R&D spend as a percentage of revenue: software benchmarks

R&D spend as a percentage of revenue is R&D expense divided by revenue for the same period. For software companies the sourced benchmark figures run from about 16% to 27%, and individual public companies range from 14% to 50%. No single number is "the" benchmark, because the samples, denominators and years differ. Most benchmark pages stop at the ratio; the useful step is turning it into a headcount.

What does R&D spend as a percentage of revenue measure?

It measures how much of each revenue dollar goes into building product. The numerator is everything the company books as research and development. The denominator is revenue, usually for the fiscal year.

What sits in the numerator varies. Salaries, benefits and tools are the core. Hosting costs and, outside GAAP reporting, stock-based compensation are treated differently from company to company, so two firms with the same engineering team can report different ratios. Public companies report R&D expense under GAAP, which covers the whole R&D function, not only engineers. Private-company surveys rely on what founders and finance leads choose to report, and those figures are usually unaudited.

What do software companies actually spend on R&D?

Here are the benchmarks we'd trust, each with its sample and date. Where a source doesn't state something, the table says so.

The Aleph figures come from Aleph's write-up of the 2026 SaaS and AI Performance Benchmarks, a secondary host for the Aleph x Benchmarkit report published 1 July 2026. We haven't read the report itself. The same page shows growth mattering: companies growing over 50% a year have a median of 33%, against 24% for those growing 11-20%. It also says the median share of R&D going to AI is 10%, from only 46 companies, so treat that one as directional.

The public-market numbers come from Damodaran's R&D-by-sector dataset at NYU Stern. ICONIQ's 2024 engineering report gives the plateau.

If you want one answer to "what percentage of revenue should be spent on R&D", it's a range: ICONIQ puts the plateau at 20-30%, and Aleph's private B2B SaaS median is 27%. Under $20M ARR, Aleph's medians are higher still, at 32%.

How do public software companies compare?

We computed these from each company's 10-K: R&D expense divided by revenue. Fiscal years end on different dates, so read the year column.

The spread is the point. Salesforce spends about a seventh of revenue on R&D and Atlassian spends about half. These ratios include stock-based compensation and every person in R&D, so they run higher than a cash-only engineering budget would. They also reflect strategy: a company with a large sales organisation or a mature install base will show a lower ratio than one still building its core product. Nobody at Salesforce is wrong and nobody at Atlassian is overspending. A ratio on its own doesn't tell you which.

Why do the benchmarks disagree?

Five things move the number.

Sample comes first. Damodaran covers US-listed firms, and Aleph covers B2B SaaS and AI-native software companies in its benchmark. Public and private samples aren't the same population.

Denominator is next. Revenue, ARR and a plan figure give different ratios for a company growing fast, and many pages don't say which they use.

Then the year. Aleph's median fell from 35% to 27% in a single year. That's an eight-point swing between the two years Aleph reports, bigger than the gap between several of the sources above.

Stock-based compensation and hosting are included in some benchmarks, excluded in others, and often not stated. Finally, mix: Aleph reports 10% of R&D going to AI at the median, and as that share grows the ratio will carry costs earlier years didn't.

When two pages quote 27% and 16%, check these five before deciding one is wrong.

What does a ratio buy you in engineers?

A ratio is only useful once it becomes a headcount. The method has four steps, and the third is an approximation you should label as one.

1. R&D budget = revenue × R&D ratio.
2. People budget = R&D budget × personnel share. ICONIQ says personnel is "usually ~70-80%" of spend.
3. Engineering budget = people budget × engineering share. ICONIQ says engineering is about 80-90% of R&D headcount. That's a headcount share, not a cost share, and engineers may cost more or less than the product managers and designers beside them. Use it to size the team, not to promise a number.
4. Engineers = engineering budget ÷ fully loaded annual cost per engineer.

Here's the arithmetic with placeholder inputs. Revenue of €10M is an assumption to replace with yours, and 27% is the Aleph median.

• R&D budget: €10M × 27% = €2.7M.
• People budget: €2.7M × 70-80% = €1.89-2.16M.
• Engineering share: €1.89M × 80% = €1.51M, and €2.16M × 90% = €1.94M.

To divide by a cost per engineer, you need that cost. As a floor, take Eurostat's 2025 information and communication hourly labour cost (covered below) for 1,920 hours a year (160 hours × 12, an assumption). In Poland that's 1,920 × €29.3 = €56,256, so €1.51-1.94M buys about 27-35 engineers. In Germany it's 1,920 × €58.5 = €112,320, which buys about 13-17.

Those are the same budget producing roughly double the team, and they're overstated in both cases, because the Eurostat figure is an average for all ICT employees, not senior engineers, and leaves out the layers in the next section. The method is the useful part. For the same exercise built bottom-up from a team of 12, see a 12-engineer Series A budget built bottom-up.

How do you calculate the fully loaded cost of an engineer?

Add layers, and label each one as a sourced figure or your own input.

Layer 1 is wage plus employer non-wage costs. Eurostat's lc_lci_lev dataset gives total hourly labour cost, including employer social contributions, for information and communication (NACE section J), used here as a proxy for ICT. It covers all ICT employees, not senior engineers, so it's a base layer, not a salary guide. The 2025 values:

The non-wage share swings from 4.1% in Romania to 34.6% in Sweden, so the same gross salary costs a different amount to employ depending on the country. Our guide to European software team costs has the fuller country table.

Layer 2 is the costs Eurostat doesn't count: recruiting, equipment, tools and licences, the share of an engineering manager's time, and the weeks a seat sits vacant or a new hire ramps up. No source here gives a number for these, so take them from your own invoices and recruiting history. Employer-side costs beyond wages walks through contributions, recruiting, equipment, ramp and severance for one developer.

Add the layers and you have your cost per engineer. We'd skip the "multiply base salary by 1.25-2x" rules of thumb you'll see on vendor blogs. None of them cites a source, and your own layers are checkable.

What should you check before you defend the ratio to a CFO?

A CFO will probe the weak point in your comparison, so find it first.

1. Which denominator does the benchmark use, and does it match yours?
2. Which peer set is it: public companies, private survey respondents, or your stage and growth rate?
3. What's in R&D at your company and at theirs, including stock-based compensation and hosting?
4. Is the argument about the trend or the absolute? Aleph's median dropped eight points in a year, so "we're at the median" needs a year attached.
5. What headcount mix does the ratio imply? A ratio that funds 30 engineers on paper and 18 in practice is a conversation about management layers, not about the benchmark.

Bring the arithmetic from the previous two sections. A ratio with the headcount worked out is a plan. A ratio alone is a talking point.

What this means for your engineering budget

Pick a range, not a point. At 20-30% of revenue ICONIQ and Aleph agree on the neighbourhood, with Damodaran's public-company averages lower at 16-20%; the 14-50% public spread says your own strategy matters more than the median. Set the ratio from your growth plan, then convert it to headcount with your own loaded cost, as above.

Two decisions follow. First, the team mix: if engineering is 80-90% of R&D headcount in ICONIQ's data, check whether your org chart is heavier on management, QA or product than that. Second, where the seats sit: the same budget buys about twice the engineers at Poland's ICT labour cost as at Germany's, a gap you can verify with the Eurostat figures above. That's arithmetic on sector averages, not a claim about what any senior hire costs.

Once the number is set, you still have to win the hire. A template for taking a hire to your CEO and CFO is the next step. For the wider set of decisions at this stage, see Startup CTO guides: leadership, diligence and the first hires.

Hiring senior engineers through HighCircl

HighCircl places senior engineers from seven European countries (Poland, Hungary, Slovakia, Serbia, Slovenia, Romania and Spain) at €45-105/hr ($50-115/hr). The margin is 20%, capped and applied on top of what the engineer earns. About 1 in 10 applicants pass vetting, and a shortlist of three to five candidates arrives in 72 hours. There's no subscription, no recruitment fee and no minimum hours. If your ratio says you need senior capacity fast, see senior engineers vetted by HighCircl.

FAQ

What percentage of revenue should a software company spend on R&D?

There's no single correct figure. Aleph x Benchmarkit's calendar 2025 median for B2B SaaS was 27%, Damodaran's January 2026 average for system and application software was 15.98%, and ICONIQ puts the plateau at 20-30% for growth-stage companies at scale. Pick the benchmark that matches your size and growth rate, and say which one you used.

Is R&D the same as the engineering budget?

No. R&D includes engineers, product managers, designers and tooling, and in public filings it includes stock-based compensation. ICONIQ puts engineering at about 80-90% of R&D headcount, so the engineering budget is a subset of R&D, though that's a headcount share, not a cost share.

What is the rule of 40?

It's a SaaS rule of thumb that revenue growth rate plus profit margin should reach 40%. The profit measure varies by source (EBITDA, operating income or free cash flow), so check how a benchmark defines it before comparing. Aleph's report puts the median for calendar 2025 at 25%, well under that line. R&D spend is one of the levers companies pull to move it.

Does the ratio include contractors?

It depends on how your finance team books them, and benchmarks rarely say. Contractors who work on the product are normally R&D expense in company accounts, but check with finance before comparing your number against a published ratio. If the benchmark excludes them, your ratio will look higher than it is.

Sources

• Aleph: median R&D percent of revenue for SaaS, 2026 benchmarks
• Damodaran, NYU Stern: R&D spending by sector
• ICONIQ: Engineering Series 2024, building engineering teams
• Eurostat: labour cost levels by NACE activity
• Salesforce: FY2026 Form 10-K, SEC EDGAR
• Atlassian: FY2026 Form 10-K, SEC EDGAR