← MATE ROV Projects

Reference Class Forecasting

I built this tool because an early project estimate rarely includes the broken parts, redesigns, and extra work that show up later. It uses completed MATE ROV projects to give a new budget or schedule a more realistic starting range.

Whole ROV cost distribution

Histogram of observed projects with a fitted normal curve · n = 378

ObservedNormal fit

Average

$3,716

Arithmetic mean

Median

$2,782

Half fall below this value

Middle 50%

$1,500–$5,306

25th through 75th percentile

Sample

378 observations

17 archive years

Spread: $2,991 standard deviation; observed range $97–$14,663.

Data quality: High- and medium-confidence archive values. Travel is excluded; 2 extreme outliers were removed using a robust IQR fence. The fitted curve is a planning aid, not a claim that costs are normally distributed.

README

About the forecast

What this is

I use reference class forecasting to compare a planned ROV with completed projects from the MATE archive. The bars show the values I found in the source documents. The fitted curve is only a simple planning model over those observations.

Data boundaries

Whole-ROV costs prioritize company specification sheets; subsystem costs come from technical-document BOMs and use four broad engineering groups. Travel, lodging, registration, and fundraising are excluded. Dollar values are nominal and are not adjusted for inflation.

How to use it

Start with the measure, then narrow the class, year, or subsystem. I normally look at the median and middle 50% before the average because a few expensive projects can stretch the distribution. Samples below 12 observations are useful as a clue, not a dependable estimate.

Source

Values were extracted from public MATE ROV technical reports, specification sheets, budgets, and bills of materials published from 2004 through 2025.

View MATE ROV archives ↗