Where does the funnel leak?

Cedar & Dune Home, a fictional online furniture shop, spent about SAR 677 thousand on five marketing channels between January and June 2024. The head of growth has to propose the budget split for the next half-year to the finance committee and asks you a plain question: where does each channel's funnel leak, and which channels deserve more or less money?

The data (campaign_funnel, synthetic) hold one row per channel per week: spend, impressions, clicks, signups and first purchases within seven days. Signups and purchases are attributed to the last channel clicked.

Stakeholders:

• Head of growth — owns the budget proposal and wants a ranked list of channels.
• Finance committee — will ask whether the differences are real and what a riyal moved from one channel buys in another.
• Channel managers — each will defend their channel; your analysis needs to survive that.

You will check the data (stage 1), locate the leak in each funnel with the right denominators and intervals (stage 2), price each channel (stage 3) and write a recommendation that is honest about what last-click observational data can and cannot show (stage 4).

All numbers come from a synthetic dataset generated for teaching; they describe no real company.