Ridgeline Outdoors: last-click ROAS versus a geo holdout

Why the two numbers disagree. Last-click credits the final interaction, so it systematically over-rewards channels that harvest existing demand - brand and high-intent search - and under-rewards channels that create it. A 9.4 ROAS on paid search is not evidence of 9.4 units of incremental revenue; a large share of those clicks would have converted anyway. The holdout is the stronger signal because it measures a counterfactual, but it is not decisive on its own: eight weeks across 41 pairs gives limited statistical power, may pick up novelty or seasonality, and spillover between matched pairs can bias the estimate.

The recommendation. Do not move €1.5M. Split the decision into a small committed shift and an evidence-building track:

• Release €400-600k into upper-funnel social with a pre-registered test design: matched regions, fixed duration, total revenue and new-customer count as the outcome, not platform-reported ROAS.
• Commission a lightweight marketing mix model on 18-24 months of weekly data. The real question is not search versus social, it is the marginal return of the next €100k in each channel.
• Run the same test on search: suppress brand bidding in a matched set of regions for four to six weeks. Auditing your own best-performing channel is what buys credibility with finance.
• Hold the remaining €900k unallocated until both report.

Reframe the KPI. Move the board from channel ROAS to blended incremental CAC and marketing efficiency ratio. Channel-level ROAS rewards credit-claiming; incremental CAC rewards finding where the next euro works hardest.

Achievable in three weeks.

1. Re-analyse the raw holdout data rather than the summary slide: report the confidence interval on the 11.3% revenue effect, check covariate balance between pairs, and convert the revenue effect into contribution margin so social and search are compared on profit, not on revenue. Search and social rarely carry the same return rate, so a revenue-only comparison is misleading in search's favour.
2. Apply the same lens to the 2.1% search decline in suppressed regions: if paid social suppression moved search revenue down, that is direct evidence of cross-channel interaction and undercuts a pure last-click reallocation.
3. Build a one-page contribution model showing blended CAC at three budget scenarios (status quo, €1.5M shift, staged €500k shift) with the assumptions stated explicitly.
4. Draft the brand-search suppression test protocol for approval, so the next piece of evidence is already in motion.

Not achievable in three weeks. A validated marketing mix model, a multi-cell incrementality programme, or a defensible causal read on channels with short flighting. Claiming otherwise and being wrong is worse than a staged answer.

Structure the commitment. Ask the board to approve €500k of the reallocation now, with a written decision gate at week 10: if the social test shows incremental revenue per euro at or above the search benchmark, release the balance; if not, redirect it. Include an explicit kill criterion and a named owner. Boards accept uncertainty when the decision rule is fixed in advance; they reject it when the rule is negotiated after the result.

The current structure makes optimisation impossible: each director is incentivised to defend a channel number rather than to find the marginal euro. Fixes, in order of leverage:

Pool the budget above the channel line. Set one performance budget owned by a single leader with authority across search, social, affiliate and retail media. Directors become channel experts who execute, not budget owners who defend. This is the change that matters; the rest is secondary.

Separate always-on from experimental spend. Carve out roughly 10% (about €720k) as a ring-fenced test fund with its own approval path. Experiment budgets die when they compete against proven channels inside the same line item, because the proven channel always looks better on a last-click report.

Create one measurement owner. A single function owns attribution, incrementality testing and the mix model, and reports to the CMO rather than to any channel. Independent measurement is the only way channel teams accept an unfavourable verdict.

Change incentives. Replace channel ROAS targets with blended incremental CAC, new-customer volume, and contribution margin. Add a shared goal so no channel wins by shifting credit.

Install a quarterly investment council. Ninety minutes, fixed agenda: what we learned, what we will move, what we will stop. Decisions are recorded with an owner and a review date. Without a forum, pooled budgets quietly revert to the old splits.

Guard against over-rotation. Reallocate in tranches of 10-20% of a channel's budget, never in one move, and give tests long enough to clear the noise floor.