01Context

A draft survey turns water level into cargo weight.

A draft survey works out how much cargo a barge is carrying from how deep it sits in the water. Read the freeboard at marked points, convert those readings through the barge’s own hydrostatic table into displacement, subtract everything aboard that is not cargo, and the difference is the tonnage.

The client is a marine surveying firm working the British Columbia coast, surveying a 17-barge fleet (named by category, under a confidentiality agreement). It is the kind of field and marine operation where the measurement is taken outdoors, in weather, on a schedule set by the tide. The certificate its surveyors sign is a commercial document. The number on it is the number a buyer pays against and a seller books, which means a mis-read table row is not a spreadsheet problem. It is a commercial one.

02Before

Every signed tonnage traced back to a copied workbook.

The calculation ran in Excel — a workbook family two decades old, carried forward and copied barge by barge. Each barge has its own hydrostatic table. A surveyor found the bracketing rows for a freeboard reading, interpolated between them, applied the corrections in the same sheet, and the result was whatever that particular copy of the workbook happened to contain.

That arrangement holds until the questions get harder. A workbook cannot tell you which copy is authoritative. It cannot refuse a reading that falls outside the table it is interpolating against. It cannot prove, months later, which inputs and which table version produced a signed certificate. And the whole thing depends on the people who know where the formulas are buried staying available.

This is the pattern Aptixx sees most often in operational software work: the business logic is sound, and the spreadsheet carrying it stopped being the right container years ago.

03The decision

Rebuild the calculation spine first, and nothing else.

Two scoping decisions shaped the build. The first was to rebuild the calculation and the certificate it produces, and to leave the rest alone. Scheduling, dispatch, invoicing and CRM were deliberately not built — they are ordinary problems with ordinary software already available, and none of them was the reason the spreadsheet carried risk. Widening the scope there would have delayed the part that mattered.

The second was to ship a production system rather than a demo. That meant authentication, roles, versioned migrations, a real test suite and real signed PDF output from the start, instead of a convincing prototype that would need to be rewritten before a surveyor could put their name to its output.

04The system

Readings in, signed certificate out.

The platform takes freeboard readings, interpolates against the barge’s hydrostatic table, applies the corrections a surveyor would apply by hand, and issues the certificate — with the arithmetic held to a standard a spreadsheet cannot enforce.

Calculation flow: freeboard readings feed hydrostatic interpolation, corrections are applied from the barge table and dock water, and the system outputs a survey certificate and a daily production report. Readings Interpolate Correct Certificate Hydrostatic table Dock-water density 6-point / 4-point 20-dp decimal Survey certificate Daily production report
The calculation spine: freeboard readings, hydrostatic interpolation, dock-water and trim corrections, then the signed certificate and the daily production report.
  • 01Six-point and four-point freeboard reading modes, selected per survey stage rather than fixed for the whole job.
  • 02Freeboard-indexed hydrostatic interpolation, with hard out-of-range gates that refuse a reading the barge’s table cannot support.
  • 03Bilinear freeboard-by-trim interpolation, so a barge that is not sitting on an even keel is read across two axes of the table, not one.
  • 04Dock-water density correction taken from each barge’s own table density, not a site-wide default.
  • 05Itemized deductible line items with per-item units, so ballast, fuel, stores and constants come off the displacement individually and stay itemized on the certificate.
  • 06Imperial readings auto-detected per barge and converted to metric, so surveyors enter values in the units they were taken in.
  • 07Branded, signed PDF survey certificates and daily production reports generated by the system itself.
  • 08Every calculation in 20-digit decimal arithmetic — never floating point.

05Outcome and evidence

In production, and checked against a certificate that was already signed.

The platform is in production use by the client’s survey team. Its correctness rests on three things: a test suite, a fleet-wide hydrostatic audit, and a re-run of a certificate that was already signed.

1,630

automated test cases across 100 files

46

versioned database migrations

17

barges in one read-only fleet audit

20

decimal places on every calculation

Run against a previously issued, signed survey certificate, the engine reproduces it to the cent.

Alongside that, a fleet-wide hydrostatic audit ran across all 17 barges, read-only against production, checking each barge’s table and the engine’s handling of it.

Every figure on this page is countable in the repository or the database it came from.

06Trust controls

Four controls that make a signed number checkable.

A system that issues commercial documents earns trust the same way a surveyor does: by being checkable, years after the fact, by someone who was not there.

A test suite that owns the arithmetic

1,630 automated test cases across roughly 100 test files cover the calculation engine and the workflow around it. Interpolation, density, trim and rounding each have their own cases, so a change that moves a tonnage has to justify itself before it merges.

Schema changes as reviewable files

46 versioned database migrations. Every change to the shape of survey data is a numbered file in the repository with a forward path, not an untracked edit applied straight to a live database.

Verification that cannot touch production data

The fleet-wide hydrostatic audit ran read-only against production across all 17 barges. Checking that the tables and the engine agree is not allowed to mutate a single survey record.

Access that matches the operation

Role-aware access with public signup disabled; the first administrator is bootstrapped by an operator. Nobody self-registers into a system that issues commercial certificates.

07Next step

A workbook your business bills against is a single point of failure.

The marine specifics are particular to draft surveys. The shape of the problem is not. If a workbook holds logic your business bills against — unit rates, yields, load calculations, allocations, compliance thresholds — and one person is the reason it still works, that is the same engagement.

A paid discovery sprint is where that starts: map the calculation, agree what is in scope and what deliberately stays out, and price the build against a defined system rather than a guess.

Field and marine operations systems →

Fixed fee, $3,500–$9,500 USD, set in writing before work starts. You keep the calculation map, the build definition and the budget band whether or not Aptixx builds it; if you proceed, the fee is credited against the build.

Next step

Bring the calculation your business bills against.

20 minutes on the workbook, who depends on it, and what a defensible version would take. Free, and the next step is clear either way.