The piece list in. The truck out.
What Heatmark does, what it deliberately doesn't do, and where the hard engineering is. Everything below is a design, tested only in a prototype with demo data. Nothing here has run in a real shop yet.
The core loop
Works from the files you already get
- KISS and detailer CSV/XML reports: carry assemblies and piece marks.
- DSTV/NC1: one file per part. NC1 has no assembly hierarchy, so it supplements a report rather than replacing it. In the prototype, the assembly is inferred from the drawing number and flagged for review.
- PDF shop-drawing BOM: extracted to a draft piece list, with every inferred field highlighted for a person to check.
- Free report template for Tekla Structures and SDS2, so a detailer's export is one click. No Tekla Open API plugin at launch.
Rev 2 won't break the board
When a revision arrives, Heatmark compares it with the last one: added, removed, changed geometry, changed quantity. Changed pieces are flagged against what's already been scanned (a beam that was cut before its length changed is the expensive case), and a person accepts the merge.
Status is an append-only event log keyed to a stable piece identity, not the mark string. This is listed as a hard part in the plan, not a solved one.
Three or four scans per piece
Cut list complete; to paint or galv; back from paint or galv; on truck. Scan the QR tag or type the mark. Android-first in a native wrapper, with a recommended rugged phone of about $150 Assumption, or a shared kiosk tablet for shops that ban personal phones.
iOS web apps lack Background Sync and can evict stored data, which is why the plan goes Android-first (CTO red team; not re-verified).
Load list, weight check, BOL
Select finished pieces, see the running weight against the truck's limit, sequence by erection sequence, print the load list and BOL, and scan pieces onto the truck. On a bad-connection morning, the device builds the BOL from cached data.
Mill certs, human-confirmed
Store and search mill test reports per job and PO. AI drafts heat number, grade, size, chemistry and mechanical values; a person confirms every field against the PDF. Heat-to-piece links are by receiving or bundle in V1. Heat-at-cut linking waits for V2, and then only via nesting or saw-line reports that already carry the stock ID: a saw operator won't type a heat number per cut.
Progress-billing backup
Earned value by tonnage or line item against the schedule of values, with a backup table of pieces shipped. AI drafts the summary; the PM approves; nothing is sent automatically. CSV/IIF export first. A QuickBooks Online push only if pilots ask, because many family shops may run QuickBooks Desktop or Sage Assumption.
What Heatmark won't do
- Nesting, 3D modeling or detailing
- CNC control
- Inventory
- Full accounting, payroll, or timekeeping of record
- AI estimating from drawings. Cut in v2. SketchDeck.ai already sells AI steel takeoff [S17], so it's no edge.
Where this could fail
- Floor adoption. If the foreman quits and nobody scans, the board is fiction. Scan-health alerts go to the owner weekly ("job 2417: 41% of pieces have no ship scan").
- Messy files. Every detailer has quirks. The moat candidate is a test corpus of 50+ real US detailer bundles, collected with permission.
- Offline sync on cheap phones. 4 to 6 weeks budgeted for that alone.
- Trimble. It controls Tekla's API and partner terms [S26] while selling PowerFab. Core input stays export-only.
- Easy to copy. The QR, Kanban and BOL layer could be cloned by EZIIL in an estimated 3 to 9 months Estimate.
Demo Iron Works
A fictional shop with three demo jobs. Everything runs in your browser; state is saved locally and can be reset. Every screen carries a demo-data banner.
Import
Drop the sample NC1 bundle, CSV BOM or drawing-BOM PDF. See confidence flags on inferred fields.
Board and stations
Kanban across cut, fit, weld, paint and ship with per-job progress by count and tonnage.
Phone scan
Camera QR or typed mark. One big button advances the status, with timestamp and user.
Load builder
Pick completed pieces, watch the weight bar against the truck limit, print a BOL PDF.
Mill certs
Upload an MTR PDF, confirm the extracted heats, link them to pieces, look up any piece mark.
Job dashboard
Estimated vs actual hours, percent complete by tonnage, a drafted progress-billing summary.