WBSO — the Dutch R&D wage tax credit

The thing most people get wrong

Start with the fact that trips up every first conversation: WBSO is not a grant. Nobody writes you a cheque. The Belastingdienst does not send a confirmation letter with your company name in bold at the top. If that is what you are expecting, you will spend three months waiting for a payment that never comes.

What actually happens is this. You run payroll the way you always do. When your accountant calculates the wage tax (loonheffing) you would normally remit to the Belastingdienst, she subtracts the WBSO credit for that month and pays the lower amount. The money never leaves the company. The “benefit” shows up as a smaller wage-tax line on the monthly payroll cycle — which, for the finance lead, looks indistinguishable from a payroll tax holiday.

That is the whole mechanic. Once you internalise it, the rest of the scheme makes sense.

Why WBSO exists

The Wet Bevordering Speur- en Ontwikkelingswerk — the Act for the Promotion of Research and Development Work — was introduced in 1994, and in its current form has been running continuously since 1996. The political logic is straightforward. The Netherlands has a small domestic market, expensive labour, and a tax regime that competes on sophistication rather than headline rates. If you want Dutch companies to keep their R&D in the Netherlands rather than offshoring it to Poland or Portugal, you have to pay for it.

The CBS and RVO publish the annual totals. WBSO disburses roughly €1.5 billion a year through about 20,000 applicants, covering somewhere north of 90,000 R&D workers. That makes it, by population-adjusted intensity, one of the most generous R&D tax schemes in Europe — comparable to the UK’s old SME R&D scheme before it was cut, and considerably more generous than Germany’s Forschungszulage on a per-employee basis.

The political consensus around it is unusual. WBSO has survived three austerity rounds, two pandemic-era budget rewrites, and several centre-right coalition governments that would instinctively prefer headline rate cuts. Even the 2024 rate adjustments, which tightened the brackets modestly, were pitched as a rebalancing rather than a retreat. For a founder, this matters: WBSO is one of the few things in the Dutch business-tax landscape you can assume will still be there in five years.

Who actually qualifies

The legislation is clear on the two axes that matter: who you are and what you are doing.

On the who: WBSO is available to any entity that runs a Dutch payroll — which means a BV, a NV, a branch of a foreign company registered for loonheffing, or a foundation that employs people. Self-employed founders qualify through a separate track (more on that below). A holding company with no payroll cannot claim. A UK Ltd with Dutch contractors billing through zzp cannot claim on those contractor hours — WBSO is a wage-tax credit, so no wage means no credit.

On the what: the work has to be technically novel to you, technically uncertain at the outset, and systematic. Those three words do a lot of work in RVO’s published handbook, and they are the filter almost every rejection comes down to.

Novel to you — not novel to the world. You do not need to be publishing papers. But configuring a CMS, implementing a third-party SaaS, or porting an existing library to a new framework does not count. If the work could reasonably be done by a competent developer following documentation, it is not WBSO-eligible.
Uncertain — when you started, you did not know whether the approach would work. The hallmark of a good WBSO project description is that it names the technical risk up front.
Systematic — you followed a plan. You can point to hypotheses, experiments, and conclusions. Not necessarily a pre-registered protocol, but not just “we tried stuff.”

Common accept patterns: ML model development where the architecture is non-obvious; embedded firmware with real-time constraints; distributed systems work where consistency guarantees are the problem; novel algorithmic work in any domain. Common rejections: A/B testing copy on a landing page; configuring Shopify or HubSpot; redesigning a UI without a performance or algorithmic problem underneath; “modernising” a codebase by swapping frameworks.

The grey zone is large. A team building an e-commerce checkout flow does not qualify for the checkout flow itself, but might qualify for the bespoke fraud-scoring layer they had to invent because off-the-shelf tools underperformed on their category. Good advisors live in that grey zone. Bad ones claim everything and get clawed back.

The money, in concrete terms

For 2026, the rates are:

36% of R&D wage costs on the first €350,000 per calendar year.
16% on anything above that threshold.
For starters (companies under five years old with no prior WBSO) the first-bracket rate rises to 50% for the first three years of claiming.
Self-employed founders receive a fixed annual deduction of €15,551, with an additional starter bonus of €7,781 for qualifying zzp.

What does that look like in practice? Take a Dutch scale-up with five developers on €90,000 gross salaries, each spending 60% of their time on qualifying R&D. That is €270,000 of eligible wage costs. At 36%, the credit is €97,200 a year — roughly the loaded cost of one additional hire. For a ten-person engineering team on comparable numbers, the maths pushes you through the €350,000 threshold: €126,000 at 36% on the first tier, plus 16% on the excess, for a blended effective subsidy of around €140,000. A senior hire, fully funded by the credit.

The arithmetic is why WBSO is, for most Dutch companies doing real R&D, the single most valuable line item on the tax return. It compounds with Innovatiebox for profitable companies (the 9% innovation box rate uses the WBSO certification as its qualifying trail), and it stacks with ZIM or Horizon grants for specific projects, though the wage costs covered by the grant are excluded from the WBSO base to prevent double-dipping.

The hours problem

If there is one operational detail that decides whether your claim survives an audit, it is this. You must track R&D hours per person, per project, in a way RVO can reconstruct three years later.

The specific format is flexible. Some teams use their time-tracking tool of choice — Harvest, Toggl, Clockify — with tags mapped to approved WBSO projects. Others maintain a separate timesheet that developers fill weekly. A few build their own: a small internal tool that pulls git commits, tags each to a WBSO project, and lets the developer confirm the hours attributed. RVO does not prescribe software. It prescribes auditability.

What fails: reconstructing hours at year-end from memory. Allocating time by project budget rather than actual effort. Using round numbers — “40 hours a week on WBSO project A” — for a developer who also attended company meetings, helped with support, and onboarded a hire. RVO audits a meaningful fraction of claims each year, and the audit begins with the timesheet. If the timesheet is fiction, the claim is fiction.

The practical rule most Dutch finance teams settle on: set up tracking in month one, insist on weekly entry, and reconcile monthly when you close the books. The overhead is ten minutes per developer per week, and it eliminates the single largest clawback risk.

Application windows and timing discipline

WBSO runs on three fixed application windows per year, aligned to the calendar. You must apply at least one month before the period starts. You cannot claim WBSO for a month that has already begun — the door is closed.

First window: applications by end of November for the following January–April period (or for the full calendar year).
Second window: applications by end of March for May onwards.
Third window: applications by end of August for September onwards.

Most established companies apply once, for the full calendar year. First-time applicants often apply for a partial year, which is fine, but means the first calendar quarter is usually lost because the internal machinery was not in place in time to submit by the November deadline.

The calendar itself is the trap. A founder who decides in February that they want to claim WBSO for the year has already missed the first window, and is applying in March for a May start — which means January through April is permanently unrecoverable. Move the decision earlier.

What RVO actually reads

The claim you submit is a project description. For each R&D project, you describe:

The technical objective — what you are trying to build.
The technical bottleneck — the specific hurdle that makes it uncertain.
The systematic approach — how you will investigate and resolve it.
The hours estimate — by month, by project, for the period.

This document is, in effect, the entire claim. The rate, the wage data, the hours — all of that is arithmetic, verifiable from payroll. The project description is where the claim lives or dies. RVO assessors, who are typically technical specialists in relevant domains, read the description and ask whether what you have written is genuinely R&D or whether it is disguised implementation work.

Good descriptions are specific, honest about uncertainty, and written in the language of the domain. Bad descriptions hedge, use marketing language, or pile on buzzwords (“AI-powered, cloud-native, real-time”) without naming an actual technical problem. The single strongest signal of a well-prepared claim is that the author could defend every sentence in a fifteen-minute technical conversation.

When to hire a WBSO advisor

The specialist market for WBSO is large, competitive, and of varying quality. A good advisor — typically a boutique firm with technical staff rather than a Big Four tax practice — can prepare the project description in one to two meetings with your engineering lead, will correctly identify the grey-zone work that qualifies, and will quote a fee that is either flat or a capped percentage of the realised credit.

For a first application, the ROI of specialist help is almost always positive. The project description is the hardest part of the claim, and getting it wrong means either a rejection or a clawback. For subsequent years, many companies internalise the process, with the advisor reviewing the description rather than writing it.

What to avoid: firms that charge 20% of the claim on an uncapped basis; firms that file dozens of identical-sounding applications and rely on volume to pass RVO’s filters; firms that tell you every project qualifies. WBSO rewards specificity. A generic claim is a rejected claim.

How it goes wrong

Three failure modes account for almost every WBSO problem we see in conversation with Dutch finance teams:

No tracking for the first two months, because the application went in late and the timesheet tool was set up in month three. The claim gets cut proportionally, and the first months — which usually have the highest R&D intensity — are lost.
A project description that describes implementation, not development. RVO flags it, requests clarification, and the company either withdraws the project or has it rejected on re-review.
Over-claiming in year one, getting audited in year two, and owing back the difference with interest. Under the Netherlands’ normal audit cadence, expect to be audited at least once in the first four years of claiming.

All three are avoidable. All three happen routinely to founders who thought WBSO was a form-filling exercise rather than a documented production process.

What’s changing

The 2024 rate adjustments modestly compressed the scheme. The first bracket’s percentage is lower than it was in 2020 (38% then, 36% now), and the starter bonus has been reshaped. There is ongoing parliamentary discussion about further tightening — not the existence of the scheme, but its generosity — as part of broader fiscal consolidation. For planning purposes, assume rates will drift gently downward over the next three to five years rather than stepping down sharply. For any claim you are preparing now, use current rates; for a multi-year hiring plan, discount modestly.

Official sources

Indicative information only. This is not legal or tax advice. Rates, thresholds, and procedural rules change annually — always verify with RVO or a licensed advisor before filing.