How Bach AI Scores Audience Health Across Every Adset

‘Audience size: 4.2M’ is what Meta shows. ‘Audience health’ is what you actually need. A 4.2M audience can be saturated, overlapping with three other ad sets, and serving the same stale segment for weeks — while the dashboard still shows the reassuring big number.

Bach AI scores audience health as a 0-100 composite of four signals: saturation, frequency, overlap, and freshness. Every adset gets a daily health score with the underlying signals visible. The brands that scale efficiently are the ones reading the score, not just the size.

The Invisible Problem

Audience size is a static input. Audience health is a dynamic state. A 4M lookalike audience launched today is one creature. The same audience after 90 days of continuous spend at ₹8K/day is another — most of it has seen the ad, the freshness has decayed, and the auction pressure has built.

The downstream effect: ad sets in poor audience health show declining CTR, rising CPM, and rising CPA — but the founder cannot tell whether the cause is the creative (fatigue) or the audience (saturation). Without a clear signal, the response is often to swap the creative, which does not fix the underlying problem.

The Four Signals

Bach AI’s audience health score combines four signals:

1. Saturation — what fraction of the addressable audience has already seen ads from this ad set. Above 60% saturated is a yellow flag, above 80% is red.

2. Frequency — the average impressions per reached user. 3.0-4.0 is healthy, 4.0-5.0 is yellow, above 5.0 is red. Read [audience overlap, the silent ROAS killer](https://www.wittelsbach.ai/post/audience-overlap-the-silent-roas-killer-in-meta-ads) for the operational consequence.

3. Overlap — what percentage of users in this audience also appear in your other active audiences. Above 30% overlap is meaningful; above 50% creates auction self-competition.

4. Freshness — how many of the users currently reachable are net-new vs already engaged. New audiences score 90+ here; stale audiences score below 50.

The Composite Score

The four signals roll up into a 0-100 health score with categorical interpretation:

• 81-100 — Strong. Healthy and ready to scale.

• 61-80 — Stable. Maintain current spend; monitor.

• 41-60 — Watch. Early decay; consider audience refresh.

• 21-40 — Decay. Saturation building; refresh recommended within 7 days.

• 0-20 — Critical. Audience exhausted; swap or expand immediately.

What the Score Actually Tells You

The composite score is useful, but the four underlying signals point at different fixes:

• High saturation + low freshness = audience exhausted; expand the lookalike percentile or add an interest expansion.

• High frequency + low overlap = audience too narrow; broaden it.

• High overlap + healthy saturation = consolidate this ad set with another to remove auction self-competition.

• Low freshness + healthy saturation = pause the ad set briefly to let the audience refresh, then re-enable.

Audience Overlap — The Underrated Killer

Auction self-competition is the most expensive form of overlap. Two ad sets bidding for the same user in the same auction drives the CPM up for the brand — the brand pays Meta more to compete with itself. Bach AI surfaces overlap pairs across all your active ad sets and recommends consolidation where the overlap exceeds 40%.

The Refresh Recommendations

Detection without a fix is just an alert. Bach AI pairs audience health flags with specific actions:

1. Audience expansion — adding interests, broadening lookalike percentile, or removing narrowing exclusions.

2. Audience consolidation — merging high-overlap ad sets under a single broader audience.

3. Audience rotation — pausing a stale audience for 14-21 days while another carries spend, then re-enabling.

4. Lookalike refresh — re-seeding the lookalike on recent purchase data so the freshness signal climbs.

5. Geo or demographic expansion — adding a new layer to inject net-new reach.

The UI — What You See

Inside Wittelsbach AI, every ad set shows an audience health gauge with the four underlying signals as small radials. The Audience tab roll-up surfaces the unhealthy ad sets first, with the recommended action and the one-click approval. Overlap pairs across the account appear as a separate panel — the consolidation recommendations sit there.

The ₹ Impact

Across Indian D2C accounts on Wittelsbach AI in Q1 2026:

• Average % of ad sets with audience health below 60: 41% on accounts older than 12 months.

• Average ROAS uplift after audience-health interventions: 19-32%.

• Overlap-driven auction self-competition eliminated: typical recovery of ₹50K-1.5L/month on a ₹15L spend account.

• Time from health flag to action: 1-2 days with Bach AI vs 3+ weeks of manual analysis.

How Wittelsbach AI Operationalises Audience Health

Audience health is the underrated half of ad set performance. Bach AI surfaces it daily, recommends concrete refresh actions, and tracks the impact of every intervention. Connect your Meta account at [app.wittelsbach.ai](https://app.wittelsbach.ai) for a free audit.

Frequently Asked Questions

How does Bach AI compute audience overlap if Meta does not expose direct overlap data?

Bach AI uses Meta’s audience overlap report combined with deep signals from delivery patterns. The overlap report itself is available through the Audience Manager. Bach AI surfaces it daily, calculates the pairwise overlap matrix across all active ad sets, and flags any pair above 30% as worth investigating. Above 50% is recommended for consolidation.

What is the practical difference between high frequency and high saturation?

Frequency measures impressions per reached user. Saturation measures what fraction of the addressable audience has been reached at all. High frequency with low saturation means a small subset is being hit repeatedly — audience is too narrow. High saturation with healthy frequency means the audience is broad but mostly exhausted — audience needs expansion or rotation. Different problems, different fixes.

Can audience health alone explain a CPA spike, or is creative usually the cause?

Both are common causes. Bach AI correlates the two — if creative fatigue scores are healthy but audience health is in decay, the diagnosis is audience-led. If both are decaying, the recommendation typically addresses both: refresh the audience and queue a creative refresh in parallel. Single-cause diagnoses are rare in mature accounts; the model surfaces the correlated picture.

How often should I refresh a lookalike audience?

Lookalikes built on a high-quality seed (recent purchasers, high-LTV customers) typically stay healthy for 60-90 days. Lookalikes built on weaker seeds (Page Likes, video views) decay faster, often within 30-45 days. Bach AI tracks the freshness signal per lookalike and recommends a re-seed when freshness drops below 60. Most active D2C brands re-seed their core lookalikes monthly.

Does audience health scoring work for CBO campaigns?

Yes. In CBO, Bach AI scores each underlying ad set independently while also tracking the campaign-level audience overlap (since CBO auto-allocates budget across ad sets, overlap matters even more). The recommendation logic adapts — under CBO, consolidation often happens via shifting budget rather than merging ad sets, since Meta is already doing some of the auto-allocation work.