ML-Powered Marketing Mix Optimization for a Premium Lighting Brand

Reallocated media budgets across channels using Bayesian modeling and predictive simulations, driving incremental revenue with no added spend.

Technology
ML & Predictive Modeling
Geography
North America
Industry
E-Commerce

Key results

18.2%

increase in revenue with the same marketing budget by reallocating across high-performing channels

33

seasonal and promotional scenarios simulated, supporting real-time budget decisions during major events

41%

improvement in ROAS on Meta and Display through saturation-adjusted reallocation

50%

reduction in planning effort using self-serve tools integrated with marketing and sales data

About the Client

A U.S.-based luxury lighting and home furnishings brand renowned for its designer collaborations and curated collections. The client operates across e-commerce, showroom retail, and trade channels, with campaign-heavy seasonal peaks and high-margin product segments.

Key Challenges

No clear attribution of sales across marketing channels, especially during cross-channel campaigns

Missed reallocation opportunities during high-stakes retail events and designer launches

Overspending on saturated channels without visibility into diminishing returns

Manual campaign planning slowed decisions between marketing, merchandising, and finance

Our Approach

Data Consolidation & Feature Engineering

  • Unified weekly-level marketing, sales, promo, and calendar data across Meta, Google, Email, CRM
  • Engineered features for seasonality, product launches, and pricing events
  • Stored in Google BigQuery to support high-speed querying and simulations

Bayesian Marketing Mix Modeling

  • Built probabilistic models to estimate individual and combined impact of marketing channels
  • Modeled saturation curves to detect diminishing returns on ad spend
  • Captured external factors like seasonality, promotions, and event-driven demand

Sales Attribution & Scenario Simulation

  • Decomposed historical sales into base vs. incremental lift by channel
  • Enabled custom scenario testing (e.g. collection drops, flash sales, gifting season)
  • Simulated 20+ budget combinations with projected ROI and sales forecasts

Interactive Planning & Budget Optimization

  • Delivered an app for planners to compare current vs. optimized allocations
  • Included export-ready views for quick execution and team alignment
  • Empowered teams to test, tweak, and validate strategy in real-time

Implemented Solution

ROAS & Saturation Dashboard

  • Interactive charts showing diminishing returns and channel performance
  • ROI trendlines based on ad spend deltas
  • Filters by product category, campaign type, and channel

What-If Scenario Builder

  • Simulations for seasonal campaigns, showroom activations, and promo events
  • Inputs for dates, budgets, product focus, and goals
  • Visual outputs: expected lift, spend effectiveness, and sales decomposition

Optimized Budget Recommender

  • Detected overspend in saturated channels and underinvestment in high-return ones
  • Suggested shifts by campaign, objective, and margin priority
  • Enabled strategy pivots toward profit-maximizing mixes

Attribution & Sales Breakdown Engine

  • Weekly attribution of sales to each channel and campaign type
  • Side-by-side actual vs predicted performance
  • Tools to validate media mix strategy and model reliability

Technologies Used

No items found.
Reimagine campaign planning with machine learning - maximize ROI across every channel and season.
Connect with us to explore how we can drive similar results for your business.

Related Case studies

previous arrownext arrow