Evaluate campaign success and decide new trading pair

Context

You are a Data Scientist at a crypto exchange. You work with Growth Marketing and Product to evaluate marketing spend and to make listing/launch decisions.

Part A — Measure whether a marketing campaign succeeded

A paid marketing campaign ran for 4 weeks across multiple channels (e.g., Search, Social, Affiliates). You have user-level event data (impressions/clicks optional), sign-ups, KYC completion, first deposit, first trade, and revenue.

Task: Propose how you would evaluate whether the campaign was successful.

Include:

• Primary success metric (pick one and justify): e.g., incremental net revenue, incremental funded accounts, incremental traders.
• Diagnostic metrics (examples): CTR, CVR by funnel stage (visit→signup→KYC→deposit→trade), CPA/CAC, activation rate, D1/D7 retention, ARPPU.
• Guardrails (examples): fraud rate, chargebacks, customer support tickets, churn, compliance/KYC failure rate.
• How you would compare performance across channels fairly.
• How you would estimate incrementality (vs. naive attribution), and key threats to validity (selection bias, seasonality, overlapping campaigns, delayed conversions).

State any assumptions you need (e.g., timezone, attribution window, whether user-level holdout is possible).

Part B — Decide whether to launch a new trading pair

Product proposes listing a new trading pair (e.g., ABC-USD).

Task: Describe the decision framework you would use to recommend “launch” vs “do not launch” (or “launch with constraints”).

Cover:

• Demand and expected usage (user interest signals, projected volume).
• Unit economics: expected fees, market-maker costs/fees, spreads, incentives, cannibalization of existing pairs.
• Liquidity and market quality: spread, depth, slippage, volatility, wash trading risk.
• Risk/compliance: regulatory constraints, manipulation risk, concentration risk, custody/technical risk.
• Launch evaluation plan: what metrics define success post-launch, how you’d monitor, and whether any experiment/rollout strategy is feasible given network effects.