On this page
This guide walks through the twelve weeks of the framework. Use it to plan your own project, or to understand what working with ML Data House looks like from one week to the next. The overview is on the framework page.
Before you start
A little preparation makes week one far more productive. Before kickoff:
- Name a business owner who can make decisions and attend the weekly demos.
- List the use cases you are considering and the outcome each should deliver.
- Identify where the relevant data lives and who can grant access.
- Run through the AI readiness checklist to find gaps early.
Month 1 · Understand & Design
Weeks 1–4. The goal of this month is clarity: what to build, how success is measured and how the solution will work.
Week 1Kickoff and access
Align on the outcome and make sure the team can start working with real data.
- Kickoff workshop with the business owner and stakeholders
- Agree the definition of success and the weekly demo cadence
- Set up access to data and systems with the right security controls
- Output
- A project charter and working access to the data and systems.
Week 2Data and systems audit
Find out what data you really have and how reliable it is.
- Catalogue sources, owners and refresh rates
- Profile quality: missing values, duplicates and drift
- Map existing systems and integration points
- Flag privacy and compliance constraints
- Output
- A data inventory and a data-quality report.
Week 3Use case and success metrics
Choose the first thing to build and decide how it will be measured.
- Score candidate use cases on impact, feasibility and data readiness
- Define KPIs, baselines and target thresholds
- Decide what to build, buy or reuse
- Output
- One prioritised use case with measurable success criteria.
Week 4Architecture and plan
Agree how the solution will work before building it.
- Design the solution architecture and data flows
- Sketch dashboards, screens or agent workflows
- List risks and how they will be handled
- Plan months two and three in detail
- Output
- An architecture document and a signed-off design review.
Month 2 · Build & Integrate
Weeks 5–8. The goal of this month is a working, integrated solution that has been tested against your success metrics.
Week 5Data foundations
Give the solution clean, reliable data to run on.
- Build ingestion, cleaning and transformation pipelines
- Engineer the features or structures the solution needs
- Add automated data-quality checks
- Output
- A tested data pipeline running on a schedule.
Week 6First end-to-end slice
Prove the approach with a thin version that works from input to output.
- Connect data, model or agent, and output in one flow
- Run an early demo with real data
- Collect feedback while changes are still cheap
- Output
- A demo of a working end-to-end slice.
Week 7Build out and integrate
Turn the slice into the full solution and connect it to your tools.
- Extend features and tune models, prompts or rules
- Build dashboards or user interfaces
- Integrate with your CRM, ERP or internal tools through APIs
- Output
- An integrated release candidate.
Week 8Evaluate and test
Check the solution against the success metrics agreed in month one.
- Evaluate against baselines and KPIs
- Analyse errors and edge cases
- Gather feedback from the people who will use it
- Output
- An evaluation report and a prioritised list of fixes.
Month 3 · Deploy & Hand Over
Weeks 9–12. The goal of this month is a safe launch and a team that can run the system without us.
Week 9Harden
Make the solution secure, fast and observable.
- Security review and access controls
- Performance and load testing
- Logging, alerting and monitoring
- Output
- A hardened build with monitoring in place.
Week 10Staging and user acceptance
Let real users try it before it goes live.
- Deploy to a staging environment
- Run user acceptance testing with real users
- Fix feedback and finalise runbooks
- Output
- User acceptance sign-off.
Week 11Production rollout
Go live carefully, with a way back if needed.
- Roll out to production, in stages if appropriate
- Watch the monitoring and compare results with the baseline
- Keep a rollback plan ready
- Output
- The system live in production.
Week 12Handover and next steps
Make sure your team can run what was built, and know what comes next.
- Deliver documentation and training sessions
- Review results against the success metrics
- Agree support options and the roadmap for the next phase
- Output
- A documented system, a trained team and a roadmap.
Common pitfalls
- Starting without data access. Delays in access are the most common reason timelines slip, so settle it in week one.
- Trying to solve everything at once. Pick one use case, deliver it, then expand.
- No business owner. Decisions stall when nobody can approve scope or priorities.
- Skipping baselines. Without a starting number you cannot show that the system improved anything.
- Treating go-live as the finish line. Monitoring and a support plan keep results from fading.
Next steps
Ready to turn this into a plan for your team? Check your readiness, or schedule a meeting and we will scope the first phase together. You can also read how the framework applies to our services.