The highest-scoring accounting niche we track — NicheScore 88 in our scored ranking of accounting niches — is also the one with an expiration date built into the statute. That's not a flaw. The finite window is the reason the niche scores 19/20 on Demand Momentum and the reason a CPA can charge $5K–$15K for an engagement a generalist would bill hourly at a third of the value.
This is the full teardown: the mechanics, the window math, the buyer, the engagement anatomy, and the exact first-20-prospects play.
What broke in 2022, and what OBBBA fixed in 2025
The TCJA (2017) contained a delayed fuse: for tax years beginning after December 31, 2021, §174 research and experimental expenditures could no longer be deducted immediately. They had to be capitalized and amortized — five years for domestic research, fifteen for foreign — and the statute explicitly swept software development into the definition. With the midpoint convention, a software company deducted only 10% of its 2022 domestic development costs in 2022.
The arithmetic was brutal for exactly one class of company: profitable-ish, bootstrapped software firms whose largest expense is developer payroll. A firm with $1.2M in domestic development costs in 2022 could deduct roughly $120K of it that year — adding about $1.08M back to taxable income, which at the 21% corporate rate meant on the order of $225K in extra federal tax on income that existed only on paper. Many paid it three years running, 2022 through 2024. Some funded it with debt.
Then OBBBA, signed July 4, 2025, reversed course: it restored immediate expensing for domestic R&D (new §174A) for tax years beginning after December 31, 2024 — and, critically for this niche, it opened a retroactive path for small businesses. Firms meeting the §448(c) gross-receipts test (roughly $31M average, which covers essentially every bootstrapped software company) may apply the fix back to 2022–2024, recovering the tax they overpaid, generally via amended returns. Larger taxpayers instead accelerate their stranded unamortized amounts forward. IRS implementing guidance issued in 2025 set the procedural rails — election mechanics, method-change procedures, filing paths.
The result: a large, identifiable population of companies is owed real money, the recovery requires competent tax work, and the recovery path has deadlines.
Why the window is finite — and why that's the point
Three clocks run on this niche simultaneously.
The refund statute. Refund claims live under the ordinary three-year statute of limitations. A calendar-year 2022 return filed in April 2023 hit its refund wall in spring 2026 — as of this writing, 2022 is largely gone for unextended filers, still open into October 2026 for extended ones. 2023 refund claims generally run into 2027, and 2024 into 2028. Every quarter that passes, the recoverable pool shrinks by a vintage.
The procedural deadlines. The implementing guidance carries its own election and filing deadlines, layered on top of the refund statute. Verify both against the current guidance before quoting any client a window — the controlling dates live there, not in blog posts (including this one).
The competition clock. In late 2025 this work belonged to whoever showed up. Through 2026, generalist firms are waking up to it and specialty R&D shops are drifting down-market. The Competition Gap score (15/20) is a 2026 number; it will not survive to 2028.
Here's why the expiration date is a feature: a closing window converts prospects who would otherwise wait forever. Advisory sales usually die of "let me think about it." This engagement carries a statutory answer to that objection. You are not manufacturing urgency — you're reading it off the calendar, which is exactly the evidence-first posture that separates a specialist from a marketer.
The buyer: bootstrapped software firms under $10M
Be precise about who this is for, because the precision is the positioning. The three-part intersection: bootstrapped software firms × amended-return expensing recovery × the closing statutory window.
The ideal profile:
- $1M–$10M revenue, software or software-enabled, with 5–40 developers on payroll. Development payroll is the qualifying cost mass; headcount is the visible proxy.
- Bootstrapped or lightly funded. VC-backed startups running deep losses had NOLs to absorb the §174 hit; profitable bootstrappers wrote checks. They remember the checks.
- No Big 4 relationship. Their current accountant is a generalist who filed the capitalized returns correctly and never modeled the recovery. That accountant is not your competition; they're your evidence.
- Founder-led finance. The founder or a controller makes this decision in one or two meetings. No procurement, no committee.
Why they're underserved is structural, not accidental: specialty R&D-credit shops price for large filers, and the sub-$10M firm's potential refund — $50K to $400K — is too small for their model and far too large for the firm to ignore. That gap is the niche.
The engagement: $5K–$15K fixed fee, four phases
Fixed fee, value-priced against the recovery — never hourly. The client is buying an outcome with a known price; you're being paid for the specialty, not the hours. Anatomy:
Phase 1 — Eligibility screen and cost inventory (week 1). Confirm the §448(c) gross-receipts test. Inventory 2022–2024 domestic research costs: developer W-2 wages, allocable contractor spend, supplies, the software-development sweep. Identify what was capitalized on the filed returns — and flag the firms that never properly capitalized at all, which is a different (also billable) conversation.
Phase 2 — Quantification (weeks 1–2). Recompute each open year under retroactive expensing. Output: a one-page recovery memo — refund by year, interaction effects (state conformity, R&D credit under §280C, NOL movement), and the net number. This memo is the deliverable the client forwards to their co-founder; write it to be forwarded.
Phase 3 — Election and filing (weeks 2–5). Choose the path the guidance prescribes for their facts — amended returns for the retroactive election versus catch-up treatment going forward — prepare the returns and required statements, file, and calendar the refund tracking. This is where your one worked engagement template does the heavy lifting; build it on the first client, reuse it on the next thirty.
Phase 4 — The advisory handoff (built into the closing meeting). More below — the refund is the entry, not the business.
Pricing bands that hold up: $5K–$7K for a single amended year with clean books; $10K–$15K for the full three-year package with credit interactions. Anchored against a $150K recovery, a $12K fee is not a cost discussion. On effort, the work compresses hard once the template exists — eligibility screening and computation drafting are precisely the kind of structured analysis AI leverage collapses from days to hours (the niche scores 18/20 on AI Leverage for this reason), which is what makes a fixed fee at these bands a margin machine rather than a discount.
The first-20-prospects play: the free refund estimate
The acquisition move that fits this niche is the free §174 refund estimate — built entirely from public information, delivered before the prospect has ever spoken to you.
The mechanics:
- Build the list of 20. Local and regional software firms, 5–40 engineers, plausibly bootstrapped (no splashy funding announcements). Sources: LinkedIn headcount filtered by engineering titles, state business registrations, local "fastest growing" lists.
- Model each one. Engineering headcount × a defensible fully loaded developer cost (~$150K) × the share plausibly qualifying as domestic §174 cost × the capitalization delta for 2022–2024 × 21%. You now hold a range, not a number — say so. The honesty is the credibility.
- Send the one-pager. "Based on your public engineering headcount, we estimate your firm overpaid roughly $120K–$210K in federal tax across 2022–2024 under the old §174 capitalization rules. Congress opened a retroactive fix in July 2025. Part of the window has already closed; the rest closes on a statutory schedule. Fifteen minutes and your actual returns will tell us whether the estimate is real."
- The diagnostic call converts. They bring the returns; you refine the estimate live; the fixed-fee quote follows the same day. A cold list of 20 built this way reliably produces conversations because the letter contains their number, not your brochure.
This is the same find → position → win arc every scored niche runs on. The estimate model, the outreach sequence, and the diagnostic-call script are exactly the kind of assets the LibSkills Accounting Pack skills draft in an afternoon — the Niche Finder scores your local market, and the outreach engine writes the sequence that books the calls.
After the refund: the advisory upgrade
The refund engagement ends. The relationship shouldn't. You are now the specialist who recovered six figures for a software company — the trust ceiling is gone, and three follow-ons are standing in the closing meeting:
- The §41 R&D credit study. Same cost inventory, different computation — for many clients you've already done 60% of the work. Annual, recurring.
- Quarterly tax planning. The firm that overpaid $200K without noticing needs estimates and planning cadence. $1,500–$3K/quarter, fixed.
- The fractional-CFO path. Some of these clients are the exact profile of the bootstrapped-SaaS fractional-CFO niche — the natural ceiling of the hourly-compliance-to-recurring-advisory conversion this whole strategy runs on.
Structure the handoff, don't improvise it: the recovery memo's final section should be titled "what this found besides the refund," listing two or three planning gaps with annualized dollar impact. The upgrade conversation then happens on the client's momentum, not your pitch.
The honest caveats
Two, stated plainly. First, verify the current deadlines before promising anyone a window — this article is dated, the guidance controls, and part of the 2022 vintage has already lapsed. Second, this niche ends. By 2028 the amendment work is done. Enter it knowing that the exit — the client base and the recurring advisory built on it — is the actual asset, and keep a scored shortlist warm for what you claim next. The free Sub-Niche Opportunity Report re-scores 25 niches across accounting, legal, and real estate every quarter for exactly that reason; and the broader playbook for converting any scored niche into a positioned practice is in the pillar guide: The Most Profitable Accounting Niches in 2026.
A closing window is only a problem for the firms that hesitate. For the one that moves this quarter, it's the entire business case.