The source is the full R&D guide, not merely a document checklist
The linked 108-page PDF is Research and Development Tax Measures, eighth edition, published on 30 January 2026. It explains project qualification, deductions and administration, then eight annexes covering legislative history, examples, costs, computations, allocation, equipment, software and food/beverages. Its monetary deduction charts still expressly describe YAs 2009–2025, and its PIC examples concern a scheme that expired after YA 2018. Those stated periods are preserved here rather than presented as an unrestricted current incentive. The edition date alone does not update every historical rate in its body. The R&D qualification principles also apply to the relevant historical PIC claims; the separate EIS guide addresses that scheme’s additional rules.
A qualifying project must satisfy three linked requirements
Section 2’s R&D definition requires a scientific or technological study with an objective of acquiring new knowledge, creating a product or process, or improving an existing one; novelty or technical risk; and a systematic, investigative and experimental study. Novelty and technical risk are alternatives, but neither replaces the objective or SIE requirement. Products can include materials, devices, services and equipment used inside the business. Basic research, applied research, experimental development and feedback research are all tested against the same requirements. A project can fail commercially and still qualify, because success is not a separate requirement. Conversely, calling expenditure innovation, development or engineering does not establish that these tests are met.
State the technical objective before work begins
Document the scientific or technological gap and the intended outcome before the R&D starts. A broad commercial goal such as selling more ovens is insufficient: the guide’s technical example is preventing overheating while doubling cooking speed, potentially by experimenting with casing materials. Market surveys about customer demand do not resolve that technical gap. A qualifying R&D component can sit inside a larger commercial project, but unrelated parts do not become R&D unless undertaken wholly or mainly to support the qualifying work. Describe the existing knowledge at commencement, what it cannot achieve, and the scientific or technological challenge the study will address.
Novelty goes beyond being new to the company
Novelty means first of its kind in Singapore in the relevant product, process or knowledge, beyond minor or routine upgrading. Merely importing an overseas product does not qualify without the objective and SIE requirements. Content-based products such as films or electronic print editions are excluded from this novelty discussion. Routine deployment of RFID to record exhibition visitors is not R&D simply because the company has not done it before. A genuinely new manufacturing process, an uncertain first adaptation across industries or a new use of materials may qualify if the remaining requirements are met. Literature and market searches must establish the relevant baseline, not merely assert that the taxpayer has a distinctive business offering.
Technical risk is uncertainty a competent professional cannot readily resolve
The uncertainty must exist at commencement and cannot be readily resolved from reasonably available knowledge, information or experience by a qualified and experienced professional in the relevant field. That professional need not be an employee. Standard soundproofing, ordinary fine-tuning, established technology integration and predictable customisation do not qualify merely because the exact configuration is initially unknown. New materials, lighter or smaller products retaining functionality, fundamental physical changes or genuinely uncertain technology integration can involve technical risk. An easy first-trial success may indicate that the outcome was readily deducible. A competitor’s closely guarded know-how can be unavailable even where the competitor already has a solution; information readily found online or procured from experts is generally available.
SIE requires planned investigation, experiments and learning
Systematic means an orderly, documented approach that another professional could reproduce. Investigative work explores the problem and the knowledge gap; it does not merely confirm a known fact. Experimental work tests possible solutions through structured steps, often repeatedly as the team learns. Keep a plan, milestones, team roles, test methods, results, evaluation, revised approaches and final conclusions. A full-time research team is not mandatory. Failed attempts and the resulting changes can be persuasive evidence of uncertainty. Random trial and error without structure, or tests demonstrating an established result, are insufficient. Supporting testing, data collection and even social-science work can qualify when wholly or mainly undertaken for the qualifying scientific or technological R&D project.
Seven excluded activity groups and ordinary deductions
The statutory exclusions cover quality control/routine testing, social sciences or humanities, routine data collection, efficiency surveys or management studies, market research/sales promotion, routine modifications and cosmetic or stylistic changes. Examples include regulatory tests of an unchanged cream, commissioning or calibrating equipment, financial/marketing/legal feasibility, simply reducing sugar concentration and changing spectacle-frame colours. These exclusions separate genuine scientific or technological investigation from ordinary business improvement. A non-qualifying project may still have a 100% ordinary revenue deduction under sections 14 and 15 where the expenditure is wholly and exclusively incurred in producing income; failure of the R&D test does not automatically mean every cost is nondeductible.
Identify when qualifying work starts and ends
A project starts once the technical objective is identified and activities directly addressing it commence. Preliminary commercial cost-benefit studies generally precede that point; later feasibility work exploring solutions can be integral to R&D. The project ends at the latest relevant completion of solution testing and conclusion, successful validation of a fully functional prototype/pilot plant, or termination. Further prototype modifications qualify only if they themselves satisfy the tests. A pilot plant used for experimental drug production can be R&D, but ordinary commercial production is not. In the fabric example, work targeting 25% stretch in both directions starts with that objective and solution development; after successful trials, board/customer demonstrations, patenting, marketing and production tooling are excluded.
The claimant must bear costs and benefit from the results
The claimant incurs the expenditure, bears the financial burden and effectively owns results that it can commercially exploit, such as know-how or intellectual property. It can undertake work itself, outsource it or participate in a cost-sharing agreement. An R&D service provider performing work for a customer on cost-plus or another fee basis is not the beneficiary and cannot claim the customer’s R&D benefit; its ordinary business costs can instead fall under section 14. A service provider’s own project improving its products or services can qualify if it bears the burden and benefits from the results. Related-party status does not by itself determine the answer.
Base deduction, additional deduction and project location
The guide describes a 100% section 14C deduction, additional 50% under section 14D(1) for YAs 2009–2018, and additional 150% for YAs 2019–2025; PIC under section 14D(2) belongs to its historical scheme. Ordinary in-house or outsourced Singapore R&D within the stated YA period can qualify even if unrelated to existing trade. Overseas R&D generally requires a trade connection for the base deduction and does not receive the Singapore additional deduction. Chart B for a mixed project first requires trade connection, then separates local enhanced treatment from overseas base treatment. These are the actual chart distinctions; the special CSA rules from YA 2018 are addressed separately. Use the applicable YA’s legislation and scheme guidance rather than extrapolating the chart’s 2025 endpoint.
Which costs qualify and what must be excluded
The base R&D deduction excludes government/statutory-board grants, amounts already deductible under section 14, section 15 prohibited costs such as share-option costs or depreciation, and capital costs of plant, machinery, land, buildings or acquired rights. The guide notes the specified statutory exceptions to the section 15 restriction. Pre-business qualifying expenses are treated as incurred at business commencement under the relevant pre-commencement rules. Additional deductions focus on staff costs and consumables, net of subsidy. Staff costs include employee remuneration, relevant specialist training/certification and associated overseas travel/living allowances, but exclude directors’ fees. Apportion part-time staff using actual R&D time and retain timesheets. Consumables are used up or irreversibly transformed, such as chemicals or components embedded in a prototype; rent, utilities and overheads are not consumables.
Outsourcing: the 60% deemed amount and subsidy allocation
For qualifying outsourced Singapore work, 60% of the fee is generally deemed eligible staff/consumable expenditure for additional deductions. A documented actual amount above 60% can be used with the required invoices and breakdown. A grant specifically funding qualifying costs is deducted directly: 60% of fees minus that grant, or actual qualifying costs minus it. A general fee subsidy is apportioned: 60% of fees net of grant, or actual qualifying costs less the grant multiplied by actual qualifying costs/total fees. The same proportional principle applies to general in-house subsidies. The YA 2013 example of a S$1 million fee gives a S$1 million base deduction and S$300,000 additional deduction at the then 50% rate. An employee cost of S$5,000 subsidised by S$1,000 leaves S$4,000 qualifying expenditure.
Annex D: the S$27 million expenditure example
For the YA 2019 illustration, local in-house costs are S$17 million staff, S$2 million consumables and S$1 million other costs; local outsourced fees are S$5 million, including S$4 million actual staff/consumables; overseas costs are S$2 million. Staff-specific grants are S$4 million in-house and S$1 million outsourced. The base deduction is S$16 million + S$4 million + S$2 million = S$22 million. In-house enhanced costs are S$15 million, giving S$22.5 million at 150%. Outsourcing gives S$3 million using the deemed base, or S$4.5 million using substantiated actual costs, so the latter is used. Additional deduction totals S$27 million and combined deduction S$49 million. No Singapore additional deduction is applied to the overseas S$2 million.
CSA payments have special rules from YA 2018
A CSA payment is the expenditure allocated to and borne by the participant, excluding any buy-in payment for joining the agreement. From YA 2018, section 14C gives the stated base deduction without excluding section 15 costs and without the related-to-trade condition. Chart C preserves separate enhanced eligibility: the relevant local work can qualify, while overseas work does not receive section 14D(1). Where the project is not trade-related, that local qualification remains subject to the chart’s location test. Enhanced qualifying expenditure is capped at the CSA payment allowed under section 14C; actual work performed and reimbursed under the CSA does not permit exceeding that cap. Related-trade CSA expense matching follows the normal specific/common income-stream rules.
Three historical CSA computations show the expenditure cap
In Annex D’s YA 2018 example, the S$1.5 million payment includes S$500,000 buy-in, leaving S$1 million deductible CSA payment; overseas staff/consumables are S$700,000. With S$800,000 local in-house costs and S$400,000 outsourced fees, the local deemed base is S$1.04 million, capped at S$1 million: base S$1 million, additional S$500,000 and historical PIC S$2.5 million total S$4 million, leaving no overseas PIC capacity. With S$400,000 in-house and S$200,000 outsourced, the local base is S$520,000: additional S$260,000, local PIC S$1.3 million and overseas PIC capped at S$1.44 million, again total S$4 million. With no local R&D, there is no section 14D(1) claim; overseas PIC is S$2.1 million and the total S$3.1 million. These are expired-PIC examples, not present-day PIC offers.
Allocate deductions to the income they support
Usually claim in the YA corresponding to the period the expense is incurred. Directly identifiable existing-product work is matched to that product’s income. New products/processes related to the existing trade or common support expenses are allocated across income streams on a reasonable basis closely connected to activity, such as turnover where appropriate. Qualifying local non-trade work is first deducted from normal-rate income; the remainder becomes normal unutilised losses for section 37A set-off against concessionary income. If only concessionary income exists, gross up the deductible amount by prevailing corporate rate/highest concessionary rate and deduct first against that highest-rate trade. Remaining losses can be set off against other concessionary streams under section 37A. Do not simply choose whichever stream gives the most convenient result.
Plant and machinery use capital allowances and balancing adjustments
Capital equipment is not an ordinary R&D expense deduction. The guide allows sections 19/19A capital allowances for Singapore R&D plant and machinery in YAs 2009–2025 even where the work is unrelated to existing trade. For non-trade local R&D, prescribed automation equipment can receive a one-year write-down from YA 2011; YAs 2009/2010 instead used working life or three years and excluded that one-year acceleration. Apply the same income-matching approach to allowances. Disposal or cessation circumstances under section 20 require balancing charges or allowances. A balancing adjustment for non-trade R&D is offset against trade income or taxed at the prevailing corporate rate as applicable. The historical periods and the asset’s actual allowance conditions matter.
Annex F: normal and concessionary-income examples
Both YA 2013 examples use S$50,000 non-trade local R&D, S$60,000 equipment written off over three years, a S$5,000 balancing charge and S$3,000 balancing allowance. With normal income S$300,000 and 10% income S$120,000, the historical deductions S$50,000, S$25,000 and PIC S$125,000, plus S$400,000 ordinary expenses, produce a S$300,000 adjusted normal loss. Adding the charge and subtracting allowances S$20,000, S$40,000 and S$3,000 gives S$358,000. The 10% stream has S$70,000; the source rounds its section 37A equivalent to S$41,000, leaving S$317,000 carried forward. Where only 10% and 5% streams exist, grossed-up R&D deductions are S$85,000, S$42,500 and S$212,500; non-trade allowance S$34,000 and balancing allowance S$5,100 produce a S$519,100 loss at 10%. Set-off equivalents S$8,500 for the normal-rate charge and S$35,000 for the 5% stream total S$43,500, leaving S$475,600. These figures reproduce the historical illustrations and their rounding.
Make the claim and retain contemporaneous evidence
Complete the R&D Claim Form for the relevant YA. Form C filers submit it with the return; Form C-S filers retain it and provide it when requested. Claims using outsourced actual costs above the 60% deemed amount need supporting invoices. Other evidence generally need not accompany the initial claim but must be available for review. Keep objectives and management approvals, baseline literature enquiries, novelty/patent evidence, professional explanations of uncertainty, experimental designs, milestones, procedures, measurements, failed/successful trials and conclusions. Obtain the corresponding research reports from an outsourced provider. Smaller businesses can support claims with working papers, email discussions and test scripts rather than a formal research plan. Evidence prepared only after the event is less useful for demonstrating what was unknown at commencement.
IRAS review, technical advice and the Assurance Framework
IRAS starts with the Claim Form and asks for further evidence if needed; it may consult IPOS International. Its Technical Advisory Panel advises under confidentiality undertakings, including where the taxpayer requests referral or wishes to pursue a claim rejected after two review rounds. IRAS makes the final decision, with the ordinary assessment objection/appeal route available. The R&D Assurance Framework can provide up to three YAs of upfront certainty, subject to requirements: robust internal evaluation and documentation, at least five in-house Singapore projects in the application YA and at least S$500,000 qualifying expenditure on them. Submit the application and complete project list at least nine months before the return deadline; IRAS selects projects for detailed review. If review is unfinished at filing, submit the completed R&D form for each project for that YA. The source lists 1800 356 8300 for sole proprietors/partnerships and 1800 356 8622 for companies.
Annex A: the policy history remains tied to its dates
Budget 2008 introduced liberalised deductions, R&D Tax Allowance and RISE. PIC arrived in Budget 2010; RDA and RISE were phased out from YA 2011, with granted RDA usable through YA 2016. Budgets 2011/2012 enhanced PIC. Budget 2014 extended the stated additional deduction and relaxed trade condition through YA 2025, and section 14E through YA 2020. Post-Budget 2017 feedback produced the CSA changes from YA 2018. Budget 2018 increased the additional rate to 150% for the specified YAs 2019–2025 after PIC ended. Section 14E lapsed after 31 March 2020. The guide’s amendment record adds software in October 2014, food/beverages in January 2015, CSA changes in December 2017, updated deductions/grant examples in January 2021, and renumbered statutory citations and the Assurance Framework in June 2022.
Annex B: four qualifying project examples
Gold-coated silver bonding wire qualifies because coating thickness and reliable chip bonding required literature searches and documented iterations, including oxidation and high-temperature tests. Lower-fat creamier ice cream qualifies through uncertain carbohydrate/fat/overrun interactions, investigations of heat, pressure and mixing, failed trials and peer discussions. The unconventional reinforced-concrete slab example required a new basement support solution on an inclined site, finite-element analysis of piles/struts and pressures, and recorded failure scenarios; the company bore costs and could reuse the result, but ordinary engineering designs are not automatically R&D. Membrane distillation for seawater qualifies through uncertain materials, configurations, flow and heat/mass transfer, followed by laboratory and large-scale testing with documented model deviations. Each example demonstrates the objective, novelty/risk, experimental method and contemporaneous evidence rather than merely an innovative final product.
Annex B: three non-qualifying examples
A VRV air-conditioning installation targeting 25% energy savings fails because the technology and achievable result were already known; modelling and fine-tuning do not make established performance uncertain. Customer-specific concrete grades fail where ordinary ingredient proportions and 28-day quality tests simply confirm expected strengths, with insufficient experimental records. Imported solar-powered golf carts satisfy the example’s first-in-Singapore novelty, but installation, commissioning and adjusting panel position do not meet the objective or SIE tests. This last example is important: novelty alone is not enough, and a failed installation followed by troubleshooting is not necessarily an experiment resolving unknown scientific knowledge.
Software: distinguish commercial objectives and routine development
Annex G covers accounting, word processing, CRM and trading applications. Productivity, customer loyalty, schedule pressure, unclear user requirements, staffing problems and missing old-system documents are not themselves scientific uncertainty. Standard inventory, imaging and authentication components must not be included without a proper basis. Integration can be difficult and resource-intensive without qualifying. Explain the technical limits of existing solutions: a reward-booking upgrade cannot rely only on cash rebates, whereas mathematical marking of unstructured student workings can identify a genuine gap. Routine coding, unit/system/user-acceptance/load tests verify intended functioning; they do not automatically investigate unknown methods. Describe the outsourced firm’s expertise and experimental work, and ensure the claimant benefits from the technology.
Software novelty: the specific technique and core matter
A biometric technique reducing false acceptance by two orders of magnitude can meet novelty, unlike minor adaptation of an existing third-party reward platform to travel. A new synchronisation method for 3D models and Street View imagery can qualify despite integrating existing tools; a new algorithm interpreting distressed swimmers’ movements can address knowledge absent from land-based gesture systems. If novel functions form the substantially significant core, essential peripheral functions may be considered with that core, subject to the other tests. A bespoke application is not novel merely because its combination is company-specific. Retain contemporaneous comparisons, technical pros/cons and explanations of why available Singapore solutions do not effectively resolve the problem.
Software technical risk and whole-project boundaries
Integrating more than ten incompatible payment options through a new secure connection can involve technical risk; re-indexing a slow CRM database with established methods does not. A derivative-trading core processing more than ten parameters within a minute without changing hardware can support whole-project treatment where that uncertain methodology is its core. In a hospital portal, only the uncertain indexing of mixed-language unstructured documents qualifies; standard scheduling, authentication, notifications and image storage remain outside. Tightly interdependent fish-behaviour monitoring and extraction components can be assessed together because they cannot function separately. The guide distinguishes these cases rather than declaring all large or complex software eligible.
Software SIE examples and the detailed trading-platform case
An e-learning method that detects, interprets and evaluates diverse unstructured workings can satisfy SIE through investigating symbol patterns and testing accuracy/reliability. New encryption with stronger security and fewer resources, rapidly adapting speech-to-text and improved lossless compression are illustrative possibilities, subject to all tests. The detailed proprietary trading example compares market alternatives, narrows three uncertain solutions with consultants, records marginal early results and investigates instability. Its research includes lightweight architecture, unique parameters, rules, time-critical queries, input filtering, execution speed, real-time intelligence and database management, with models and algorithm studies retained. Ordinary vendor selection, maintenance and debugging are not qualifying research. The insurance automation example fails because iterations and seamless interfaces merely implement established finance, treasury and HR rules, with no identified unknown technique.
Food and beverages: technical objectives and exclusions
Annex H examines products’ aroma, taste, texture, nutrition, calories, solubility and shelf life, and processes’ energy, time, spoilage and nutrient performance. Routine Brix, CO2, pH or sensory QA, nutritional laboratory analyses, new-supplier checks, certification maintenance and ordinary pathogen tests are excluded. Calibration, routine flavour extension and hygiene updates are not R&D; purely visual packaging or shapes are cosmetic. Qualifying objective examples identify uncertainty in lactobacilli chicken feed, flavour extraction replacing 0.2% MSG and reducing sodium 25%, or micronutrient fortification without damaging herbal soup properties. Adjusting bak kwa roasting time/temperature is routine. A preservative-free sparkling coconut drink claim is insufficient unless it explains existing storage limitations and acidity, nutrient, flavour and shelf-life challenges. Each objective example still needs the other R&D tests.
Food novelty: substantial characteristics, not a new flavour name
Novelty may concern physical, chemical, biological or sensory characteristics, or a genuinely new production and preservation process. The examples include an enzyme mimicking exercise effects, air-fried instant noodles with directly infused flavours, and precisely controlled self-heating packaging reaching about 50°C in two minutes. Their patent or technical evidence supports the novel aspect, but the other tests still apply. Textured branding cans are aesthetic changes, and grape-flavoured cheese using a known additive process is routine despite a different flavour. Overseas import alone does not qualify. Compare existing products and methods using contemporaneous literature, formulation, process and packaging analysis, rather than marketing descriptions.
Food technical-risk and SIE comparisons
Qualifying risk examples investigate heat-stable plant flavours, protective cultures preserving fruit ten extra days without chemical preservatives, and the source’s proposed bio-based film with unresolved interactions, solubility and degradation behaviour. These are case descriptions, not blanket scientific claims about all polyethylene. Routine use of known preservatives or converting batch production to continuous operation with predictable QC does not qualify. SIE examples include alternative 3D seasonings tested for viscosity, drying and heat resistance; yeast-extract umami with sodium-reduction experiments; and proprietary poultry-waste digestion, biogas, odour and compost trials. By contrast, salt blends with 50%/25% less sodium, optimising a tea machine and repositioning a pouch handle/straw are routine. Schedule, budget and regulatory delay are not technical risk. Record each iteration, its learning and how it closes the identified knowledge gap.
Three detailed qualifying food and process case studies
A clear-beverage additive qualifies where solubilising natural flavour/colour oils is unknown outside a competitor’s proprietary process; the company records pH, emulsions, addition order, catalysts, purification and pilot trials. The urban vertical farm investigates a novel hydraulic rotation system, drainage, sunlight, growth and energy: the source’s case uses 6–9 metre towers, 1 mm/second rotation, 0.5 litre water and 60 W per tower, with at least fivefold output and up to twelve crop cycles. These are that project’s reported parameters, not universal requirements. The paper beer bottle investigates pulp thickness, rigidity, wet/impact resistance, microbial and light protection, flavours and shelf life; the example reports over 70% production-energy reduction. Each case retains documented failed designs, revised experiments and measurable conclusions, so a green commercial goal alone is not the basis for qualification.
The two detailed food failures illustrate the boundary
A seafood-frying premix aiming for crisp texture, golden colour, adhesion and a spiky surface fails because the ingredient interactions and formulation methods were known; comparing commercial premixes, creating prototypes and recording results remain routine product development. Redesigned baby-food pouches fail where animal shapes, colours, fonts, matte/gloss finish and clearer nutritional displays are aesthetic rather than improvements to shelf life, durability or integrity. Consumer preference surveys, market statistics and supplier design plans do not become food-science experiments merely because organised and documented. The guide’s examples are general: variations can alter the outcome, but a claim must explain those actual scientific or technological differences.
Official source
This article independently explains the substantive contents of the official PDF, including the relevant conditions, procedures and annexes. The linked document remains the authoritative source for its original wording, and later changes should be checked separately.
Read the official PDF ↗