# Incrementality test brief

*One page, written before the test starts. Template from the Vexpower MMM library: vexpower.com/mmm/incrementality-testing*

## 1. The decision
- **Decision this test informs:** (e.g. "Keep, cut or scale Meta prospecting for Q1")
- **Decision owner:**
- **Date the decision has to be made:**
- **What we'd do today without the test:**

## 2. The question, as an estimand
- **Treatment:** (what changes: ads on vs off, +30% spend, new creative)
- **Comparison:** (no ads, business-as-usual spend, old creative)
- **Population / unit:** (users eligible for the campaign; DMAs; email list)
- **Outcome KPI:** (orders from our own data, revenue, margin, new customers)
- **Test window:** start / end, plus a cool-down of ___ days for carryover
- **One-sentence estimand:** "The effect of ___ compared with ___ on ___ among ___ over ___."

## 3. Design
- **Design:** (user-level holdout / geo test / switchback / on-off)
- **Randomization or matching method:**
- **Control share:** ___%
- **Expected conversions per week in test population:**
- **Smallest lift worth detecting:** ___% (and why that size matters for the decision)
- **Planned duration:** ___ weeks (from the power calculation, not from the calendar)
- **Spend in test cells:** $___

## 4. Threats and how we'll handle them
- **Contamination / spillover:** (other campaigns reaching control, people travelling between regions)
- **Seasonality or promotions in the window:**
- **Changes we will NOT make mid-test:** (budget, bidding, creative, audiences)
- **Novelty or wear-out:**
- **Measurement gaps:** (offline sales, untracked devices, consent loss)

## 5. How we'll read it
- **Primary metric and method:** (relative lift with 95% interval; iROAS with interval)
- **Breakeven iROAS:** ___ (from margin, not revenue)
- **We will not stop early unless:**

## 6. What we'll do with each result (fill in before launch)
| Result | Action |
|---|---|
| Lift detected, iROAS interval above breakeven | |
| Lift detected, iROAS interval straddles breakeven | |
| Lift detected, iROAS below breakeven | |
| No lift detected (interval includes zero) | |

## 7. After the test
- **Where the result is recorded:** (evidence register)
- **Will it feed the MMM?** As a prior / calibration input / validation only (pick one, so it isn't counted twice)
- **The next question this raises:**