messages on screen — not a bigger production budget.
• If you can only do one thing this month, build a creative pipeline before you touch targeting.

Is AI search actually eating my organic traffic, or is that agency talk?

• It’s real but smaller than the panic suggests. AI search engines influence around 12–18% of total web referral traffic globally, up from 5–8% in late 2024 (upGrowth, updated Apr 2026).
• 25–35% of that AI-influenced traffic is misattributed or untracked in standard analytics, so some of what looks like a traffic drop is a measurement gap.
• The traffic that does arrive is better: ChatGPT referrals average about 3:10 on site with a 35% bounce rate, Perplexity about 3:30 and 32%, against roughly 2:15 and 48% for non-AI Google organic (upGrowth).
• Before concluding your SEO is dead, check whether your analytics is even labelling AI referrers. A lot of it lands in direct.

Should I be optimising for AI citations then?

• That ground moves under you. Community sources like Reddit and YouTube account for roughly 48% of AI citations, owned domains and editorial the other 52% (AirOps).
• But Reddit’s citation share fell 23% in a single month between October and November 2025, and Perplexity’s Reddit citation share dropped about 86% almost immediately after the Reddit–Perplexity lawsuit (CMSWire, 15 Apr 2026, citing Tinuiti and Conductor).
• A channel whose supply can drop 86% because of a court filing is not a channel you build a plan on.
• What’s durable is the same as it was: pages that answer a real buying question better than anyone else’s, on a domain you own. Those get cited by AI and rank in search, and nobody can revoke them.

What does hiring someone actually cost in India, and when is it worth it?

• Freelancer ₹5,000–30,000/month. Small agency ₹15,000–40,000. Mid-size ₹40,000–1,20,000. Large ₹1,20,000–2,50,000+ (BeTopSEO, 11 May 2026).
• SEO specifically: basic ₹10,000–25,000, standard ₹25,000–60,000, advanced ₹60,000–1,00,000+. Google Ads management ₹10,000–50,000 with ad spend separate.
• Rough rule: if your monthly ad spend is below about ₹1,00,000, a retainer that costs ₹40,000 is eating the budget that should be buying impressions. Hire a freelancer or learn it yourself.
• The honest version of my own pitch: below roughly ₹1 lakh a month in spend, most brands should not hire GrowthLife or anyone like us. Above it, the argument for an agency is creative volume and reconciliation discipline, not secret targeting.
• If you just want a second opinion on your site before spending anything, GrowthLife’s audit tool gives a free score, and the paid tier is ₹149 for a week or ₹399 for a month. That’s the level of commitment this question deserves.

What I’d actually do

• Spend one hour this week reconciling last month’s ad spend against bank deposits net of returns. Whatever that ratio is, that’s your real ROAS. Plan from it.
• Move budget to Reels placements and commit to 8–12 fresh vertical assets a month. That combination does more than any bid strategy change available to you.
• Leave AI search alone as a channel. Just confirm your analytics tags AI referrers, and keep writing the pages that answer real buying questions.