BoundForge — Scope of Work & Line-Item Overrun Command Center

Every scope-of-work template helps you write the document, and then it is finished. Nothing you own goes back and reads the words of each line against the hours it went on to cost.

A project ran 40% over. Which of its twelve lines did the running — and what did those lines have in common as they were written? BoundForge keeps every line of every scope exactly as it went to the client, the hours it was quoted at, the hours it took, and every change order raised against it, and answers that question from your own book.

The number nobody computes

THE COST OF AN OPEN LINE. Every line is scored by how many of five things it says in writing: how many, how many rounds, what finished means, what it does not include and what the client must supply. Each is a yes or a no about the words on the page, recorded when the scope is entered — never inferred later from how the line turned out. In the worked studio that ships with it, a line written with none of the five ran at ×1.60 of what was agreed; one written with all five at ×1.13. Open lines ran ×1.35 of bounded ones, with an interval of ×1.29 to ×1.41 off 109 and 143 delivered lines.

The identity that has to close

ACTUAL + UNDER-RUN = QUOTED + CHANGE-ORDERED + ABSORBED. It closes to exactly nought on every line, under every setting, and the residual is printed on screen whether it is nought or not. Absorbed is never typed: it is what is left when the quote and the change orders are taken off the hours worked, so there is nothing in it to round away. In the worked studio it is 3,809 hours — 30% of every hour on the book, never quoted and never billed. Coming in under is kept in its own column rather than netted off, because it is a different mistake.

The screen that argues with the rest of the application

ESTIMATING OR SCOPE? On a line with all five boundaries the scope could not grow without a change order — so whatever overrun is left there is you estimating low, and it lands on every line whatever the line says. Those lines still ran at ×1.13, which makes 31% of the overrun estimating. No amount of tighter wording will ever touch that part, and the screen says so — with the assumption behind the split printed, in both directions it can fail.

The missing boundary, without a regression

Which single boundary costs most to leave out? The honest answer is smaller than a model: take only the lines carrying exactly four of the five, and compare each group with lines carrying all five. One thing changes at a time. In the worked studio the boundary most clearly shown to cost something is revision rounds, ranked by the low end of its interval rather than the biggest estimate, and two of the five have not yet been shown to cost anything at all — the screen says which.

A boundary makes growth billable

A change order needs something on the page to point at. Fully bounded lines got 25% of their extra hours billed; lines with no boundaries got 0.9%. Priced at $125.04 an hour, the book realised $89.70 — both observed, neither a guess at what the hours were worth.

What it refuses to do

No ratio at all under thirty delivered lines; none for a score under eight, or a client, kind of work or word under six, with the floor printed beside every refusal. Lines still running are kept out of every ratio; dropped lines sit in the partition and out of every ratio; the quarter still running is withheld from the trend. It will never say a client was difficult, never price the absorbed hours as a loss, never grade the prose, and never write a reading back into a line.

What else is in it

Twenty-one screens. The SOW Linter: type a line, and it reads the words for the five boundaries, reads your book for what lines like it have cost, flags only the phrases that ran long in your book, and marks the gaps in square brackets. By kind of work, where the open-against-bounded gap is tested inside each discipline so it cannot simply be the work talking. By client, by quarter, priced-agreed-realised rates per scope, what is running now, and how much book you would need before a movement means anything. Nine calculators and six documents that fill themselves from your figures, including a reply to a change request that quotes the line back — and says plainly that keeping to the original scope is fine too. Double-click the file and it opens in your browser: no account, no installation and no network request of any kind.

A 19-sheet companion workbook. Every observation is a typed value and every conclusion a live formula. Tick one boundary or change one line’s hours and every median, the partition, the split and every cut move with it. No cell is an array formula, so it behaves the same in Excel, Numbers, LibreOffice and Google Sheets.

Two guides and a licence. Twenty minutes from opening the file to your first real answer, and a screen-by-screen reference that spends as much time on where the software stops as on what it does.

One note on your own data: you need the scopes you have already sent and the hours each line took. Tick the boundaries from the words, before you look at the hours. Thirty delivered lines is enough to read the first ratios; a year of scopes is enough for most of the screens.