How to Implement R&E Amortization Under Section 174 for Programming Costs?

I've been digging into the recent IRS guidance about R&E amortization under Section 174 (where programming costs need to be amortized over five years, unless it's maintenance programming which can be expensed in the current year) and I'm scratching my head on implementation. Here's what I'm considering: going through all our codebase for the year and marking each code line as either maintenance or new development. If programmer Jane wrote 15,000 lines of code and 3,000 was maintenance, I'd expense 20% of her salary this year and amortize 80% over five years. But I see some issues with this approach: * Time spent coding isn't consistent per line - maybe those 3,000 maintenance lines took just 2 weeks while the remaining lines took 50 weeks. Should I instead track task time and categorize tasks as maintenance vs. new development? But then what about mixed tasks? * Lines of code vary wildly between languages and coding styles. If one dev works in both Python and Java in the same year, the line count comparison becomes meaningless. Has anyone implemented this Section 174 stuff in practice? I've talked to three accountants and they all seem equally confused about real-world application. Any insights from those who've figured out a workable approach?

0 coins

18 comments

Logan Greenburg

•8M

Tax accountant here. The R&E amortization under Section 174 is definitely causing headaches for many tech companies. Your approach shows good thinking, but I wouldn't recommend the line-by-line code analysis - it's too granular and would be difficult to defend in an audit. Instead, consider tracking time by project or feature, where you categorize each project as either maintenance or new development. If a project has elements of both, you'll need to make a reasonable allocation based on the primary purpose. Documentation is key here - maintain records of how you classified each project and why. Most of my clients are using timekeeping systems where developers log hours to specific projects, then the accounting team works with tech leads to determine the appropriate classification. The focus should be on function and purpose rather than volume of code.

0 coins

Logan Greenburg

•8M

Projects should be granular enough that you can reasonably classify them as predominantly maintenance or new development. For large initiatives, I recommend breaking them down into smaller components that can be more clearly categorized. This might mean working with your project management team to restructure how work is tracked. For DevOps engineers and infrastructure work, look at the primary purpose of their activities. If they're setting up new infrastructure for new capabilities, that's likely amortizable R&E. If they're maintaining or scaling existing systems without adding new functionality, that may qualify as maintenance. Documentation of your methodology is crucial here - be consistent and keep records of your reasoning.

0 coins

Lucas Bey

•8M

I went through this exact same headache last quarter. After spending weeks trying to figure it out, I found a tool called taxr.ai (https://taxr.ai) that really helped us tackle this Section 174 issue. They have a specific module for R&E expense classification that analyzes your development processes. What worked for us was uploading our sprint documentation and git commits, and their system helped identify patterns of maintenance vs. new development work. It's not perfect, but it gave us a much more defensible methodology than our original manual approach. The best part was that it produced documentation explaining the classification rationale, which our accountant said would be helpful in case of an audit.

0 coins

Lucas Bey

•8M

Yes, it does integrate with Jira! That was actually one of the main reasons we chose it. You can connect directly to your Jira instance and it pulls in all your tickets, epics, and sprint data automatically. Saved us tons of time compared to manual exports. For the tax classification concerns, the tool doesn't make final determinations - it suggests classifications based on patterns it identifies, but you (or your tax professional) make the final call. What I found valuable was that it provided a consistent methodology and documentation of our reasoning. Our tax accountant reviewed everything and made adjustments where needed, but said the approach was much more systematic than what most companies are doing.

0 coins

Caleb Stark

•8M

I tried taxr.ai after seeing it mentioned here and I have to admit I was wrong about my skepticism. The tool actually helped us identify patterns in our development work that we hadn't noticed. Our team was classifying certain types of database optimization as new development, but the analysis showed they were more aligned with maintenance characteristics according to the IRS guidance. The documentation it generated was detailed enough that our tax firm felt comfortable with our approach. We ended up with about 35% maintenance (expensed) and 65% capitalized R&E, which seems in line with industry benchmarks for our type of software. I'm actually sleeping better knowing we have a defensible methodology in place.

0 coins

Jade O'Malley

•8M

Has anyone tried contacting the IRS directly about this? I've been waiting on hold for days trying to get clarification on these Section 174 rules for our software company. It's ridiculous how hard it is to get actual guidance from them. A colleague mentioned a service called Claimyr (https://claimyr.com) that apparently helps you get through to an actual IRS agent. There's a video showing how it works: https://youtu.be/_kiP6q8DX5c Has anyone tried this to get Section 174 guidance? I'm desperate for official answers at this point.

0 coins

Jade O'Malley

•8M

It's not magic - from what I understand, they use an automated system that continually calls the IRS and navigates the phone tree until it gets a place in the queue, then it calls you and connects you when an agent is available. So instead of you waiting on hold personally for hours, their system does it for you. I've been told it works because they've built technology to navigate the IRS phone system efficiently. It's not about "cutting the line" - it's about automating the awful hold process so you don't have to do it yourself.

0 coins

Ella Lewis

•8M

I stand corrected about Claimyr. I tried it yesterday out of desperation after our CPA gave us conflicting advice about our Section 174 calculations. I was connected to an IRS representative in about 90 minutes without personally waiting on hold. The agent wasn't able to give detailed guidance on implementing Section 174 for software development specifically (apparently they need more industry-specific training), but she did connect me with a specialist who helped clarify some of the basic requirements. The specialist confirmed that we need a "reasonable method" for separating maintenance from new development, and that time tracking by project function is acceptable if consistently applied. Worth the try if you need direct IRS input, though don't expect software-specific expertise.

0 coins

Andrew Pinnock

•8M

We just implemented a hybrid approach at our company that seems to be working well. Rather than counting lines of code (which is problematic for all the reasons you mentioned), we: 1. Had each development team estimate the percentage of maintenance vs. new development for their area 2. Set up time tracking codes that developers use when logging hours 3. Created a review process where tech leads and finance meet quarterly to review the classifications 4. Document everything with written justifications for how we classified each major component Our CPA seemed satisfied with this approach, though she emphasized that we need to be consistent and have solid documentation of our methodology.

0 coins

Andrew Pinnock

•8M

Yes, we've definitely seen that transition from new development to maintenance over time. As features mature, work on them tends to shift from primarily new development to mostly maintenance and refinement. We actually created a simple lifecycle model where new features start as 100% development, then after initial release they transition to a mixed classification, and finally to predominantly maintenance after they've been in production for a certain period. The exact timing varies by feature complexity, but having this framework helps us be more consistent in our classifications over time.

0 coins