A weight estimate is not a spreadsheet chore, it is the stability calculation in disguise. Every row in it has a mass and a height, and the second column is the one that decides whether the boat is any good. The total weight sets how deep she floats; the weighted mean height sets KG, and KG is the only term in GM = KB + BM − KG that anyone can still change once the hull is drawn.
The discipline is simple and almost nobody keeps it: no weight enters the budget without a height. A mass with no VCG is not an estimate, it is a liability sitting at whatever height the optimist who wrote it was imagining.
The arithmetic is a lever arm
KG is a weighted mean: KG = Σ(w·h) / Σw. That makes the sensitivity easy to state and easy to remember: moving a weight w up by h raises KG by w·h/Δ, and costs exactly that much GM. On a 5-tonne boat, a 100 kg mast-mounted sensor lifted a metre is 20 mm of GM — gone, permanently, for one line item.
Growth works the same way and is worse, because it is invisible. Structure comes in over estimate; cable trays get added; a second radio goes on the arch. Each is small, each is high, and they arrive one at a time after the stability book was signed.
Try it — a four-line weight budget
Why adding weight is not simply bad
Watch the total, not just KG. Adding mass sinks her, and sinking her raises KB and changes BM — a deeper hull has more of its volume below the waterline and, on most forms, a worse I/∇. So a weight added low can improve GM twice over: it pulls KG down and lifts KB. A weight added high does the opposite twice over. That is why a lever arm on its own is not an answer and the whole budget has to be re-floated.
It is also why ballast is a poor first answer and a fine last one. Ballast fixes GM by making the boat heavier, which costs speed, payload and freeboard — all of which you were presumably trying to buy with the weight that caused the problem. Moving something down is free; adding lead is not.
The margin belongs in the budget, not in your head
Carry an explicit weight margin and an explicit VCG margin, and write both into the budget as rows rather than leaving them as reassurance. They are different numbers: growth during build tends to be a percentage of the structure, while growth in service is almost all outfit and sensors, and outfit and sensors are high. A 5% weight margin applied at the existing KG is not conservative at all — it assumes the extra weight arrives with exactly the same distribution as the boat, and it never does.
The honest version is two rows: a mass margin at the mean height of the things you expect to grow, and a KG margin on top of it. Classification rules and naval specifications both require something of this shape; the reason is that the alternative — an unallocated margin — is always spent before delivery.
Late in the build you must add 150 kg. Same boat, two options.
It is 3% of displacement either way, and the two options differ by metres of lever arm. Low, it lowers KG and raises KB, improving GM on both counts; high, it does the reverse on both. The mass is the small part of the story and the height is the whole of it — which is exactly why a weight list without a VCG column is not a weight list.
Then go and weigh it
Every estimate is wrong and the error is almost always in one direction: heavier and higher. The correction is an inclining experiment — shift a known weight a known distance, measure the heel, and read GM straight off tanθ = w·d/(Δ·GM). With GM known and KB and BM computed from the hydrostatics, KG falls out as KB + BM − GM. That is the moment the budget stops being an estimate.
You can get most of the way there for free with a stopwatch. Roll period gives GM through T = 2πk/√(g·GM), and a boat that rolls slower than the calculation says is telling you KG is higher than the book claims — which is the direction the error always goes. Both routes are in the stability tool; the hydrostatics behind them are in displacement and hydrostatics.
One last discipline: keep the budget alive after delivery. The lightship condition drifts over a service life — paint, modifications, stores that never come off, water in places there should not be water. A boat that was fine on trials can be marginal five years later without anybody making a decision about it, and the weight list is the only place that shows up before the day it matters.