Why Meta Says 'Audience Too Small' on a 5 Lakh Lookalike: The Real Eligibility Reasons

You built a 1% lookalike off your 90-day purchasers. Meta tells you the audience size is 5,00,000. You launch a campaign. Ads Manager throws: 'Audience too small to deliver ads.' But 5 lakh is well above Meta's 1,000-person floor. What's going on?

Audience size in the picker is not the audience size Meta actually uses for delivery. The picker shows the potential pool. The delivery engine applies a second layer of eligibility filters most Indian D2C brands never see.

First: Confirm It's Really an Eligibility Issue

Three signals look identical but require different fixes.

• 'Audience too small' at campaign creation — eligibility filters are reducing the pool below threshold.

• 'Low estimated audience' warning (yellow, not red) — campaign will run but delivery may be limited.

• 'Ad set may not deliver' — different issue, usually budget or bid related, not audience size.

The first one is what we're fixing. Confirm by checking the audience itself: Audiences > select the lookalike > size and status indicator should read 'Ready'.

The Eligibility Filters That Shrink Your Pool

Six layered filters run on every Meta audience the moment a campaign launches. Each can cut the eligible pool by 60-90%.

1. Geo filter — your campaign targeting India alone may be wider than the lookalike's actual geo. Indian lookalikes built from a small city seed get geo-locked.

2. Demographic intersection — age 18-65 + gender filter + interest layer + lookalike = an intersection that can shrink 5 lakh down to 40,000 instantly.

3. Placement filter — Reels-only + audience that mostly uses Feed = tiny eligible delivery pool.

4. Device & OS filter — auto-applied for some objectives. iOS-heavy seed → Android-heavy lookalike = mismatch.

5. Language filter — Hindi-tagged interest groups exclude English-only users in your seed.

6. Recent overlap filter — Meta excludes users already in other active audiences for the same advertiser.

The Diagnostic: Where Your 5 Lakh Becomes 5 Thousand

Run this audit before you change anything.

• Open the ad set. Note every targeting layer added on top of the lookalike.

• In Audience Insights, check the seed audience's actual age, gender, device, and city distribution.

• If the seed skews 25-35 women / iOS / Mumbai-Bangalore, a campaign targeting 18-65 / India / Reels-only will mismatch.

• Remove all interest layers from the ad set. The lookalike already encodes interest similarity — stacking interests on top is the most common mistake.

The Fix: Strip the Stack, Don't Grow the Seed

Most brands assume the answer is a bigger seed. Often it's the opposite — strip everything off the lookalike and let Meta deliver.

1. Remove all interest layers. Lookalikes already capture interest patterns from the seed.

2. Widen age to 18-65 unless your product has a hard age constraint (alcohol, kids' products).

3. Keep gender 'All' — Indian D2C brands often over-segment on gender, halving the pool.

4. Enable Advantage+ Placements. Letting Meta choose placements reopens 4-5x the pool.

5. Use the 1-2% lookalike size, not 0.5%. Smaller lookalikes look attractive but rarely scale past ₹3,000/day.

6. Avoid stacking exclusions — past-purchaser + past-engager + past-visitor exclusions can shave 30-50% of the pool.

When the Seed Itself Is the Problem

Lookalikes built from seeds under 500 events deliver poorly even at 1%. Five hundred purchases, 1,000 add-to-carts, or 2,000 lead events is the realistic minimum. If your seed is below that, Meta uses a fuzzy approximation that's prone to eligibility cuts.

Indian D2C brands with under 500 purchases should use value-based lookalikes from view content + add to cart rather than purchase-only. Larger seed, more stable delivery.

When 'Too Small' Actually Means 'Too Cold'

Sometimes Meta's message is a polite way of saying it can't find users with high purchase propensity inside the geo. A 1% lookalike from a Mumbai-only seed expanded to all of India may technically be 5 lakh but contain fewer than 1,000 users with strong conversion intent in the next 30 days. Meta won't deliver if the predicted conversion rate is below the auction floor.

Fix: lift the [purchase signal quality](https://www.wittelsbach.ai/post/conversion-api-capi-for-meta-ads-complete-india-d2c-setup-guide) via CAPI before rebuilding the lookalike.

How Wittelsbach AI Detects Audience Eligibility Issues Automatically

Bach AI runs the targeting stack of every campaign, calculates the real eligible pool after all filters, and flags when the effective audience drops below scale threshold — before you spend a rupee. Run a free Meta Ads audit at [app.wittelsbach.ai](https://app.wittelsbach.ai).

Pair this with our [audience overlap](https://www.wittelsbach.ai/post/audience-overlap-the-silent-roas-killer-in-meta-ads) deep-dive to make sure your exclusions aren't doing more harm than good.

Frequently Asked Questions

What's the minimum Meta lookalike audience size for stable delivery in India?

For India-wide campaigns, 10-20 lakh is the sweet spot for 1% lookalikes. Below 5 lakh, delivery becomes erratic and the algorithm exits the learning phase too late. Tier 1 metro-only campaigns can run on smaller pools (2-3 lakh) because user density and intent are higher. For state-specific delivery, aim for 1-2 lakh minimum, and accept that learning will take 7-10 days instead of 3-5.

Should I use a 1%, 2%, or 5% lookalike for Indian D2C scaling?

Start with 1% for ROAS protection, then expand to 2-3% once you're past ₹5,000-10,000/day. 5% lookalikes are too dilute for most Indian D2C categories — you're effectively running broad with a weak signal. The exception: high-AOV products (₹3,000+) where the seed is already qualified buyers; 3-5% works because the underlying signal is dense enough to survive dilution.

Can I combine multiple lookalikes into one ad set?

Yes, but Meta will treat the union as a broader audience and may default to the largest overlap zone. Combining a 1% purchaser-LAL with a 1% engager-LAL usually produces a slightly cheaper but lower-quality audience. The cleaner approach is to run them as separate ad sets inside a CBO campaign and let the algorithm allocate budget where ROAS is strongest. See our [CBO vs ABO guide](https://www.wittelsbach.ai/post/cbo-vs-abo-in-meta-ads-which-budget-strategy-wins-for-d2c-in-2026).

Why does Meta's audience size estimate change every time I refresh?

The estimate is sampled, not exact. It bounces 5-15% between refreshes because Meta queries a subset of its index each time. Bigger swings (30%+) usually indicate a true backend update — Meta adjusted demographic, behavioral, or interest mappings in your geo. The picker also recalibrates after iOS/Android attribution changes, so swings during a known Meta update window are normal.

Is a value-based lookalike better than a regular purchase-based lookalike?

For D2C brands with AOV variance (some customers spend ₹500, others ₹5,000), value-based lookalikes consistently outperform. They optimize for users who resemble high-LTV customers, not just any buyer. The catch: value-based requires the purchase event to pass a 'value' parameter via the pixel or CAPI. If your pixel only fires a flat 'Purchase' event, value-based lookalikes will silently fall back to standard behavior. Verify pass-through before relying on the audience.