Optimizing Media Budget Allocation through Geo-Incrementality for Smartbox

The context

As the European leader in gift boxes, Smartbox makes massive media investments across three major bottom-of-funnel channels: Meta DPA, Google PMax, and Google Search, deployed across four strategic markets: France, Spain, Italy, and Belgium.
Despite these substantial investments, the brand lacked a harmonized, comparative view of the true profitability and incremental efficiency of each channel. Until then, evaluation relied on proprietary methods provided separately by ad platforms (e.g., Conversion Lift on one platform vs. Geo-Lift on another).

The key challenge for Smartbox was to answer 3 fundamental questions for its acquisition strategy:

What is the true incremental efficiency (iROAS) of each bottom-of-funnel media channel?
How can Google and Meta be objectively compared using the same neutral methodological framework?
How can media budgets be intelligently reallocated across channels and countries to maximize overall value creation?

The solution

To address this measurement challenge, Numberly and Smartbox designed and executed a unified, experimental Geo-Lift protocol deployed simultaneously across 4 countries for 3 channels (totaling 12 Geo-Lifts), built on postal-code level data across three key steps:

1. Python-based Data Science & postal code modeling: to form statistically representative and marketing-realistic geographic groups, a Python simulation algorithm developed by Numberly analyzed historical sales data by postal code. It generated perfectly balanced combinations of zones in terms of historical performance.

2. 5 Airtight Isolation Test Cells: building on these homogeneous geographic groups created by the algorithm, the territory was split into 5 airtight cells to isolate the net impact of each channel:

Everything On (~45%): All channels active to establish the Business-As-Usual baseline.
PMax Only (~15%): Strict isolation of Google PMax.
Search Only (~15%): Strict isolation of Google Search.
Meta BOFU Only (~15%): Strict isolation of Meta DPA.
Media Holdout (~10%): Neutral control zone with zero media exposure to measure organic baseline activity.

3. Deployment & Execution: Tailored campaigns were launched by region, enabling the team to isolate the net impact of each channel against the Media Holdout zone and calculate a strictly comparable iROAS (Incremental Return on Ad Spend).

The results

The framework delivered unprecedented clarity on true channel performance and enabled immediate budget optimization.

The true impact lies in decision-making: by replacing the over-attributed ROAS reported by ad platforms with a pure incremental indicator (iROAS), the project transformed measurement into a direct, actionable budgeting tool. Now that the impact measurement framework has generated key insights, Smartbox has reallocated its budget across strategies and countries, and will continue leveraging these findings in the coming months. Similar studies covering other levels of the funnel will also be implemented for Smartbox.

4,9 record iROAS in Belgium

achieved by Google Generic Search, demonstrating very high incrementality for paid search channels in these markets

iROAS = Incremental ROAS

4,1 record iROAS acheived

by Google Performance Max Non-Brand (in Italy, vs. 2.2 in France, 2.1 in Belgium, and 1.2 in Spain), highlighting major efficiency variations across countries

0,98 lowest iROAS recorded

attributed to Meta DPA (in Italy), indicating performance near or slightly below the break-even threshold, driving immediate budget reallocations