Last-click ROAS misallocates budget across channels with long lag
A home-services lead-gen company runs Search, Display, and YouTube with a $200k/month budget. Search has a 14-day average conversion lag; Display and YouTube have 45-60 day lags because homeowners research slowly. The team uses last-click conversion tracking with a 30-day window for all channels. At the monthly budget review, Display shows 1.2 ROAS and YouTube 0.8, while Search shows 6.0. The CFO wants to cut Display and YouTube and move 70% of budget to Search. Search is already impression-share limited at 90% on top terms, and brand terms are included. The marketing team has no incrementality tests and uses platform-reported conversions. Budget decisions are made monthly, and there is pressure to show immediate return. The board review is in two weeks.
1. Evaluate whether the reported ROAS by channel is a valid basis for a 70% reallocation.
Read the worked answer
It is not valid. Last-click with a 30-day window undervalues Display and YouTube because their average lag is 45-60 days; many conversions fall outside the window, and click paths often end in Search. Search gets credit for demand created earlier. Search at 6.0 ROAS may also be inflated by brand and by conversions that would have happened anyway. Display at 1.2 and YouTube at 0.8 may be understated. In addition, Search is impression-share limited at 90% on top terms, so a 70% budget shift cannot spend efficiently; marginal Search returns may fall sharply. The CFO's reallocation would optimize to a measurement artifact and could reduce total pipeline. The right conclusion is that the current table cannot support the decision, not that Display and YouTube are necessarily profitable.
2. Propose a measurement plan and budget decision process that accounts for lag and attribution.
Read the worked answer
Build a plan with four parts: 1) Extend the conversion window to 90 days for all channels and report conversions by conversion time, not click time. 2) Use data-driven attribution for cross-channel reporting while keeping last-click for channel diagnostics, and document the difference. 3) Run a geo-based holdout or incrementality test on Display and YouTube with a 4-6 week test window. 4) Calculate marginal ROAS by incrementally increasing Search budget in a controlled test, respecting impression-share limits. Use a two-speed process: monthly guardrails on CPA and lead volume, quarterly reallocation based on incremental ROAS. Do not move 70% in one step. Use a 10-15% tranche test if action is required before the board review, and present the confidence range.
3. If the CFO insists on reallocating now, what is the safest move and how would you frame it?
Read the worked answer
Shift a small tranche, about 10-15%, from the lowest-confidence channel to Search only up to the point where impression share and marginal CPA remain acceptable. Do not cut Display and YouTube by 70%. Keep YouTube and Display live long enough to observe lagged conversions, and set a holdout before the next review. Frame the move as a test, not a verdict: state that last-click undercounts long-lag channels and that Search is near saturation on top terms. Provide a decision rule for the next review based on incremental cost per lead and holdout results. This protects the CFO's need for action while avoiding a reallocation based on biased attribution.