Most disallowed credits aren't disallowed because the work didn't qualify. They're disallowed because nobody can prove it did, three years later.
Claims are rarely lost on the merits. They are lost on the file.
A business that genuinely performed qualifying work, claimed a reasonable amount, and cannot produce the evidence three years later is in a worse position than a business that claimed less and documented it properly. That asymmetry is the whole reason a study is worth commissioning rather than estimating.
For a research credit
An examination will generally want to see, project by project:
- What the business component was — the specific product, process, formula or software
- What the technical uncertainty was at the outset, stated concretely
- What alternatives were evaluated, and how
- Who performed the work, and the basis for allocating their time
- How wages, supplies and contract research were quantified
The recurring failure is the time allocation. A business says an engineer spent 60% of the year on qualified work. Where does 60% come from? If the answer is "that feels right," it will not survive. If the answer is a project-level methodology applied consistently and documented at the time, it generally will.
For cost segregation
Different evidence, same principle:
- The methodology used, and whether it was engineering-based or a residual estimate
- Construction documents and cost records supporting the component allocation
- The basis for each classification — why this item is 5-year and not 39-year
- Placed-in-service dates
An engineering-based study with documented allocation is what IRS guidance describes as the most defensible approach. Residual or rule-of-thumb studies are cheaper to produce and considerably weaker when someone asks how a number was reached.
The word that carries the weight
Contemporaneous. Documentation created as the work happened is evidence. Documentation reconstructed after a notice arrives is an argument.
Both may be truthful. They are not treated the same way, and they should not be.
What this means practically
If you are doing qualifying work right now, the cheapest examination defense available is capturing it while it is happening. Project narratives written in the year of the work. Time allocation methodology decided in advance and applied consistently. Test results and design iterations retained rather than discarded once the problem is solved.
That is most of what a properly run study builds — not just the calculation, but the file behind it.
Our position on it
We build every file expecting it to be examined, because occasionally one is. Audit defense is part of the engagement rather than an upsell: we hold the documentation, we handle the substantiation, and we work alongside your CPA if a notice arrives.
And where the substantiation cannot be built, we say so and stop. An unsupported claim is not worth the exposure — not to you, and not to us.