Chris Peters | R&D Tax Advisors

R&D Tax Credit for Startups: IRS 4-Part Test, Internal-Use Software Rules, and Audit-Ready Documentation

This episode explains the IRS framework for evaluating R&D tax credit eligibility in clear startup terms, including how to apply the four-part test to specific “business components” rather than an entire product, using an example to show how to isolate the true areas of technical uncertainty and experimentation. It clarifies the difference between external-facing software (standard four-part test) and internal-use software (IUS), which must meet an additional three-part test—innovation, economic risk, and why an off-the-shelf solution wouldn’t work—creating a higher seven-part threshold. The video emphasizes the need for contemporaneous, audit-ready documentation and suggests using existing artifacts from tools like Jira, Confluence, and GitHub, plus quarterly check-ins, to compile evidence over time and reduce audit risk while capturing meaningful credits.

R&D Payroll Tax Credit for Pre-Revenue Startups: How to Claim Up to $500K (Form 6765 + 8974)

The script explains how pre-revenue startups can use the federal R&D tax credit to offset payroll taxes (including the 7.65% payroll tax) without giving up equity, with examples of companies capturing roughly $50K–$197K and one engineering firm calculating about $150K from $1.5M in qualified wages. It outlines why the IRS enabled payroll offsets for startups, allowing up to $250K per year and later up to $500K, and gives qualification rules (the “555 rule”: within the initial five-year phase, under $5M gross receipts, and the offset available for up to five years). It details the filing process using Form 6765 with the payroll election and Form 8974 to apply the credit on quarterly payroll filings, clarifies that credits can be carried forward up to 20 years, and stresses meeting the IRS four-part test and strengthening documentation—especially ahead of expanded Form 6765 requirements becoming mandatory in tax year 2026—using existing engineering artifacts like Jira, Confluence, and GitHub.