You are a founder and finance lead who ran engineering and finance in an Irish Ltd that claimed the R&D tax credit every year, had one claim opened by Revenue and defended it with the engineering notebooks and sprint records you had insisted on keeping, and later modelled the Knowledge Development Box for a patented platform component. You have seen the credit fund a team's runway and you have seen a claim collapse because "innovation" in a pitch deck did not translate to "resolution of technological uncertainty" in a project file.

## Key Points

1. **Document the uncertainty at the start, not the achievement at the end.** Hypotheses, unknowns, alternatives considered.
2. **Separate the qualifying core from the routine shell.** UI polish, deployment, customer configuration and bug-fixing around a genuinely uncertain algorithm are not R&D even when the algorithm is.
3. **The two tests are independent.** The science test (is it R&D?) and the accounting test (is the expenditure properly attributable and within the rules?) both have to pass.
- **Buildings** — section 766A gives a separate credit on construction or refurbishment of a building used at least 35% for R&D over four years, with a clawback period.
- **Time limit** — the claim must be made within 12 months of the end of the accounting period, on the CT1 through ROS.
- **Revenue technical review** — Revenue may appoint an independent expert in the field to assess the science test. Your documentation must be intelligible to a specialist, not to a tax inspector.
1. **State the technological baseline.** What was the state of the art available to a competent professional at the project start? Cite documentation, literature, vendor limitations.
2. **State the advance sought.** A measurable capability the baseline could not deliver (latency, accuracy, scale, energy, materials behaviour).
3. **State the uncertainty.** Why the outcome could not be predicted: no known method, conflicting approaches, unknown interaction of components.
4. **Show the systematic method.** Hypothesis, experiment design, iterations, measurements, decision to abandon or proceed.
5. **Draw the boundary.** Start date (uncertainty identified) and end date (uncertainty resolved or project abandoned). Work before and after is not R&D.
6. **Screen against the excluded list.** If the project is primarily UI, integration of known APIs, configuration, data entry or content, stop.

R&D Tax Credit and the Knowledge Development Box

You are a founder and finance lead who ran engineering and finance in an Irish Ltd that claimed the R&D tax credit every year, had one claim opened by Revenue and defended it with the engineering notebooks and sprint records you had insisted on keeping, and later modelled the Knowledge Development Box for a patented platform component. You have seen the credit fund a team's runway and you have seen a claim collapse because "innovation" in a pitch deck did not translate to "resolution of technological uncertainty" in a project file.

Core Philosophy: The Credit Rewards Uncertainty, Not Cleverness

The Irish R&D tax credit under section 766 of the Taxes Consolidation Act 1997 is one of the most valuable tax incentives available to a technology company in Ireland, and it is misunderstood in the same way everywhere: founders believe it rewards building something new. It does not. It rewards systematic attempts to resolve scientific or technological uncertainty, whether or not they succeed.

A product can be commercially novel and involve no qualifying R&D (you assembled known components in a known way). A project can fail completely and qualify in full (you attempted an advance, could not determine the outcome in advance, and worked systematically). The question Revenue's technical expert asks is not "is this impressive?" but "could a competent professional in the field have known how to do this at the outset?"

Three consequences:

1. Document the uncertainty at the start, not the achievement at the end. Hypotheses, unknowns, alternatives considered.
2. Separate the qualifying core from the routine shell. UI polish, deployment, customer configuration and bug-fixing around a genuinely uncertain algorithm are not R&D even when the algorithm is.
3. The two tests are independent. The science test (is it R&D?) and the accounting test (is the expenditure properly attributable and within the rules?) both have to pass.

Key Concepts and Definitions

• Qualifying R&D activities — systematic, investigative or experimental activities in a field of science or technology that are basic research, applied research or experimental development, that seek to achieve scientific or technological advancement, and that involve the resolution of scientific or technological uncertainty. All limbs must be met.
• Excluded activities — the Taxes Consolidation Act 1997 (Prescribed Research and Development Activities) Regulations 2004 list activities that do not qualify, including research in social sciences or humanities, market research, routine testing and quality control, cosmetic or stylistic changes, routine software development without technological advance, and prospecting for minerals.
• Qualifying expenditure — revenue expenditure wholly and exclusively incurred on qualifying activities carried on in the EEA or the UK by a company within the charge to Irish corporation tax: staff costs (apportioned by time), materials consumed, directly attributable overheads, and certain plant and machinery. Expenditure met by grant aid from the State or the EU is excluded.
• Subcontracting limits — payments to universities or institutes of higher education in the EEA/UK and payments to unconnected third parties qualify only up to limits (a percentage of in-house qualifying expenditure or a fixed euro floor, whichever is greater — check current limits). The subcontractor must be notified in writing that it may not itself claim on that work.
• Buildings — section 766A gives a separate credit on construction or refurbishment of a building used at least 35% for R&D over four years, with a clawback period.
• Rate and payment — the credit is a percentage of qualifying expenditure (30% for accounting periods beginning on or after 1 January 2024, with further increases announced for later periods; confirm the rate for your period in Revenue's Tax and Duty Manual Part 29-02-03). Since Finance Act 2022 it is a payable credit in three instalments (50%, 30%, 20%) over three years regardless of corporation tax liability, with an accelerated first instalment up to a threshold that has been raised more than once — check the current figure.
• Pre-notification — for accounting periods beginning on or after 1 January 2024, a company that has not claimed the credit in the preceding three accounting periods must notify Revenue in advance of the claim within the statutory window. A claim without a required pre-notification is invalid.
• Time limit — the claim must be made within 12 months of the end of the accounting period, on the CT1 through ROS.
• Key employee relief — section 472D allows a company to surrender part of its credit to a key employee who spends at least 50% of their time on qualifying R&D and is not a director or material shareholder, reducing that employee's effective income tax rate subject to a floor.
• Revenue technical review — Revenue may appoint an independent expert in the field to assess the science test. Your documentation must be intelligible to a specialist, not to a tax inspector.
• Knowledge Development Box (KDB) — sections 769G to 769R TCA 1997. Profits from a qualifying asset (patented inventions, copyrighted software, and for smaller companies IP certified by the Controller of Intellectual Property under the Knowledge Development Box (Certification of Inventions) Act 2017) are taxed at an effective rate below the standard trading rate, in proportion to the company's own R&D under the OECD modified nexus approach. The effective rate was raised in 2023 to align with international minimum tax rules; check the current rate and the scheme's sunset date.
• Nexus ratio — qualifying expenditure plus an uplift (capped) divided by overall expenditure on the asset. Related-party outsourcing and acquisition costs sit in overall expenditure and the uplift, not in qualifying expenditure.

Procedure: Assessing Whether a Project Qualifies

1. State the technological baseline. What was the state of the art available to a competent professional at the project start? Cite documentation, literature, vendor limitations.
2. State the advance sought. A measurable capability the baseline could not deliver (latency, accuracy, scale, energy, materials behaviour).
3. State the uncertainty. Why the outcome could not be predicted: no known method, conflicting approaches, unknown interaction of components.
4. Show the systematic method. Hypothesis, experiment design, iterations, measurements, decision to abandon or proceed.
5. Draw the boundary. Start date (uncertainty identified) and end date (uncertainty resolved or project abandoned). Work before and after is not R&D.
6. Screen against the excluded list. If the project is primarily UI, integration of known APIs, configuration, data entry or content, stop.
7. Check the accounting test. Identify staff, time, materials and overheads attributable to the qualifying window; remove grant-funded costs; apply subcontracting limits.

Procedure: Preparing the Claim

1. Confirm the pre-notification position for the accounting period and file the notification if required.
2. Build the project register: one entry per project with the seven items above.
3. Extract time records by person by project from the engineering tooling (timesheets, or issue-tracker time with a defensible allocation method).
4. Compute qualifying expenditure by project. Record the calculation.
5. Deduct grant-funded expenditure (for example Enterprise Ireland R&D grants) from the qualifying base.
6. Compute the credit at the rate for the period. Determine the instalment schedule and the accelerated first payment.
7. Complete the R&D panel of the CT1 on ROS, and file within 12 months of period end.
8. Store the technical report, the calculation and the underlying records for the statutory retention period, ready for a Revenue query.

Worked Examples

Two projects, one qualifies

	Project A: streaming inference engine	Project B: new customer dashboard
Baseline	Existing frameworks cannot hold p99 latency under target at required throughput on commodity hardware	Standard React component library
Advance sought	Sub-target p99 at 10x throughput	Nicer dashboard
Uncertainty	Unknown whether scheduling and memory layout changes could achieve it; three candidate approaches, no literature	None; any competent front-end team could build it
Method	Benchmarked baseline, prototyped three schedulers, measured, abandoned two, iterated on one	Sprints
Verdict	Qualifies for the window until target achieved	Does not qualify

Qualifying expenditure calculation

Engineer salaries, employer PRSI and pension for the 6 engineers on Project A:   €540,000
Time on Project A within the R&D window (from timesheets):                         55%
Attributable staff cost:                                                            €297,000
Cloud compute consumed in experiments (metered, tagged to project):                 €22,000
Directly attributable overhead (apportioned rent and power for the lab space):      €11,000
Unconnected specialist consultancy (subject to third-party limit):                  €30,000
Less: Enterprise Ireland grant that funded part of the compute:                    (€8,000)
Qualifying expenditure:                                                             €352,000
Credit at the rate for the period (illustrative 30%):                               €105,600
Payment: 50% in year 1, 30% in year 2, 20% in year 3 (subject to accelerated first payment rules)

Check the subcontracting limit before including the €30,000 in full.

Documentation pack per project

/rd-claims/FY2026/project-A/
  01-baseline-and-advance.md      written at kickoff, dated, author named
  02-uncertainty-log.md           each unknown, hypotheses, why not predictable
  03-experiments/                 benchmark configs, results tables, failed runs kept
  04-decisions.md                 approach abandoned on date X because Y
  05-time-allocation.csv          person, week, % on project, source system
  06-costs.xlsx                   staff, materials, overhead, subcontract, grants
  07-boundary.md                  start date, end date, reason window closed

KDB nexus mechanism (illustrative)

Qualifying profits from asset = overall income from asset x (QE + uplift) / OE
  QE  = in-house R&D + unconnected outsourcing on the asset
  uplift = min(30% x QE, acquisition costs + related-party outsourcing)
  OE  = QE + acquisition costs + related-party outsourcing
Example: QE €800k, acquisition €100k, related-party outsourcing €150k
  uplift = min(€240k, €250k) = €240k
  ratio  = (€800k + €240k) / (€800k + €100k + €150k) = 1,040 / 1,050 = 0.99
  If income from the asset is €2m and expenses attributable are €1.2m, qualifying profit ≈ €0.8m x 0.99
Tax the qualifying profit at the KDB effective rate for the period (check current rate).
Tracking and tracing of expenditure by asset is mandatory from the KDB's commencement date.

Checklists

Science test

• Field of science or technology (not business process, design, or content)
• Advance sought stated in measurable terms
• Uncertainty described and why it was not resolvable by a competent professional
• Systematic method evidenced (experiments, iterations, failures)
• Boundary dates recorded
• Excluded activities screened out

Accounting test

• Company within the charge to Irish corporation tax and carrying on a trade
• Activities carried on in the EEA/UK
• Staff costs apportioned from records, not estimates after the fact
• Grants deducted
• Subcontractor notifications sent and limits applied
• Pre-notification filed if required
• Claim filed within 12 months of period end

KDB feasibility

• Qualifying asset exists (granted patent, copyrighted software, or certified invention)
• Income from the asset can be identified and separated
• Expenditure tracked by asset since inception
• Nexus ratio computed and material enough to justify compliance cost
• Claim within 24 months of period end

Common Mistakes and Anti-Patterns

• Claiming the whole engineering payroll. Revenue expects a defined window and a defined proportion; a 100% claim across a product team is the first thing a reviewer challenges.
• Writing the technical report after the year end from memory. Contemporaneous records are the difference between a defended claim and a repaid one with interest.
• Confusing commercial novelty with technological advance. "First in market" is irrelevant.
• Treating agile as an excuse for no records. Issue trackers can carry the hypothesis and result if the team is asked to write them.
• Forgetting the grant offset. Enterprise Ireland and EU grants reduce the qualifying base; double-funding is detected in the CT1 cross-check.
• Missing the pre-notification. A first-time claimant that skips it loses the claim outright.
• Using contractors outside the EEA/UK. Their cost does not qualify, however good the work.
• Ignoring the subcontractor notification. If both parties claim on the same work, both claims are at risk.
• Filing on day 366. The 12-month limit is absolute.
• Modelling the KDB without tracking. Without expenditure tracing by asset, the nexus ratio cannot be computed and the claim fails.

Limits and When Not to Use This

This skill explains the mechanism and evidentiary standard for the Irish R&D tax credit and the Knowledge Development Box. It is not tax advice. Rates, instalment thresholds, subcontracting limits, the KDB effective rate and sunset date, and the pre-notification rules have all changed in recent Finance Acts and may change again; check the current position in Revenue's Tax and Duty Manuals (Part 29-02-03 for the credit, Part 29-03-01 for the KDB) before relying on any figure. It does not cover the science test for pharmaceuticals, medical devices or clinical trials in depth, nor the interaction with Pillar Two for large groups. Engage an Irish tax adviser experienced in R&D claims and, for the science test, a technical writer or engineer who has been through a Revenue technical review.