Meta — Spend Efficiency

CPM 349.5 KES (amber zone, up from 339.3). CTR strong at 4.62% (green). CVR 0.48% — down from 0.87%, driving ROAS to 1.10x (red, well below 2.54x break-even). CAC inflated to 1,573 KES (+56% vs Feb). Frequency reported as 0.00 in source parquet (field appears unpopulated at aggregate level). Source: Meta placement_insights W3 parquet (2026-04-24_0829 run), months 2026-04 and 2026-02.

Spend + ROAS — 16-Month Dual View

The single chart that makes compression visible: bars show spend scaling while the ROAS line collapses. The red dashed line is break-even. Every month the ROAS line is below it, the channel runs at negative margin.

Spend peaked at 605K KES in Jan-2026 while ROAS hit its first sub-2.0x month. The two shaded regions mark platform events (attribution window change Jan-2026; CPM algo rollout Mar-2026) that accelerated an existing downward trend. Action: do not scale spend until ROAS recovers above 2.54x.

ROAS Compression Scatter — Spend vs Efficiency

Each bubble is one month. The downward-right trajectory is the compression story: as monthly spend grew, ROAS fell. The annotation anchors the full magnitude.

The compression is not random volatility — it is a monotonic trend from top-left (small spend, high ROAS) to bottom-right (large spend, sub-break-even ROAS). Increasing budget accelerated the decline. The channel is in saturation, not a bad month.

ROAS by Placement — Apr-2026

Only 3 placements are at or above break-even. The top placement (facebook/feed) takes 66% of spend at 1.18x ROAS — below break-even.

17 of 20 placements are below break-even in Apr-2026. Audience Network rewarded_video is the only placement performing near break-even at 2.18x — and it receives only 0.34% of spend. The concentration risk: 66% of budget on facebook/feed at 1.18x ROAS.

Source: placement_insights.parquet, 2026-04-24, W3 worker

Apr-2026 Spend Concentration — Placement Share Donut

Where the money actually goes. Two placements (facebook/feed + instagram/feed) absorb 81% of spend. Both are below break-even.

66% of April spend went to facebook/feed at 1.18x ROAS — every KES spent there returns 1.18 KES. The only near-break-even placements (rewarded_video 2.18x, fb_stories 2.08x) together receive less than 0.34% + 1.83% = 2.17% of spend. The donut is inverted: the largest slice has the worst efficiency.

Placement Efficiency Quadrant — ROAS vs Spend Share

Upper-left = efficient with low spend (underweighted opportunities). Lower-right = expensive bleeders (overweighted losers). The correct response: grow upper-left, shrink lower-right.

Bleeders (lower-right, high spend share, low ROAS): facebook/feed (66%, 1.18x) and instagram/feed (15%, 0.89x) dominate the lower-right — they take the most money and return the least. Underweighted opportunities (upper-left): rewarded_video (0.34%, 2.18x) and fb_stories (1.83%, 2.08x) are the only placements approaching break-even; both receive under 2% of spend. The corrective action is a placement exclusion + reallocation, not a budget cut.

Source: placement_insights.parquet, 2026-04-24, W3 worker

Placement Drill-Down

Full Apr-2026 placement data sorted by spend. Use this to verify any exclusion decision.

What To Do at MER 2.36x (Feynman + Kahneman Lens)

The Feynman plain read: You are spending KES 1.00 to get back KES 1.10 on Meta. Your blended business needs 2.54x to break even (MER target). The gap is not a bad week — it has been open for 10 consecutive months and widening since Jan-2026. The cause is not one bad campaign. It is the combination of (1) CPM inflation (what you pay per 1,000 impressions has risen structurally), (2) budget concentration on the two largest placements which are both below break-even, and (3) audience saturation — you have been showing the same Nairobi audience the same ads for 16 months.

The Kahneman bias audit: Before you act on this page, challenge two System 1 instincts. First: "cut Meta spend entirely." Sunk-cost reversal feels satisfying, but if you cut without understanding which placements and audiences are efficient, you will also cut the 2.18x rewarded_video and 2.08x fb_stories — the only placements near break-even. Second: "scale what's working." rewarded_video at 2.18x sounds good versus 1.10x account average. But 2.18x is still below break-even (2.54x) — do not anchor on relative improvement when the absolute is still negative margin.

The honest three-step action at current MER 2.36x:

1. Immediate (this week): Add placement exclusions for the zero-ROAS placements (audience_network/an_classic, marketplace, threads/threads_feed, facebook_reels_overlay, biz_disco_feed, right_hand_column). Total spend reclaimed: ~9,835 KES/month. Small in absolute terms but removes confirmed waste with no signal value.

2. Structural (next adset cycle): Reduce daily budget cap on instagram/feed (-50%) and reallocate toward audience_network/rewarded_video and facebook/facebook_stories — the two placements at or approaching break-even. Do not shift the full instagram/feed budget in one move; run a 2-week test at -25% to measure elasticity before committing.

3. Diagnostic (before any scale decision): The compression is account-wide, not placement-specific. The root cause likely involves CPM inflation and creative fatigue. Until the CPM root-cause page (G3) is acted on — reducing CPM via new creative and audience refresh — reallocation within placements is rearranging deck chairs. The MER will not recover to 2.54x without either lowering CPM or raising AOV.

Kahneman pre-mortem: If you implement all three steps and ROAS is still below break-even in 60 days, the most likely explanation is not poor execution — it is that the Nairobi audience pool for this product is genuinely saturated at the current CPM floor. The correct next move in that scenario is a product-side intervention (AOV uplift, bundle pricing) not an ads-side one.

Source: placement_insights.parquet + monthly ROAS series, 2026-04-24, W3 worker