How to Extract Subcontractor Invoice Data to Excel
No Templates, No Rekeying

Why Most Extraction Tools Give Up on Subcontractor Invoices

If you've tried a standard PDF-to-Excel tool or a generic invoice OCR on a subcontractor payment application, you already know the result: it pulls the vendor name and total correctly, then falls apart on everything else. The retainage line reads as a negative charge. The schedule of values becomes an unlabeled grid of numbers. The job number — if it even appears on the page — gets lost in a header block the tool didn't recognize as relevant.

This isn't a failure of OCR. It's a failure of assumptions. Generic extraction tools are built for standard commercial invoices: vendor at the top, line items in the middle, totals at the bottom. They expect one document structure and one set of fields. A subcontractor invoice breaks every one of those assumptions:

• Multiple cost breakdowns on one page. A subcontractor billing for concrete work might list labor, materials, equipment, and retainage as separate line groups — each with its own subtotal — on a single invoice. Generic tools read this as "a table with too many columns" and produce garbled output.
• Construction-specific fields that standard OCR has never been trained to recognize. "Work Completed This Period," "Materials Stored," "Retainage Withheld," "Total Earned Less Retainage" — these aren't in the field dictionaries of generic invoice tools. The tool either skips them or mislabels them as "Other Charges."
• Job numbers and cost codes that live in the margins. A job reference might appear in a small box in the top-right corner, in the subject line of the subcontractor's cover email, or embedded in the project name — not in a labeled field. Standard extraction doesn't look for it because it doesn't know it should be looking.

The Fields That Matter in Construction (and Why Generic Extraction Misses Them)

A subcontractor invoice isn't just a bill — it's a job cost document, a compliance record, and a payment authorization rolled into one. The fields you need to pull aren't the same set you'd extract from a supplier invoice for office supplies. Here's what's on every subcontractor invoice and why each one matters for more than just cutting a check:

None of these fields appear on a standard commercial invoice template. That's why a tool trained on "Invoice Number, Date, Vendor, Total" returns blank cells for the columns that actually drive your job cost reporting.

How Column-Name Extraction Finds What You Need — Whatever the Format

The technique that makes this work is column-name extraction: instead of telling the tool where to look on the page (by drawing rectangles around each field, the way template-based OCR works), you tell it what you're looking for — and the AI finds the corresponding value by understanding what it means. For a deeper dive into how this approach works with invoice data specifically, see our guide to extracting specific fields from any invoice format.

The input is a list of column names you type — exactly the headers you want in your output spreadsheet:

The AI reads each document, locates the value that corresponds to each column name, and fills the row. "Sub Name" on one invoice might be in a "Contractor:" label in the header; on another it's the bold text centered at the top of the page with no label at all. The AI recognizes both because it's matching meaning, not position — the same way a human AP clerk scans a page and finds the subcontractor name in half a second regardless of where it sits.

This is fundamentally different from template-based tools, which work by recording pixel coordinates ("the vendor name is 2.3 inches from the top and 1.1 inches from the left"). A template breaks the moment a subcontractor changes their invoice software or adds a new field. Column-name extraction doesn't break — it's looking for the information, not the pixel.

Step-by-Step: Subcontractor Invoice to Job Cost Spreadsheet

Here's the actual workflow, from receiving a batch of subcontractor invoices to having a populated job cost spreadsheet. Each step exists because the previous one would create a bottleneck without it.

1. Define your columns once, reuse every draw period

Before processing any documents, decide what columns your spreadsheet needs. For a GC or specialty contractor tracking subcontractor costs by job, a practical column set looks like:

• Sub Name — who's billing you
• Invoice # — their reference number (essential for audit trail)
• Date — invoice or period-ending date
• Job # — which project this cost belongs to
• Cost Code — CSI division or your internal code (e.g., 03-3100 for cast-in-place concrete)
• Work Completed — dollar value of labor and installed materials this period
• Materials Stored — materials on site but not yet installed
• Total Billed — Work Completed + Materials Stored (the gross draw)
• Retainage — amount withheld (usually 5–10%)
• Net Due — what you actually pay this period
• Change Order # — reference any COs included in this draw

Save these columns as a preset so you're not re-typing them every month. The same column set works for every subcontractor — the AI handles the format variation on the input side.

2. Upload — one invoice or twenty, same flow

Drop your PDFs into the upload area. Subcontractor invoices arrive as email attachments (PDF), scanned paper from the job trailer, or phone photos of a handwritten bill. The extraction engine handles PDF, JPG, PNG, and WebP — no pre-processing, no 300 DPI requirement, no "please ask your concrete sub to reformat their invoices."

If you're processing a full month of draws, you can upload all invoices at once. The tool processes them in parallel — extracting a single page typically takes 5–10 seconds — and merges all results into one spreadsheet.

3. Review the extracted spreadsheet — not individual documents

Instead of opening each PDF and typing fields one at a time, you open one spreadsheet where every row is a subcontractor invoice and every column is a field you defined. You're reviewing data in the format you'll actually use it — not transcribing from one format to another. Spot-check high-value fields (Total Billed, Net Due) against the source documents for the first few invoices. The verification pass is discussed in detail below.

4. Export and feed your job cost system

Download as Excel (XLSX) or CSV. The spreadsheet is structured with your column headers, one row per invoice. From here, the data flows into your existing workflow — whether that's a QuickBooks import, a Sage 300 job cost module, a Procore cost ledger, or a shared Excel workbook that tracks project spend against budget. The point is that you're now moving structured data between systems, not rekeying numbers from paper.

Handling Retainage, Change Orders, and the Lien Waiver Paper Trail

Three aspects of subcontractor billing create data that standard extraction ignores — but in construction, they're the difference between a clean draw package and a disputed payment application.

Retainage. Most subcontracts withhold 5–10% from each progress payment. Your column set should include a Retainage field so the spreadsheet reflects both the gross amount billed and the net amount payable. If a sub calculates retainage differently than your contract specifies (e.g., withholding on labor only, not on materials stored), this column makes the discrepancy visible in the spreadsheet — before you approve the draw, not after the owner's accountant catches it.

Change orders. A subcontractor invoice that includes change order work should reference the CO number and the approved amount. Without those references in your spreadsheet, six months later you're staring at an invoice total you can't reconcile against the original subcontract value. Include a Change Order # column. If the invoice references multiple COs, the AI will capture all of them as a comma-separated list — you'll see at a glance whether this draw is base contract work or approved extras.

Lien waivers. Many GCs require a conditional lien waiver with each pay application. While the extraction tool doesn't generate lien waivers, having the Net Due amount extracted and verified means the waiver amount matches the actual payment — eliminating the most common source of lien waiver rejections. If you're managing 20+ subcontractors per project, cross-referencing waiver amounts against payment amounts by hand is a full-day task. An extracted spreadsheet turns it into a VLOOKUP.

What You Can't Skip: The Verification Pass

AI extraction replaces the first data entry — the act of reading a field on a page and typing it into a cell. It does not replace verification, and no extraction tool at any price point can. For printed data on clean documents, extraction accuracy reaches upwards of 99%. But "99% accurate" on 200 invoices means two invoices may have an error. Those two are the ones that create reconciliation headaches and payment disputes.

The practical approach is a tiered verification pass — not checking every cell on every invoice, but targeting the fields where errors have financial consequences:

• Tier 1: Spot-check all dollar fields on every invoice. Total Billed, Retainage, Net Due. These are the fields that determine payment amounts. Glance at each row against the source document. A missing digit in a $48,700 invoice becomes $4,870 — and you'll catch it in two seconds of visual comparison.
• Tier 2: Verify job numbers on the first invoice from each subcontractor. If the AI consistently extracts the correct job number for a given sub's format, subsequent invoices from that sub are likely correct. If it's wrong on the first one, adjust the extraction and reprocess.
• Tier 3: Skim cost codes for reasonableness. If a plumbing sub's invoice gets coded as Division 03 Concrete, you'll notice because it doesn't match the trade. Most miscodes are obvious in context.

This tiered approach takes 2–3 minutes for a batch of 20 invoices — compared to the 15–20 minutes per invoice that manual entry requires. The bottleneck shifts from "getting data into the spreadsheet" to "verifying the data that's already there." That's a vastly smaller time commitment.

Frequently Asked Questions

Does this work with handwritten subcontractor invoices?

Yes, with qualifications. The AI reads handwritten text — printed handwriting more reliably than cursive, and clean handwriting more reliably than rushed scribbles. A plumber's handwritten invoice with clearly printed numbers and dollar amounts will extract well. A faded carbon-copy with smeared pencil marks and overlapping text will have lower accuracy. If handwritten invoices are a significant portion of your intake, the verification pass described above becomes more important — but even at 85–90% accuracy on difficult handwriting, you're typing corrections instead of retyping entire documents.

What about AIA G702 payment applications specifically?

The column-name approach works on AIA G702 forms because the fields you need — Contract Sum to Date, Total Completed & Stored to Date, Retainage, Total Earned Less Retainage, Less Previous Certificates, Current Payment Due — are text-labeled on the form. The AI reads these labels and extracts the adjacent values, the same way it would on any other document. For a dedicated guide on AIA forms, see our AIA G702 data extraction guide.

Can the tool calculate Net Due from Gross Billed and Retainage?

Yes. Using computed columns, you can define a column like "Net Due (Total Billed − Retainage)" and the AI performs the subtraction during extraction. You get the calculated result directly in the spreadsheet — no post-extraction formulas needed. This is useful when subcontractors show retainage as a percentage but don't state the net payable amount explicitly on the invoice.

Does this integrate with Sage 300 / Viewpoint / QuickBooks?

The tool exports to Excel (XLSX) and CSV — both of which can be imported into any construction accounting system that supports file-based data import. There's no direct API integration with specific ERP platforms. The workflow is: extract to spreadsheet → review → import into your accounting system. For most small to mid-size contractors, this replaces the manual entry step without requiring an ERP integration project. Larger firms running Procore or Viewpoint Vista may want a tool that feeds directly into their cost ledger — but those platforms typically cost $15,000+/year and require dedicated implementation. If the bottleneck is "we can't get data off the page," the spreadsheet-first approach solves it without the overhead.

What happens if a subcontractor changes their invoice format?

Nothing needs to change on your end. Because the extraction matches by field meaning rather than page position, a new invoice layout from an existing subcontractor is handled the same way as an invoice from a brand-new subcontractor. No template updates, no retraining, no configuration. This is the single largest operational difference between template-based OCR and column-name AI extraction — and the reason the approach scales across dozens of subcontractors with zero maintenance.