Business Context
A retail advertiser is reviewing a quarter of performance across Meta Ads Manager for campaigns running on Facebook Feed, Instagram Reels, and Stories. Spend increased, but finance and the account team disagree on whether the campaigns are actually healthy because ROAS, CPA, and projected customer value are moving in different directions.
Metric Scenario
In Q2, the advertiser spent $1.2M on Meta campaigns and generated $4.8M in attributed 7-day click / 1-day view purchase revenue from 24,000 purchases. Average first-order value was $200. Historical analysis shows new customers acquired through Meta have 12-month LTV of $320 on average, while returning customers driven by retargeting have incremental 12-month value of only $90. In Q3, spend rose to $1.5M, attributed revenue rose to $5.1M, and purchases rose to 30,000. However, the mix shifted from 40% new customers in Q2 to 25% in Q3. The VP of Marketing asks whether campaign health improved or worsened, and whether budget should continue shifting toward lower-CPA retargeting campaigns in Meta Ads Manager.
Requirements
- Define ROAS, CPA, and LTV precisely in this Meta advertising context, including attribution and customer-type caveats.
- Interpret the Q2 to Q3 movement in these metrics and explain why the metrics may point in different directions.
- Decompose campaign health into at least four drivers and identify which driver appears most concerning.
- Recommend how you would evaluate prospecting vs retargeting performance going forward.
- Name the guardrail metrics you would monitor before recommending a budget reallocation.
Data Available
- meta_ads_campaign_daily: campaign_id, ad_set_id, placement, objective, spend, impressions, clicks, attributed_purchases, attributed_revenue
- pixel_conversions: event_time, user_id, event_type, order_id, order_value, new_vs_returning_customer
- crm_orders: order_id, customer_id, first_order_date, repeat_orders_12m, gross_margin, refund_flag
- audience_membership_log: user_id, audience_type (prospecting/retargeting/lookalike), date
- geo_device_breakout: date, country, device_os, placement, spend, purchases, revenue
