Technology roadmaps as quantitative models
Most roadmaps are arrows on a timeline. MIT's ATRA method turns them into quantified models built on figures of merit. This article describes its twelve sections, what a public library of student roadmaps shows, and a checklist for assessing a roadmap.
Most organisational technology roadmaps share the same structure: a timeline across the top, coloured swimlanes and an arrow pointing to the upper right. Such a slide cannot answer what would change if, for example, battery performance improved by 10% a year instead of 5%.
Olivier de Weck addressed this limitation with the Advanced Technology Roadmap Architecture (ATRA), the method behind MIT's course on Technology Roadmapping and Development and its online counterpart, the Management of Technology programme I followed. The central idea is that a roadmap should be a model that can be queried, not only a picture. This note examines what this requires in practice.
A household planning to buy a home faces the same distinction. "We will buy in 2030" is a picture. A spreadsheet that links the monthly savings, house prices and mortgage rates to a purchase date is a model: it shows how far a rise in mortgage rates moves the date, or how much an extra €200 saved each month brings it forward. A technology roadmap built on figures of merit is that spreadsheet for a technology.
Roadmapping started in industry long before it reached textbooks. Motorola and Philips used it in the 1980s, and Phaal, Farrukh and Probert (2004) gave it the layered form most people recognise: market and business drivers on top, products in the middle, technologies at the bottom, all against time. Kerr and Phaal (2020) traced those industrial roots. De Weck's contribution, published as Technology Roadmapping and Development (2022), adds quantification. Knoll, Golkar and de Weck (2018) showed the model-based approach on a solar-electric aircraft, the same example that now opens the MIT roadmap library.
Four questions and twelve sections
ATRA organises the work around four questions, in de Weck's words: Where are we today? Where could we go? Where should we go? Where we are going! Each roadmap answers them through a fixed set of sections:
| # | Section | What it requires |
|---|---|---|
| 1 | Roadmap overview | What the technology is and how it works |
| 2 | DSM allocation | Which other roadmaps it depends on or feeds |
| 3 | Model in OPM (ISO/PAS 19450) | Objects, processes and the figures of merit attached to them |
| 4 | Figures of merit | Name, unit, definition and historical trend |
| 5 | Alignment with strategic drivers | Targets on those figures of merit, with dates |
| 6 | Positioning against competitors | Where the organisation and its competitors sit on the same charts |
| 7 | Technical model | Morphological matrix, tradespace, sensitivities |
| 8 | Financial model | The technology's value, typically as NPV |
| 9 | R&D projects and prototypes | The portfolio that moves the figures of merit |
| 10 | Publications, presentations, patents | The external evidence |
| 11 | Technology strategy statement | One paragraph and one "arrow" chart |
| 12 | Roadmap maturity assessment | Optional: how good is the roadmap itself? |
Sections 4, 7 and 8 carry most of the analytical content. A figure of merit is a quantity physics or economics can reason about, as in my note on figures of merit and their theoretical limits. The technical model turns design choices into figures of merit; the financial model turns figures of merit into value. Once those three exist, the "what if the battery improved faster" question has an answer, and the arrow on the slide becomes a derived result rather than an aspiration, in the same way that the purchase date in the household spreadsheet follows from the savings and the rates.
Levels
Roadmaps are nested. The MIT library identifies each one by a code whose first digit is its level: 1 for a market or industry, 2 for a product or service, 3 and 4 for the technologies inside it. The solar-electric aircraft is 2SEA, a product; its batteries and solar cells would have level-3 or level-4 roadmaps of their own. The DSM in section 2 records those links, so that a target on the aircraft's endurance becomes a target on the battery's specific energy.
Observations from a roadmap library
Since 2019, students of MIT's class 16.887/EM.427 have written their roadmaps in a public wiki, roadmaps.mit.edu, all with the same structure. The wiki also files them in the 5 × 5 grid I described in a functional matrix for classifying technologies: what the technology acts on, and what it does. The topics are diverse (small launch vehicles, lab-grown meat, random forests, space electric propulsion, continuous security monitoring), which makes the library a fair sample of what engineers choose to roadmap when they are free to choose.
The figure below uses the library's own index. It illustrates two observations: most roadmaps sit at product level and in the "transform" and "transport" processes, so storing, exchanging and controlling are under-roadmapped; and a roadmap's completeness can be scored section by section.
- Distribution by cell. At the time of writing, the wiki's taxonomy table places 81 distinct roadmaps 82 times (one, the mRNA vaccine, sits under two processes). Transform and transport hold 73% of the placements; store and exchange five each.
- Distribution by level. About 62% of the placements are level-2 product roadmaps. Market-level (level 1) and deep technology roadmaps (level 4) are rare. Students, like companies, tend to roadmap concrete products.
- Scoring a roadmap. Press "Presentation only": a polished overview and strategy statement, competitor charts, and nothing underneath. It scores 36%. Press "Typical first draft" and then advance the sections one at a time. The radar fills only when the technical and financial models exist.
- Most efficient improvement. Starting from the first draft, which single section raises the score most per hour of work? Usually the figures of merit, because sections 5, 6 and 7 all depend on them.
Limitations of the score. It is a simple average of twelve ordinal states, so it ranks roadmaps rather than measuring them; a "reviewed" financial model built on an unquantified technical model remains unsupported. That is the same trap as averaging readiness levels, which I discuss in assessing system readiness from component technology readiness levels. Use the score as a checklist, not as a KPI.
A roadmap demonstrates its value when a quantitative question receives a quantitative answer. The twelve sections make this possible: figures of merit with units and trends, a model that links design choices to them, a value for moving them, and a portfolio of projects that moves them. Presentation matters as well, once the underlying content exists.
References
- R. Phaal, C. J. P. Farrukh and D. R. Probert, "Technology roadmapping, A planning framework for evolution and revolution", Technological Forecasting and Social Change, 71(1-2), 5-26, 2004.
- D. Knoll, A. Golkar and O. de Weck, "A concurrent design approach for model-based technology roadmapping", 12th Annual IEEE International Systems Conference (SysCon), 2018.
- C. Kerr and R. Phaal, "Technology roadmapping: Industrial roots, forgotten history and unknown origins", Technological Forecasting and Social Change, 155, 119967, 2020.
- O. L. de Weck, Technology Roadmapping and Development: A Quantitative Approach to the Management of Technology, Springer, 2022.
- ISO/PAS 19450:2015, Automation systems and integration - Object-Process Methodology.
- MIT Technology Roadmapping, Technology roadmaps (wiki of the class 16.887/EM.427), roadmaps.mit.edu.