A self-led, step-by-step guide for New Zealand founders who want to claim both R&D tax credits themselves, without paying an accountant or an R&D consultant a slice of the money. It answers the two questions every claim turns on, in order: is your work genuinely novel, and what can you actually claim and how. It covers the 15% R&D Tax Incentive and the 28% R&D Loss Tax Credit as one journey, from working out which pot of money is yours through to protecting the claim if Inland Revenue asks questions later.
Cormac Denton | Playdough
builds The DIY RDTI: Claim New Zealand's R&D Tax Credits Yourself
A self-led guide to claiming both of New Zealand's R&D tax credits yourself, so the money supports your R&D instead of a consultant's bank account.
Cormac Denton is the founder and fractional CFO at Playdough, a Chartered Accountancy and Fractional CFO practice. He spent about a decade in senior finance and operational roles across London and Wellington at brand names, like Telefonica (o2), Westfield and 2Degrees.
He has worked as an internal CFO at growth companies, including media tech company Mood Media; Canada’s fastest-growing company, 72x revenue growth in five years, Kapura; now Star Group, helping the hospo group 5x Revenue, 10x Profit in four years, and as GM Services and Finance at Reveal, a deep-tech company doing the kind of R&D this guide covers.
The Playdough team works with founders looking to start, scale and grow, where often R&D is a core part of the cash flow before revenue arrives.
Cormac is exactly the person a founder would hire to handle an R&D claim. He is also the one telling you to try it yourself first.
🚨 General information, not tax advice. This edition was last reviewed July 2026. The rules, thresholds and dates may change, and Budget 2026 measures are proposals until they become law. Check rdti.govt.nz and ird.govt.nz, and take advice from an R&D Expert on your own situation before you rely on any of it. R&D Experts are usually a chartered accountant or an R&D specialist experienced in RDTI claims. Start with the free MBIE review before paying anyone.
The DIY RDTI, in the order founders ask:
This is New Zealand's RDTI. Australia runs a different scheme with the same name and much higher rates. If an AI told you 43.5%, it was reading the Australian one.
Part 1: The Money Map
There is more than one pot of money for Kiwi startups doing real development. They have different rules, and some stack together. Here is the whole map, in plain English. Read it first, then use Part 2 to check eligibility and Part 4 to write your application.
R&D Tax Incentive (RDTI), 15% credit.
The main one, and non-dilutive. 15% of your eligible R&D spend, every year, for work you are already doing. Profitable? It reduces your tax. Loss making? It comes back as cash, but only up to the PAYE, ESCT and FBT you paid that year, so the cash back this year is capped at those taxes and you may not get the full 15%. If you are part of a wholly-owned group, labour taxes from other group companies can lift this cap. The rest is not lost, it carries forward to cut future tax (subject to shareholder continuity).
You need $50k+ of eligible spend a year, waived if an approved research provider does the R&D. Apply during the year or after it ends, but before the cut-off: the last day of the 3rd month after your balance date (30 June for a 31 March year). It runs through MBIE and Inland Revenue jointly, in myIR, not Callaghan anymore. Almost every startup doing genuine development should be claiming this, and it is mostly admin, so do it yourself. Spending over $2m a year on R&D? A different route (CAM) applies, with a much earlier deadline, and most founders do not use it.
R&D Loss Tax Credit (RDLTC), 28% loss cash-out.
For loss-making startups. Turn this year's tax losses into cash now, instead of carrying them forward. You get up to 28% of your R&D tax losses as cash now, separate from and on top of the 15%. Pre-profit, the 15% has little tax to offset, so this is what puts real cash on your burn. You need to be NZ resident, loss-making, with R&D a decent share of your wages (roughly a fifth, a set formula applies), and you own the IP. It runs through Inland Revenue, on your income tax return. The big catch: it is not a grant, it is an interest-free loan. You repay it out of tax once profitable, or sooner on a recovery event (selling more than 90% of shares, selling the R&D IP, liquidation, moving offshore). Model it first. Part 6 covers this in full.
In-year payments, up to 80% as a loan.
Cannot wait until after balance date? Draw up to 80% of your expected credit during the year, up to 3 draws, once you have approval in place. It is an interest-free government loan administered by Tax Management NZ (TMNZ). Two separate things here: the TMNZ loan is live now, and a direct quarterly payment of the credit is one the Government has proposed from 2027-28, not law yet.
The free helper most founders miss, $0.
MBIE Innovation Services will check your eligibility and read your draft application at no cost. It is free, non-binding, and they know what a good application looks like. Use this before you pay anyone. Request it via the short form at rdti.govt.nz free support, and a Customer Engagement Specialist from a science or tech background will read your draft.
Grants, 40% co-fund, a separate track.
These are not tax credits. You apply in advance, they are competitive, and they are paid in arrears, run by MBIE not IRD. The New to R&D Grant co-funds 40% of eligible costs, up to $400k, for businesses new to R&D. You qualify if you have spent under $150k on R&D in 3 years, can fund the other 60%, and intend to keep going. The R&D Experience Grant (hire students) and R&D Career Grant (hire graduates) are open too.
Got another grant for the same work? Pull it out.
This is the double-dip rule, and it catches people. If a government grant (an MBIE grant, an MPI grant, a Callaghan-legacy grant, any of them) has already funded some of your R&D spend, that funded spend is out of your RDTI claim. You cannot claim the same dollar under both. You can still claim the RDTI on your own R&D spend that the grant did not cover. The trick is separating grant-funded costs from company-funded costs, cleanly, with records that show the split. Grant co-funding also cannot come from other government money.
The two dates that kill good claims.
Everything else can be fixed or deferred. These cannot. General Approval: the last day of the 3rd month after your balance date (30 June for a 31 March year). Supplementary return: 30 days after your income tax return is due, off the normal calendar, so it is the one most missed.
See Your General Approval Deadline (2025/2026 Tax Year) for more details.
Which best describes your startup?
Pick the one that fits you best, then follow that row.
- Physical / deep-tech: building a product, material, process or machine, not software? Same credits apply. RDTI (15%), plus RDLTC if loss-making.
- Profitable: doing genuine R&D and paying tax? RDTI (15%). Straightforward, do it yourself.
- Loss-making: burning cash, no profit yet? RDTI + RDLTC together. This is where the real money is, and where the numbers need care.
- Tight on cash: cannot wait a year for the credit? Add in-year payments through TMNZ.
- New to R&D: just starting, spent little so far? Start with the New to R&D Grant, then move onto the RDTI.
- Just a wrapper: mostly using an existing AI model or building software using standard features? Probably none of these yet. Honest answer: come back when you are advancing the method itself.
.png){kind=logo}
Not sure if you're doing R&D?
The hardest question isn't how to claim the RDTI - it's whether your business is actually doing R&D in the first place.
Read Expert Edition NZ Government Support: A Step-by-Step Execution Guide first to work out where your startup fits, then come back here once you've confirmed you're on the R&D pathway.
Part 2: Is Your R&D Actually Eligible?
A free, honest self-check. Three knock-out gates decide whether you can claim at all, before we look at whether your work qualifies. All three must be a yes. Then four tests, all four must pass.
The three knock-out gates (all must be yes)
1. Does your business own the results of this R&D? | You must own the R&D, or be able to use it at no extra cost. If you do the R&D on contract and the client owns the results, that is the agency trap, and it is the most common false start. You generally cannot claim the RDTI on that work, they might. You may still have eligible R&D in your own product or internal tools, so it is worth separating that out with an R&D Expert. |
2. Do you run a business through a fixed establishment in New Zealand? | The RDTI is for NZ businesses carrying out R&D through a fixed establishment here where substantial business is carried on, with at least one NZ-resident director. If you are an overseas-owned subsidiary the rules get technical (ownership, who the R&D is for, offshore limits), so check with an R&D Expert before relying on a claim. |
3. Are you a Crown Research Institute, Health New Zealand or a tertiary education organisation (or controlled by one)? | These entities, and businesses they majority own or control, are excluded. If you only receive funding from one but are not controlled by it, check the detail with an R&D Expert. |
This guide assumes one NZ company that owns its R&D. If you are overseas-owned, part of a group, using an approved research provider or heavy on contractors, or spending over $2m, extra rules apply. Get those checked.
The four tests (all must pass)
1. Does your work solve a real technical unknown? | Judged at the time you started: could a skilled person in your field have known the answer then, using what was publicly available? Not whether the answer exists now. If the answer became public later, or a competitor solved it after you began, that does not disqualify you. Only your core activity, the part that actually resolves the uncertainty, sits at the heart of a claim. Supporting activities exist only to serve the core R&D. The trap is dressing up ordinary business-as-usual work as supporting. Record the search you did at the start, dated: what you Googled, the papers, patents or competitor products you checked, and why none of them answered it. That dated search is your proof of novelty. |
2. Did you work at it methodically? | A plan, a test, results you wrote down. Not just trying things until one stuck. Notes made at the time carry a claim: the uncertainty, what you tested, the R&D time spent, and how each cost links to the work. Records reconstructed at year end are weak and are exactly what gets a claim questioned. Thin records are the top reason claims get stuck. |
3. Is your R&D mainly software or AI? | Counts: a novel algorithm, a new machine learning model where no solution exists, new cryptography, a genuinely new architecture for a hard performance or scale problem, advanced threat protection. Does not: standard payroll, accounting or CRM builds on known tech, or wiring existing tools and APIs together, or wrapping an existing model. The word AI does not decide it, the tests do. Using an existing model or API to build a product is almost never core R&D. Fine-tuning a model to solve a genuinely unsolved task can be core; wiring RAG over an existing model is not. |
4. Do you have $50k+ of eligible R&D spend? | Careful here: eligible spend is not your total project cost. It is only the R&D share, after stripping out non-R&D time, admin, sales, and excluded costs. Most founders overestimate this at first. The floor is waived when an approved research provider does the R&D.
Check the official approved research providers list. |
Your result
This is a self-check, not approval. Only Inland Revenue confirms eligibility, and the free MBIE review is your safety net. If you cannot explain the technical unknown to a ten year old, do not lodge yet.
- Looks worth pursuing. On your answers this has the shape of eligible R&D. That is a good sign, not a decision: only Inland Revenue can confirm it. Find your deadline, draft your application, and get the free MBIE check before you lodge.
- Worth a closer look. There is real R&D here, but at least one answer sits in the grey. That usually means narrowing to the part that genuinely qualifies.
- Take advice before you rely on this. At least one answer points away from eligible R&D. Sometimes there is still a narrow claim in there, sometimes not.
Whichever result you land on: do not pay anyone yet. Use the free MBIE service first. If they see a real but complex claim, a paid conversation earns its keep, and a fixed quote beats a percentage.
 This is a self-check, not approval. Only Inland Revenue can confirm your R&D is eligible, after they assess your General Approval application.
Worked examples
Picture a small Waikato agritech startup, three founders, building software that predicts pasture growth from satellite imagery so farmers know when to move stock. Call them Paddock. It all felt hard to build, so their instinct was to claim the whole year of costs.
What wins: New Zealand has constant cloud, and Paddock found no existing method could infer pasture growth through the clouds at paddock resolution. They did not know at the start whether it was possible. They tried one approach that failed, then a second that partly worked, and measured both against real farm data. That uncertainty, tested methodically and recorded at the time, is what the four tests reward.
What loses: the farmer app, the login, the mapping layer, the weather API wired in. Skilled engineering, real product, no technical unknown. Every piece already existed and worked. They were assembling proven tools, not advancing a method. The claim that gets stuck lumps it together, because the moment IRD asks what was technically unknown about the app, the whole thing wobbles. The claim that stands names the cloud cover problem as the core activity, the pipeline feeding it as supporting work, and leaves the dashboard out. Carving out what does not qualify is what protects what does.
Picture a solo founder out of Nelson building a low-power remote sensor pod that survives high-salinity tidal surges in Tasman Bay while transmitting real-time data by satellite. If it works she will call it the Bobi Buoy. She has real problems to solve: bio-fouling mitigation, power management and thermal degradation, and material stress and sealing. The work needs external testing by consultants, 3D printing, test materials, and a part-time PhD student, over about two years. An MBIE grant covers the first six months, $20,000, tied to specific bio-fouling milestones.
What wins: off-the-shelf enclosures fail as bio-fouling blocks the sensors, and standard battery chemistry degrades in prolonged sub-zero marine conditions. The right mix of coating frequency and battery insulation has no off-the-shelf blueprint, so it needs systematic trial and error. That uncertainty, tested methodically and recorded at the time, is what the four tests reward.
What loses and the grant trap: the $20,000 grant funds bio-fouling work over that six-month window, so those grant-funded costs cannot also be claimed under the RDTI. But R&D on other activities in the same period, say power management, can still be claimed, as long as you meet all the conditions, including at least $50k of eligible R&D spend for the year excluding the grant-funded amounts. Records and timekeeping matter here, because both MBIE and IRD will check there is no double dipping.
Part 3: Who Runs It Now
Callaghan Innovation was disestablished on 1 December 2025. Its RDTI support responsibilities moved to MBIE. Inland Revenue decides claims. Everything goes through myIR. Free support is now MBIE Customer Engagement Specialists via rdti.govt.nz. If a page says Callaghan runs it, it is old.
Don't stop at the RDTI.
New Zealand offers far more than R&D tax credits. Explore the NZ Government Support Directory to find grants, funding, capability programmes, and other government support your startup may be eligible for. Updated for 2026.
Part 4: Claiming, Step by Step
- Apply (General Approval) in myIR: rdti.govt.nz/how-do-i-apply
- Claim (supplementary return) later: rdti.govt.nz/how-do-i-claim
- Check your exact deadline: rdti.govt.nz/due-date-calculator
- Free draft review: request at rdti.govt.nz/we-provide-free-tailored-support
Step 1. Enrol in myIR. | One-time, and it unlocks the RDTI account. |
Step 2. Write your General Approval. | Describe your R&D the way Inland Revenue will read it. IRD reads hundreds of claims and spots templates. Use your own uncertainty, your own tests, your own numbers. They do not want hundreds of pages about what you do, your go-to-market, or how good your team is. Rather concise context of the business, and what was technically unknown, and how you tackled it. Describe your core activity, and the part that actually resolves the uncertainty, so they can reach the same conclusion.
Illustrative example only, to show the shape of a strong answer:
• ✗ Instead of: "We built a smart matching feature customers love."
• ✓ Write: "We could not tell if matching would stay above 90% accuracy past 10M records. Existing methods failed, so we designed and tested approaches, measuring accuracy and speed."
If your uncertainty paragraph contains any of these words, reword it: commercial, traction, market, customer, revenue, sell, sales, profit, growth, scale. IRD does not care what it does for the business, only what was technically unknown. Good wording never rescues thin evidence. |
Step 3. Total your eligible spend. | For each person, annual cost multiplied by the percentage of their time on R&D, with a basis you can defend (timesheets). Then contractors doing R&D, consumables used up, depreciation on R&D gear, and overheads genuinely tied to R&D. Eligible spend is not total project cost. Strip out admin, sales, management, land, interest, and grant-funded costs. If the eligible total lands under $50k, you cannot claim on your own spend this year. A "Marketing Manager" can get knocked back even for real R&D, so record what they actually did, not the title.
Three things to note:
• Your 15 percent is a credit, not always cash. If you are profitable it lowers the tax you owe. If you are loss making, how much comes back as cash is capped, and Part 5 is where that matters most.
• Only up to 10 percent of your total eligible spend can be R&D performed overseas, for example an offshore contractor. The rest must be done in New Zealand.
• All these figures are net of GST, if you are GST registered. |
Worked example (the 15% credit in dollars): your project cost $200,000, but after stripping out admin, sales and non-R&D time, $120,000 is defensible eligible R&D spend. Your credit is 15% of $120,000 = $18,000. Profitable? It cuts your tax bill by $18,000. Loss making? It comes back as cash up to the PAYE, ESCT and FBT you paid, and any excess carries forward. | |
Step 4. Get the free review. | An MBIE Customer Engagement Specialist reads your draft before you lodge. Request it at rdti.govt.nz/we-provide-free-tailored-support. |
Step 5. Lodge General Approval in myIR | before your deadline.
One approval can cover up to three income years. You can apply during the income year or after it ends, but never before it starts. |
Step 6. File the Supplementary Return | within 30 days of your income tax return due date. It sits off the normal calendar, so it is the one most missed. |
Before you lodge:
Your General Approval deadline (2025/26 income year)
The rule: the last day of the 3rd month after your balance date. Confirm the exact date with the due date calculator and use the official General Approval template on the apply page.
Balance date | General Approval deadline |
31 Oct 2025 (early) | 31 Jan 2026 |
30 Nov 2025 (early) | 28 Feb 2026 |
31 Dec 2025 (early) | 31 Mar 2026 |
31 Jan 2026 (early) | 30 Apr 2026 |
28 Feb 2026 (early) | 31 May 2026 |
31 Mar 2026 (standard) | 30 Jun 2026 |
30 Apr 2026 (late) | 31 Jul 2026 |
31 May 2026 (late) | 31 Aug 2026 |
30 Jun 2026 (late) | 30 Sep 2026 |
31 Jul 2026 (late) | 31 Oct 2026 |
31 Aug 2026 (late) | 30 Nov 2026 |
30 Sep 2026 (late) | 15 Jan 2027 (IRD shifts with holidays) |
 Big spender note. Over $2m a year of R&D uses a different route (CAM) with a much earlier deadline: the last day of the 6th month before year end, so 30 September for a 31 March year. Most founders use General Approval, not this.
Part 5: The Loss-Maker's Stack (RDLTC)
The 15% is not the real money if you are pre-profit. The 28% is.
A loss-making startup spending $310k on R&D can put roughly $87k of cash back this year via the RDLTC (Research and Development Loss Tax Credit), on top of a 15% RDTI credit.
The RDTI gives you 15% back, but if you are loss-making there is little tax to offset, so it comes back as cash only up to a cap tied to the PAYE, ESCT and FBT you actually paid. Real, but modest. The R&D Loss Tax Credit is the bigger lever: it lets you cash out this year's tax losses at 28 cents in the dollar, now, instead of carrying them forward to a profitable year that might be five years away. Together the two schemes can put back a meaningful share of every R&D dollar you spend. Loss-making and doing genuine R&D is the one situation where this system is actually generous. It is also where the numbers need the most care.
Do you qualify? Four checks.
1. You are a company. | NZ-resident, unlisted, and not 50% or more owned by a public body. Sole traders, partnerships and look-through companies are out; their losses flow to the owners instead. Check here: https://companies-register.companiesoffice.govt.nz/ |
2. You are in a tax loss for the year. | What you will be loading into your IR4, Company Return. |
3. You pass the wage intensity test: at least 20% of your total labour cost is R&D labour. | R&D labour includes salaries for people doing R&D and the R&D portion of any shareholder salary. Contractors doing R&D count at 66% of what you paid them, net of GST. If your whole team is building, you will clear this easily. If most of your payroll is sales and admin, run the numbers before assuming. |
4. You own the R&D and its results. | If someone you contract for owns the output, or you cannot use it freely, you are out. This is the same ownership test as the main RDTI. |
What you get
28% of your eligible R&D losses as cash, capped at the lowest of:
- $560,000 (the absolute ceiling)
- your net loss for the year Ă— 28%
- your total R&D expenditure Ă— 28%
- 1.5 Ă— your R&D labour expenditure Ă— 28%
The $560,000 cap is per Inland Revenue's current rules and IS 23/09. It is 28% of the $2 million maximum loss you can cash out, so $2M of losses and $560,000 of credit are one and the same cap, just stated two ways. Fully phased in and current for 2026, and nothing in the Budget 2026 material touches it, though it is election year.
For a labour-heavy team the binding number is usually the total R&D expenditure cap, not the 1.5 times labour cap. The 1.5 times labour cap normally impacts when a lot of the R&D spend is non-labour, such as heavy contractor use.
How to apply
A separate application in myIR, not part of your RDTI General Approval. You claim it after your income tax return is filed, and no later than 30 days after your IR4 due date. There is no discretion for late filing on this one. Miss it, lose it.
The one thing to understand before you take the cash: this is not a grant. It is an interest-free loan from the government against your future profits. You repay it out of income tax once you are profitable, which is fine, that was the deal. Or you repay it early if a recovery event happens. That is Part 6, and you should read it before you file, not after.
Worked example: pre-revenue, first full R&D year
The company for the year. Revenue nil, total costs $400,000, so the tax loss is $400,000. Of those costs, this is the R&D split:
The money | Amount |
R&D labour (salaries of the people doing the R&D) | $250,000 |
Other R&D costs (contractors at 66%, cloud compute, depreciation on dev gear) | $60,000 |
Total R&D spend | $310,000 |
Everything else (admin, sales) | $90,000 |
Total costs | $400,000 |
Can you include any of that $90,000? Not the sales and general admin itself. Selling, marketing and running the company are not R&D, so they stay out. The one thing worth checking is overheads that genuinely support the R&D, for example the R&D team's share of rent, power and IT. A reasonable, defensible share of those can move into your R&D spend. Sales and admin as such cannot. If you are not sure, the test is simple: would this cost still exist if the R&D stopped? If yes, it is not R&D.
Check the 20% wage intensity: R&D labour $250,000 plus 66% of R&D contractor cost, over total labour ($250,000 R&D + $90,000 admin and sales + 66% of all contractor cost). Comfortably over 20%.
Your credit is the lowest of these four:
The cap | How it is worked out | Amount |
Absolute ceiling | 28% of the $2 million maximum loss you can cash out | $560,000 |
28% of your net loss | 28% Ă— $400,000 | $112,000 |
28% of your total R&D spend | 28% Ă— $310,000 | $86,800 |
28% of 1.5 times your R&D labour | 1.5 Ă— $250,000 = $375,000, then Ă— 28% | $105,000 |
The lowest is $86,800, so that is what lands in the bank. For a team whose spend is mostly wages, it is the total R&D spend cap that decides it, not the labour multiple.
Why 1.5 times labour? The scheme anchors your credit to what you pay people, because R&D is mostly wages and wages are the hardest number to inflate. The 1.5 multiple lets you count a reasonable layer of non-labour cost on top, up to half as much again as your R&D wages. It is a guardrail: your cash-out cannot run far ahead of your actual R&D payroll.
Should you apply for the RDLTC at all?
Usually yes. Cash now beats a loss carried forward to a profit year that may never arrive on the timeline you hope. I would think twice in two situations. First, if you are close to a sale inside two or three years, because the clawback (see Part 6) can pull a chunk back out of the deal at the worst moment. Second, if your loss is small and your R&D labour is modest, where the cash is minor and the admin plus the future repayment obligation may not be worth it. For a genuinely loss-making team building for the long haul in New Zealand, take it.
Part 6: The Exit Landmine
Read this before you cash out losses, not during your exit.
The balance of the R&D Loss Tax Credit is repaid early, in full or in part, if a "loss recovery event" happens.
The triggers:
- You sell more than 90% of the company's shares. Selling up to 90% does not trigger it. That threshold is exact and it matters: a partial exit, a big new round, founders taking money off the table, none of that trips the wire at 90% or below. A full acquisition does. The amount is the lesser of the value of the shares sold times 28% or your outstanding balance. In a near-full acquisition that usually means the whole outstanding balance comes back, which is why you model it into the price, not as a surprise in due diligence. Example: you cashed out $200k of RDLTC and sell 100% of shares for $3m. Clawback is the lesser of $3m x 28% = $840k or your $200k balance, so the full $200k repays at your terminal tax date.
- You sell the R&D intellectual property itself. Asset sales and IP-only deals catch people here. There are specific triggers, the lesser of 28% of the sale price, or your outstanding balance. It is worth talking to an R&D Expert, as judgements are needed if the sale is taxable and at market value, which impacts treatment and if a clawback is triggered.
- A liquidator is appointed. Full outstanding balance.
- The company migrates offshore or stops being a company. Full outstanding balance.
- The company is removed from the register without a liquidator. Winding a company down by letting it be struck off under section 318 of the Companies Act, instead of liquidating it, does not make the credit disappear. A company that is off the register is no longer a company, which is a recovery event on its own, so the full outstanding balance falls due. Inland Revenue is a creditor for that amount and can object to the removal until it is paid, so a quiet strike-off to save the cost of a liquidation crystallises the whole balance rather than avoiding it.
Two more things nobody tells you. Until the credit is repaid, you cannot attach imputation credits to dividends, which mostly does not matter pre-profit but is worth knowing. And repayment is due at your terminal tax date after the event, so the timing is unforgiving.
The practical rule: the RDLTC is brilliant for a startup planning to build in NZ for years. It needs modelling for a startup that might be acquired inside two or three. Not a reason to skip it, a reason to take it with your eyes open. If your company is winding up, being liquidated, or being struck off under section 318, make sure you check with an expert.
Part 7: What Changed and What's Coming
Live now: the in-year payments loan via TMNZ, an interest-free government loan of up to 80% of your expected credit, up to three draws per income year, available once approval is in place.
Proposed, not law: direct quarterly in-year payment of the credit (limited to labour-related taxes paid), Inland Revenue discretion to accept late returns and fix minor errors (from 1 April 2027), a mining eligibility expansion, and the internal software cap cut from $25M to $3M. Most are proposed to apply from the 2027-28 income year and are expected to be introduced in a tax bill later in 2026. None of this is law yet.
The software cap cut does not touch startups building software for sale. External software sits outside that cap entirely.
Part 8: Records That Survive a Review
Claims do not fail on wording. They fail on evidence.
Everything in Part 4 about writing a good application is true and none of it saves you if the records underneath are thin. IRD can review a claim, ask questions before paying, and audit for up to four years after your claim is processed. What makes an audit easy is boring but critical: notes made at the time. IRD can review for up to four years, but you must keep your tax records for seven, so don't throw them out.
The records that matter, in order
- Time. Who spent how long on the R&D, recorded as you go. The percentage you claimed for each person has to trace to something: timesheets, sprint logs, a weekly note (Date. What we did not know this week. What we tried. What the result told us). A number invented in June for the year that ended in March is exactly what gets questioned.
- The uncertainty, dated. A short note from when you started: what you did not know, why it was not knowable, what you planned to try. Two paragraphs written then beats ten written later.
- The tests and the results. What you tried, what you measured, what failed. Failed R&D is still R&D, but only if you wrote it down.
- The cost links. Every dollar in the claim traces to the activity. Contractor invoices that say what the work was, not just "development services."
What triggers questions: eligible spend that is a suspiciously round share of total spend. A "Marketing Manager" claimed at 80% R&D with no explanation of what they actually did. Grant-funded costs left in the claim. Numbers that moved between the General Approval and the Supplementary Return with no story. Silence when IRD asks for more.
If IRD comes back with questions: answer fast and completely. The 10-week payment target only runs from a complete return, and it stops every time they wait on you.
What happens if you get it wrong: If IRD later declines a claim you have already been paid on, for example through an audit, you repay it, and a careless or unsupported claim can attract use-of-money interest and shortfall penalties. That is why records and the free MBIE check matter.
Where founders actually get stuck
Is it really novel?
Founders see free money, and in software especially, a lot of the work is using the same tools everyone has to get a different outcome. That is not R&D. The test is simple: you should be able to explain what was genuinely unknown to a ten year old, and show how you planned for the risk and worked through it. If you cannot, it probably is not R&D.
R&D versus commercialisation.
This is where founders get most confused. A customer asks for a change so the product performs better for them, so you build that one thing for that customer. The question IRD will ask is whether that is really R&D or just commercial work with a known outcome. If there was no genuine technical unknown, it is commercialisation, not R&D.
Loading in every cost and hoping.
The admin person may well have touched the project, but think about proof. You want the highest level of proof so the claim is accepted quickly. Keep everything simple, make every number trace back to your submission pack, and run each cost through one check: would this survive if IRD asked me to justify it?
Real world examples of questions from the IRD
Why is your sales person included in your R&D submission?
Very common when we take on a client who has DIY'd it. Usually the intent and the calculation are fine. The problem is the bookkeeping. They run Xero, then decide to chase a different revenue story, and someone goes back through Xero and re-tags a heap of things to get cleaner year on year reporting for investors. Suddenly the numbers behind the claim have moved. The lesson: lock down your submission, your key rationales, your tests and your significant spends at the time. If you do not want to download or tag every invoice, screenshot the list so it is easy to work through if IRD reviews.
How do you get to this number?
Common, and the core issue is nearly always how you wrote it. Either you assumed the reviewer already knows your field and wrote too little, or you assumed they know nothing and wrote a small novel about your expertise. Remember the person reading it is human. Keep it concise. Say what is novel, how you know it is novel, what searches you did, and which specific part is novel. Make everything else minimal and go heavy on the novelty. If they need to research your field to follow it they will, but give them the breadcrumbs to know where to look.
Your minimum viable record-keeping system
You do not need a fancy system. Word and Excel are enough. Timesheets are ideal but not always practical, so at a minimum keep the invoices to one side when you file the submission, so if IRD reviews you, you answer in an afternoon instead of a panic. The best hack is to make the records a byproduct of work you already do. If you run sprint reviews, tie the R&D notes into your OKRs or 90 day plans. Capture how and when you report progress to the wider team, in a specific Slack channel or your all hands, as a few bullet points on a regular cadence. That rhythm is what proves to IRD the R&D was real and happening at the time, not something written up for compliance after the fact.
Writing your first government application?
The WFW Expert Edition NZ Government Support: A Step-by-Step Execution Guide includes practical advice on writing strong government funding applications—from explaining technical innovation clearly to avoiding common mistakes that slow approvals. The same principles apply when writing a strong RDTI claim.
Part 9: The Honest Maths, DIY vs Paying Someone
Before you decide whether to pay anyone, do the maths on what you could actually get back.
The maths at founder scale. $80k of eligible spend is a $12k credit. Success-fee firms charge a percentage of the credit, as much as 20%. The DIY cost is founder hours, mostly on the first claim.
What the fee actually buys, unbundled: eligibility scoping, drafting the General Approval, compiling the expenditure, filing, and handling Inland Revenue's questions. For a straightforward single-project claim, this edition plus the free MBIE review replaces most of that. Some of it, on a complex claim, genuinely earns a fee.
When paying is the right call, honestly: large or multi-project claims, messy records, complex ownership, loss-makers who want the RDLTC optimised and the exit modelled, when your time is better spent in the business, or an Inland Revenue review already underway. Outside those, do not pay anyone yet. A fixed quote normally beats a percentage.
The free layer, first. MBIE Customer Engagement Specialists check your eligibility and read your draft at no cost. Use it before you spend anything.
So the order that keeps the money in the company is simple. Work out how much you could get back. If the claim is simple, Inland Revenue and the Government want you to do it yourself, so use the free MBIE help if you need it. Reduce the fees you would otherwise pay by being able to explain plainly what makes your R&D novel and being clear on the costs involved. Reuse the questions others have already been asked by Inland Revenue, because they want clear information and the ability to sign you off quickly, so do not make it hard. And once you are through, share the pathway with the next founder. If your situation is complex, or you want a safety blanket, talk to an R&D Expert. Otherwise, try free first.
Summary: The Whole Claim at a Glance
A founder who reads nothing else lands here and knows where they are and what to do. Do it yourself if: single project, clean records, you own the IP, not near an exit. Pay if: multi-project, messy records, group or overseas, a loss-maker optimising, or IRD already asking.
Stage | You are | What to do | Deadline | Free help |
Understand | New to all of it | Read the money map, pick your path | n/a | n/a |
Check | Unsure you qualify | Run the self-check: 3 gates, 4 tests | n/a | MBIE eligibility chat |
Claim | Ready to apply | Draft GA with the template, total eligible spend | Last day of 3rd month after balance date (30 Jun for 31 Mar) | Free MBIE draft review |
File | GA approved | Supplementary Return in myIR | 30 days after tax return due date | n/a |
Cash (loss-makers) | In loss, wage test passed | Separate RDLTC application with your IR4 | 30 days after IR4 due date, no late discretion | n/a |
Protect | Cash received | Records weekly, model the exit clawback before any sale | Ongoing | n/a |
Playdough is a Chartered Accountancy and Fractional CFO practice that helps founders start, scale, and grow their businesses. Their team combines practical finance, tax, compliance, and strategic commercial support, with a particular focus on companies where R&D, fundraising, and cash flow are critical to growth. If your claim is straightforward, Cormac's advice is to try it yourself first. If it becomes more complex, Playdough is the kind of team that can help.
Quick Navigation:
Back to /w=1920,quality=90,fit=scale-down){kind=logo}
All facts verified July 2026 against rdti.govt.nz, ird.govt.nz, mbie.govt.nz, the Beehive Budget 2026 release, IS 23/09, and the Motu five-year evaluation. Budget 2026 items are proposals until enacted. General information, not tax advice.