Case study · Product development
Phoenix – Automated MMM Platform
X Lab knew Marketing Mix Modeling inside and out. Their team had the vision – they just needed a partner to bring it to life at scale. We spent 6 months building the first iteration of Phoenix together, and now it’s a platform they sell to enterprise clients.
the problem
Great methodology. Hard to scale.
X Lab had a proven methodology and a hard ceiling on growth. Scaling meant hiring – and training someone to work independently took 12 months. The answer was to turn the methodology into software. The challenge was finding a team who could understand the domain deeply enough to build it right.
- 250–400 hours per model, 3 months to deliver – revenue capped by analyst headcount.
- Model refreshes happened once every 2–3 years – clients making decisions on stale data.
- Onboarding a new analyst took 12 months before they could run a project independently.
The numbers first
What changed for X Lab’s business
The headline numbers are striking. The business model shift underneath them is the real story.
From services to platform revenue
Before Phoenix, X Lab sold time and expertise. Projects ended, invoices closed. With Phoenix, they sell access to a platform – recurring revenue that doesn’t require a new project kickoff every time.
From a handful of clients to unlimited capacity
The old model had a natural cap: how many projects could their senior team absorb? Phoenix removes that ceiling. The platform handles the routine modeling work. Analysts focus on complex, high-stakes engagements where their judgment still matters most.
The process behind 6 months from kickoff to enterprise-ready
We didn’t start writing code in week one. We started by learning to think like X Lab’s analysts.
Stage I
Discovery & Architecture
The first month was immersion. X Lab’s team walked us through the MMM methodology – not the textbook version, but how it actually worked in practice. The edge cases, the variable selection logic, the judgment calls that determined whether a model was trustworthy. From that, we built two things in parallel: a technical architecture and a design system foundation.
Stage II
Core MMM Engine & System Development
Backend and frontend moved simultaneously – algorithm engine, data pipelines, and API on one side; UI, workflows, and interactions on the other. Weekly reviews with X Lab, feedback directly into the next sprint. No big integration surprises. Continuous delivery, constant iteration.
Stage III
Iteration & Production Readiness
Real users on the platform meant real feedback. We upgraded the core MMM module based on what clients actually asked for. Built the budget optimization block – a feature that emerged from real usage, not the original spec. Completed the design system, hardened security, ran performance and accessibility reviews.
What it looks like
The platform in action
The Phoenix workspace – from variable selection to model output, one unified workflow.
Results
From services to platform. From months to weeks.
50×
faster model delivery
36×
more frequent refreshes
6 mo
kickoff to enterprise-ready
83%
training time reduced
What clients say
Real feedback from engineering leaders we’ve worked with.
“Working with this team means I don’t have to wonder whether something will get done – it will, and it will be done well. CodeWeaver helped us to build a platform that genuinely transformed how we deliver MMM to clients, but what I value most is how they work: proactively, with real ownership, and with the kind of quality that makes me trust them with the parts of the product that matter most. They don’t wait to be asked – they bring ideas, flag risks early, and consistently deliver beyond what was scoped. That reliability is rare, and it’s what makes this partnership work.”
Anna Skrzydło
Head of Product & Analytics, xLab
What Phoenix actually does
Intelligent Variable Selection Engine
Phoenix analyzes the client's data and automatically suggests optimal variable combinations – tests that used to take days now happen in seconds. Analysts validate and approve results rather than running them manually from scratch.
Guided MMM Workflow
The platform walks users through each modeling decision with context-aware guidance. The full workflow – data loading, model building, scenario planning, budget optimization, export – now runs in under a month. What used to take three months.
Budget Optimization Engine
Once the model is built, Phoenix automatically generates allocation scenarios – showing X Lab's clients exactly how to distribute their media budget for maximum incremental return. Output is exportable and ready for client presentations.
White-Label Ready Architecture
The platform is designed for X Lab to sell under their own brand to enterprise clients. Role-based access, multi-tenant architecture, scalable infrastructure – built from day one to support X Lab's commercial model, not just internal use.
Design System Foundation
Every interface was built on a coherent design system – which means extending Phoenix is fast, consistent, and doesn't require rebuilding patterns from scratch. X Lab's product roadmap has a solid technical foundation to grow from.
