Home Improvement Retailer Case Study

GEO Test Proved Paid Search ROI 

A statistically clean test turned a hunch about nonbranded search into a defendable growth case. 

  • 100+

    Stores tested in a statistically valid geo experiment

At a Glance

Client
Large home improvement retailer
Engagement
Paid Search, Marketing Mix Modeling
Project
Geo Testing
Focus
Prove incremental ROI of nonbranded paid search before reallocating spend 

Overview

Platform metrics looked promising, but promising isn’t proof.  

A large home improvement retailer needed a defendable way to measure the return on nonbranded search investment before changing channel allocation.  

We treated data and creative as one system of proof, not separate workstreams, and built a geo test designed to hold up under scrutiny. 

The Proof

The test measured a consistent, statistically significant sales lift from nonbranded search. Results stayed aligned with industry norms and validated the channel across markets. The region saw over 2 percentage points of growth in total revenue. 

  • Over 2 Percentage Points

    Revenue growth across region

  • Statistically Significant

    Result validity

The Goal

Measure the true incremental impact of nonbranded paid search before the retailer committed to a bigger channel shift. 

The Challenge

Platform metrics and informal pre- and post-reads suggested the channel was working, but the client needed a comprehensive ROI measurement they could trust. The test design also had to account for shoppers who visit multiple stores and for real market differences across a large region. 

A person shopping in a home improvement store

The Solution

We designed a geo test across a region with more than 100 stores. We segmented stores using historical sales patterns and trade area signals like local demographics, weather, and macroeconomic factors, then matched each store with a statistical twin to create a clean control set. That architecture gave the client an apples-to-apples read on performance and made the result explainable. 

A person shopping for a weedwacker

Large Hospital Network Case Study

Modeling Drives Efficiency & More New Patients

Bayesian marketing mix model identified opportunities for smarter use of marketing dollars and incremental hospital growth within a set budget. 

  • 30%

    Improvement in cost per acquisition efficiency

At a Glance

Client
Large Hospital Network
Engagement
Marketing Mix Modeling
Project
Dual Patient Decision Path Map
Focus
Increase Marketing Efficiency
Duration
Two Years

Overview

The work turned spend decisions into a repeatable model for growth.

A large hospital network needed a clearer way to grow marketing-driven new patient acquisition inside a fixed budget.  

We built a connected marketing mix modeling system that surfaced the truth in the system: two different patient decision paths, each with different timing and response patterns.  

This gave a defendable view of channel ROI, diminishing returns, and a mix that could drive more new patient volume with less waste.  

The Proof

The hospital network shifted spend with more confidence and better visibility into marginal gain. Those gains came from a model the team could explain, defend, and use again as the budget moved. They improved cost per acquisition efficiency over two years all while reducing overall working media spend. 

  • 8%

    Reduction in working media spend

  • 2

    Patient decision paths modeled

The Goal

Increase the number of marketing-driven new patients using the existing budget, while also quantifying how a sister brand’s marketing influenced hospital new patient volume.

The Challenge

The hospital network needed to grow acquisitions without adding spend. They also needed a clearer read on which channels actually drove new patients, how returns changed as spend increased, and how outside factors like COVID-19 affected demand.

A nurse pushing a patient in a wheelchair. Both are wearing face masks.

The absence of that clarity made incremental budget decisions harder to defend. 

The Solution

We collected historical marketing and operations data, then added relevant contextual inputs, including regional COVID-19 case counts. Using a Bayesian marketing mix model, the team analyzed new patient volume patterns and identified two underlying patient types: one with a longer decision process and one that moved faster.

A doctor and a nurse in face masks examining a patient on a monitor.

The model quantified channel-level ROI, diminishing returns, and the optimal mix for maximizing new patient volume. With that architecture in place, the hospital network shifted channel mix and made incremental spend decisions based on predicted marginal gains across channels.