Methodology
How our figures are produced
Every chart on this site links here. The short version: we report the range of published estimates rather than picking one, we say which scope decision explains the gap between them, and we test each headline against something countable before we publish it.
1. Published estimates are collected, not averaged
For each market we gather the size and growth figures already in circulation, with each publisher's stated scope recorded next to its number. Estimates for the same market routinely differ by two to three times. That spread is almost never a disagreement about growth — published CAGRs tend to cluster tightly — it is a disagreement about what counts as the market. A report whose headline is a single number is reporting the analyst's scope decision, not a finding about the market, so we publish the spread and name the scope difference behind it.
2. Headlines are grounded against a public register
Before a figure is published it is divided through by something a reader can count: FDA device and approval listings, CMS provider counts, the IAEA DIRAC register of radiotherapy centres, ClinicalTrials.gov trial counts, WHO and OECD statistics, or revenue a company discloses in its own filings. The resulting per-unit number is checkable arithmetic, and it is often decisive — an implied $976,000 per treatment centre per year, for instance, rules out a narrow reading of a market's scope on its own.
A grounding calculation is a plausibility test, not a market estimate. Where we use one, the report says so, and says what the register itself leaves out.
3. What we leave out
Sections appear only when there is evidence behind them. In practice that means several things are routinely absent, on purpose:
- Vendor market shares. Private-company share is almost never disclosed. Our competitive sections name the players and describe their position in words. Where the only figures in circulation are a decade old, we say that instead of repeating them.
- Regional and country splits that no source provides. An omitted section is honest; an apportioned percentage is not.
- Segment percentages where only the segment list is known. The dimensions are listed without numbers attached.
- Recent-development lists padded with unsourced items.
4. Internal arithmetic checks
Each report is checked mechanically before publication. The stated CAGR has to reconnect the base-year value to the forecast value across the stated forecast period, to within a rounding tolerance; the plotted series has to begin at the base year on exactly the base-year figure; every chart has to carry a source; and the methodology section has to name at least three datable publications or registers. A draft that fails any of these is not published.
Forecast curves between two published endpoints are modelled at a constant rate. That is a presentational choice, not a prediction of the path, and the caption under each chart says so.
5. Corrections
Figures get superseded. If a number here conflicts with a source you hold, write to us and say which figure and which source — we would rather correct a report than defend it. Contact the research desk.