Acquisition Metrics

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.

CPM Trend

What this shows: CPM climbed 2.7x from 130.79 KES (Jan-2025) to 349.48 KES (Apr-2026), peaking at 489.41 KES in Dec-2025. All months sit well above the 145 KES Kenya benchmark. Why it matters: Jan-2026 attribution window deprecation and the Mar-2026 algo rollout (shaded) accelerated auction pressure — the dip to 339 KES in Feb was a brief reprieve, not a trend reversal. Action: Do not interpret the Feb-2026 CPM dip as relief. The structural floor is 330+ KES. Efficiency gains must come from audience quality and bid strategy, not CPM reduction.

Spend Concentration — Apr-2026 (Age × Gender)

Where the budget actually landed in April. Segment order = spend descending.

What this shows: 35-44 male holds the largest named-segment share (11.4%) and delivers only 1.07x ROAS — every KES spent returns 7 cents net loss. The top-3 segments (35-44 male, 45-54 male, 25-34 male) absorb 25.4% of named-segment spend. 25-34 male is the only segment above 1.2x ROAS. Why it matters: The budget is concentrated in two bleeders (35-44 male + 45-54 male) that together account for 19.3% of named spend at sub-1.1x ROAS. This mirrors the 2025 April misallocation pattern from G1. Action: Shift budget weight from 35-44 male and 45-54 male toward 25-34 male and 55-64 male (1.21x ROAS). No placement should absorb >8.33% of budget (12-segment even split) unless ROAS exceeds account average.

Source: demographic_insights.parquet, 2026-04-24_0832_w4_demographic_insights_meta_metadata

Spend-Efficiency Quadrant — Apr-2026

Each bubble is a segment. x = share of April spend. y = ROAS. Size = absolute spend (KES). Upper-left = efficient and cheap. Lower-right = expensive bleeders. No segment clears the 2.54x break-even.

What this shows: 25-34 male sits upper-left (6% spend share, 1.36x ROAS — the only segment above account average). 35-44 male and 45-54 male are lower-right bleeders: high spend share, below-average ROAS. 18-24 female is extreme lower-left — tiny spend, catastrophic 0.14x ROAS. Why it matters: No segment clears break-even (2.54x). The account is structurally loss-making across all 12 demographic segments. The quadrant makes the bleeders vs underinvested dynamic visible: the best-ROAS segment (25-34 male) has 47% less budget than the worst large-spend segment (35-44 male). Action: Shift ~30K KES from 35-44 male (bottom-right) toward 25-34 male (upper-left). Run a 2-week ramp test before committing full reallocation, consistent with G1 conservative +16% methodology.

Source: demographic_insights.parquet, 2026-04-24_0832_w4_demographic_insights_meta_metadata

CAC per Purchase — Ranked

Cost per purchase by segment, sorted lowest to highest (cheapest customer acquisition at top).

What this shows: 25-34 male has the lowest CAC (1,224 KES/purchase) — 22% below account average. 18-24 male is an outlier at 6,363 KES (based on n=1 purchase — single-event noise, not a signal). 35-44 female and 45-54 female both exceed 1,850 KES CAC with sub-1.0x ROAS — doubly punished. Why it matters: CAC × ROAS together reveal the truly dangerous segments: high acquisition cost AND low return. 35-44 female at 2,239 KES CAC + 0.94x ROAS is the worst unit-economics combination in the account. Action: Treat 18-24 male as noise (n=1). Flag 35-44 female and 45-54 female for budget reduction — both exceed 1,850 KES CAC at sub-parity ROAS. 25-34 male remains the scale-up candidate (cheapest, best ROAS).

Source: demographic_insights.parquet, 2026-04-24_0832_w4_demographic_insights_meta_metadata

ROAS by Placement (Apr-2026)

What this shows: Only rewarded_video (2.18x) and facebook_stories (2.09x) are within striking distance of break-even. No placement clears the 2.54x contribution-margin hurdle. feed absorbs ~85% of spend at only 1.12x. threads_feed and both Instagram placements are deeply sub-scale. Why it matters: Placement-level bleeders (Instagram stories 0.51x, Instagram reels 0.60x) are destroying return on a portion of the budget that could be shifted to top-performing placements. But feed is the primary lever by absolute KES — even a 0.2x ROAS improvement on feed would outweigh pausing all Instagram placements. Action: Exclude instagram_stories and threads_feed from active campaigns (combined small spend, consistent low ROAS). Test increasing budget allocation to rewarded_video and facebook_stories in a dedicated adset.

Drill-down

Source: Meta placement_insights W3 parquet (2026-04-24_0829) + demographic_insights W4 parquet (2026-04-24_0832). Named demographic segments only (excludes unknown/unclassified rows which represent 48% of spend — a separate CAPI/pixel matching gap, not a targeting signal).

Source: placement_insights.parquet, 2026-04-24_0829_w3; demographic_insights.parquet, 2026-04-24_0832_w4