Meta's lookalike audience tool is one of the most powerful prospecting mechanisms available to Indian advertisers — and one of the most commonly misused. The default advice ("upload your customer list, build a 1% lookalike") ignores the single most important variable in Indian markets: your customer list is almost certainly a blend of buyer types with radically different value profiles, and when you feed that blend into Meta's algorithm, you get an average of all of them.
In India, where cash-on-delivery orders and UPI prepaid orders often coexist in the same product catalogue, the difference between building a lookalike from your full customer list versus your top-LTV customers can be a 40–60% difference in prospecting ROAS. This guide covers how to build lookalikes that actually work in Indian markets.
Why Indian Lookalike Audiences Underperform by Default
The problem begins with the source audience. Most Indian brands upload their full customer database — everyone who has ever placed an order — and ask Meta to find more people like them. But "people like them" is meaningless when your customer base includes:
- COD customers in Tier 2/3 cities with a 25–40% return rate
- Prepaid UPI customers in metros with a 4–8% return rate
- One-time buyers who purchased during a sale event
- Repeat customers who purchase every 45–60 days
- Gifting buyers who bought once in December and never returned
When you mix all of these into a single source audience, Meta builds a lookalike that finds people who could be any of them. The result is an audience that looks diverse on paper but converts poorly in practice, because the signal is diluted.
The LTV Segmentation Approach for Indian Source Audiences
Before you build any lookalike, segment your customer data by value. The segments you need for Indian markets are slightly different from US best practices:
Source Audience Size: What Actually Works in India
Meta recommends a source audience of at least 1,000 people for lookalike creation, with 10,000+ for best results. In Indian markets, the quality threshold matters more than the size threshold.
Minimum viable: 500–1,000 high-LTV customers (better than 10,000 average-LTV customers)
Optimal range: 2,000–5,000 high-LTV customers
Diminishing returns above: 15,000–20,000 customers (signal dilution begins)
Source: Meta Business Help Centre + internal AdsSarthi account analysis
The counterintuitive finding — consistent across Indian accounts — is that a source audience of 2,000 top-LTV customers routinely outperforms a source of 20,000 all-customers. The algorithm is finding more signal in the smaller, higher-quality list.
Lookalike Size Selection for Indian Markets
Meta builds lookalikes on a 1–10% scale, where 1% is the closest match to your source audience and 10% is the broadest. For India, this maps to approximately:
- 1% lookalike: ~10–12 million people (most similar, highest CPM, highest conversion rate)
- 2% lookalike: ~20–24 million people
- 3–5% lookalike: ~30–60 million people (good balance of scale and quality)
- 6–10% lookalike: ~60–120 million people (broad reach, lower conversion rate, useful for awareness)
The standard advice to always start with 1% is wrong for Indian markets at lower budget levels. At ₹500–2,000/day, a 1% lookalike of India's population is still 10 million people — your budget is too thin to get meaningful reach within it. A 2–3% lookalike with sufficient budget to generate 30–50 daily events will outperform a 1% lookalike running on a starvation budget.
The COD vs Prepaid Split Strategy
This is the most India-specific lookalike tactic and the one most commonly overlooked. If your brand accepts both COD and prepaid orders, build separate lookalike audiences from each payment method cohort and test them separately.
The reason is straightforward: COD buyers and prepaid buyers have different demographic and behavioural profiles on Meta. COD buyers tend to be more price-sensitive, more likely to respond to discount-led creatives, and more concentrated in Tier 2/3 markets. Prepaid buyers tend to be more brand-motivated, more responsive to quality and aspiration messaging, and more concentrated in metros.
Exclusion Strategy: What to Always Exclude from Lookalikes
Building the right source audience is step one. Excluding the right people from your lookalike delivery is equally important, and more commonly skipped. For Indian markets, always exclude:
- Existing customers (all-time): Upload your full customer list as an exclusion. Don't waste prospecting budget on people who already know you.
- Recent website visitors (30–60 days): These people are already in your retargeting funnel. Don't also reach them through your prospecting lookalike — it wastes budget and creates attribution confusion.
- Your page engagers (90 days): Anyone who has engaged with your Meta page or Instagram profile in the last 90 days is warm, not cold. Keep them in retargeting.
- Lookalike overlap audiences: If you're running multiple lookalikes simultaneously, use Meta's audience overlap tool to check for significant overlap (>20%) between them. Overlapping lookalikes compete against each other in auction, inflating your CPMs.
For a deeper look at managing audience overlap in Indian Meta campaigns, see our post on fixing Meta audience overlap for Indian advertisers. And to understand how iOS tracking gaps affect your source audience quality, read our guide on the Meta Conversions API setup for Indian stores.
Festival-Specific Lookalike Strategy
India's festival calendar creates predictable demand spikes that lookalike audiences can be specifically engineered to exploit. The tactic: 45–60 days before a major festival (Diwali, Durga Puja, Eid), build a lookalike from your previous year's festival buyers — specifically customers who purchased in the 2-week window around that festival.
These "festival-intent" lookalikes consistently outperform year-round customer lookalikes during the run-up to the corresponding festival, because Meta's model is finding people who share the behavioural and demographic patterns of buyers who purchased under festival conditions — with the season-specific motivations that entails.
Value-Based Lookalikes: The Next Level
If your pixel and CRM are passing purchase value events to Meta (which they should be — see our AdsSarthi attribution features), you can build value-based lookalikes rather than simple binary lookalikes. Instead of telling Meta "find people like my buyers," you tell it "find people like my highest-value buyers, weighted by order value."
Value-based lookalikes require a minimum of 100 purchase events with values in the last 60 days, and they require that your pixel or Conversions API is correctly passing purchase values. When set up correctly, they consistently outperform standard lookalikes by 15–30% on ROAS for Indian D2C brands in the ₹500–5,000 average order value range.
Testing and Iteration Framework
The right lookalike testing framework for India uses Meta's A/B test feature to compare source audiences directly, with identical creatives and budgets. Run each test for a minimum of 7 days and a minimum of 50 conversion events per cell before drawing conclusions. The most common mistake is ending tests too early — Indian Meta campaigns often need 3–5 days to exit the learning phase, leaving only 2–4 days of stable data in a week-long test.
Test one variable at a time: source audience quality vs. size, lookalike percentage, COD vs. prepaid source. Don't test multiple variables simultaneously — you won't know which change drove the result.