MIT feasibility study: example + template for SMEs

Key points: A MIT feasibility study is intended to establish, within a maximum of 12 months, whether an innovative idea is technically feasible. You receive 35% in subsidy (up to EUR 20,000) on the project costs. MIT feasibility is administered regionally and is not assessed on quality: applications are handled on a first-come, first-served basis, with a lottery on opening day if oversubscribed, so a complete application on day one is decisive. This article shows a successful example, including the structure, the budget and the crucial components that every application must contain.

What a MIT feasibility project is (and is not)

It is: a well-defined technical study with a clear question. Examples:

• Can an ML model predict customer churn with >70% accuracy using our data?
• Is a real-time fraud-detection pipeline technically feasible within our existing payment infrastructure?
• Which architecture is needed to centrally process sensor data from 50 production sites?

It is not: marketing validation, business-plan research, or building a product (for that you need MIT R&D collaboration or WBSO).

The difference lies in the outcome: a feasibility project produces a report, not a working product. The report answers "is this technically possible?" and forms the basis for a follow-up project.

Example case: AI demand forecasting for a wholesaler

Consider a fictional but representative example. The actual application was successfully awarded.

Company situation

A wholesaler in technical components with EUR 25M revenue, 38 FTE, and stock at a single Dutch distribution centre. The problem: stock-outs during seasonal peaks and overstock during dips, both of which cost money. Current approach: inventory parameters set manually per item based on experience.

The technical question

Can a machine-learning model, based on 4 years of sales history, calendar features and external signals (weather data, currency, sector indicators), predict weekly demand per SKU with better accuracy than the current naive moving-average baseline, measured via WAPE?

Why this is a good technical question:

1. An outcome in numbers (a WAPE figure)
2. A baseline to benchmark against (moving average)
3. A well-defined scope (one DC, weekly granularity, limited feature set)
4. Uncertainty that genuinely requires research (we do not know which model family wins or which features provide lift)

Project plan structure (RVO format)

A good project plan for MIT feasibility is 8-12 pages. Broken down into:

1. Company profile (1/2 page)

• Revenue, FTE, sector, prior WBSO history (relevant for starter status)

2. Innovation challenge (1-2 pages)

• The technical uncertainty: why is this not simply "using Excel better"?
• What is the current state of the art in your sector?
• What is the gap?

3. Research question (1/2 page)

• One sentence: the main question
• 2-4 sub-questions
• Clear measurement criteria

4. Method (3-4 pages)

• Which data do you use?
• Which methods will you test? (e.g. ARIMA, Prophet, gradient boosting, neural forecaster)
• Which evaluation strategy? (cross-validation scheme, baseline comparison)
• Which iterations do you expect?

5. Work plan + timeline (1-2 pages)

• Gantt chart with work packages
• Per work package: objective, deliverable, hour estimate, person responsible

6. Budget (1 page)

Budget breakdown for this example:

7. Expected result + follow-up (1/2 page)

• What does the report deliver?
• If the result is positive: follow-up project (MIT R&D collaboration, WBSO implementation)
• If the result is negative: what did we learn, what do we do instead?

8. Team + capacity (1/2 page)

• Who does the work, what is their experience/CV
• Own people vs external partners

What research results did this project deliver?

After 4 months of research, the findings were:

• Best model: LightGBM with lagged features + calendar + weather data. WAPE on the test set: 22% vs baseline 41%.
• The main lift comes from: seasonal features and sector indicators. Weather data had less impact than expected.
• Implementation architecture: a validated design for batch training (weekly) + real-time inference (per SKU).
• Estimated ROI for the follow-up project: inventory costs 12-15% lower, stock-outs reduced by 30%.

Final report: a 24-page document plus reproducible notebooks in a Git repo. The basis for a follow-up project via MIT R&D collaboration.

Ten tips for your own MIT application

1. Write the technical question first. A vague question produces a vague application.
2. Benchmark explicitly. "Better than X" is stronger than "improve". State the baseline.
3. Quantify everywhere. WAPE, accuracy, recall, lead time. Avoid "better performance".
4. Plan iteratively. Three small iterations are better than one large development phase.
5. Own-FTE hours at EUR 50/h. That is the maximum; higher rates are reduced.
6. External expertise is welcome. RVO views collaboration with external technical partners (such as Stratalytic) as positive.
7. No marketing in the project plan. Speak only about the technology; save the business impact for the closing paragraph.
8. Submit on time. MIT feasibility is allocated regionally on a first-come, first-served basis, with a lottery on opening day if oversubscribed; the budget is usually gone almost immediately. Each region opens on its own date, often in spring (Noord-Holland and Zuid-Holland opened around 7 April in 2026). So make sure your application is fully ready before the regional opening day.
9. Avoid follow-up costs in the application. Budget only for the research; implementation does not belong here.
10. Read the general terms. RVO publishes exactly what does and does not qualify. Reading the specific exclusions (such as routine activities) prevents rejections.

How Stratalytic approaches this

We write MIT feasibility applications for SME businesses and then carry out the feasibility study ourselves. This has three advantages:

• Speed. The application is ready in 2 weeks (versus 6-8 weeks for DIY).
• Reviewer perspective. We know which wording RVO lets through.
• End-to-end. Application approved? We are ready to start the research on day one.

See how we combine the roadmap + application or schedule a call directly.

Frequently asked questions

How long does a MIT feasibility project take? Officially a maximum of 12 months. Realistically for SMEs: 3-6 months.

Which costs can I claim? Internal labour costs (max EUR 50/hour), external expertise, material and testing costs, and software/licence costs for the research. Marketing and commercial costs are not eligible.

What is the difference between a good and a bad project plan? A good plan formulates one clear technical question with an outcome in numbers. A bad plan piles up vague ambitions without concrete research steps.