ooligo

Aon Radford

compensation-benchmarking salary-surveys · salary-benchmarking · executive-compensation · market-practice-studies
Recruiting & TA
7.0 /10

What it is

Aon Radford is the compensation survey that technology and life-sciences companies have priced jobs against for more than 50 years, by Aon’s own count. Today it is sold as the Radford McLagan Compensation Database, which folds Aon’s Radford surveys (tech, life sciences) and McLagan surveys (financial services) into one dataset run by Aon plc’s Human Capital Analytics business. In its March 2026 release, Aon described the database as relied on by more than 8,500 organizations and covering over 30 million employees across 115 countries and 150 job functions.

The difference from the live-data tools is the method. Radford is a participant survey: each company submits a full employee-level data file, Aon’s analysts validate it through several rounds of quality checks, and the results are published as benchmarks for matched jobs and levels. Aon calls this “non-crowdsourced” data, and that is the selling point. Alongside the core database Aon sells market practice studies on salary increase budgets and turnover, incentive plan design, sales compensation design, severance and early-career pay.

The question it answers: what is the defensible market rate for this job, in a form a compensation committee, an auditor or a proxy adviser will accept?

Why it shows up in recruiting stacks

Most recruiters never log into Radford. They meet it as the source behind the salary bands the compensation team hands them, and as the reason an offer cannot go above the band’s maximum without sign-off. For a pre-IPO or public tech company, the band structure and the executive peer group usually trace back to Radford cuts.

The March 2026 release narrowed the gap with live tools. Aon added API connections to client HRIS and ATS systems for submission, a real-time validation dashboard that flags inconsistencies before the file is accepted, a job-matching agent that takes natural-language instructions, an AI Compensation Assistant for querying pay data, and real-time labor-market signals shown next to the survey data. It also added AI job families such as head of AI, machine learning engineer and AI ethics. The survey itself still moves on a publication calendar, not continuously. Aon’s 2024 transition FAQ said the combined database would follow Radford’s quarterly publication schedule.

Pricing reality

Aon publishes no prices. Its own wording is that fees to participate in and access the database vary with your benchmarking goals. Access is give-to-get: the way in is a complete, timely data submission that passes Aon’s checks. For several of the market practice studies, participants who submit receive the overall results at no charge, and non-participants can buy them after publication.

The best outside number is Vendr’s buyer data for Aon, which shows a typical range of about $4,900 to $21,100 a year and an average contract of about $22,000 across 18 deals. Those figures cover all Aon products bought through Vendr, not Radford alone, so treat them as a rough guide, not a quote. Cost climbs with each extra country, industry cut and custom peer group. The bigger cost is internal: someone on the compensation team has to match your jobs to Radford’s job architecture and clean the submission every cycle, and that analyst time recurs every year.

Best for

Compensation and total rewards leaders at US-headquartered tech, life-sciences and financial-services companies with roughly 500 or more employees, especially pre-IPO and public ones. The job is setting the company-wide band structure and the executive and board pay that a compensation committee approves. Recruiting teams benefit downstream, through bands that hold up when an offer is challenged.

Do not buy it if you are a startup under about 200 people pricing offers week to week, or if no one on staff owns job matching. You will pay for a dataset you cannot keep current.

Versus the alternatives

Mercer is the other large survey house. Pick it when you are a non-tech enterprise or your peers and auditors already benchmark against Mercer’s surveys. Since August 2026, Mercer insights also reach Compa customers through Compa’s Analyst Agent.

Pave is the fastest-growing real-time alternative, with 9,000+ companies on its platform and a free Market Data Lite tier. Pick Pave when you need fresh numbers for offers, equity benchmarks for a venture-backed company, or merit-cycle planning in the same tool. The two are not exclusive: you can run Pave for offer pricing and keep Radford for the band structure and the compensation committee.

Ravio is the live-data pick for European headcount. Choose Ravio when most of your hiring is in the EU and you need pay-transparency reporting. Compa reads your own offer and ATS data, which helps when the question is what competitors paid in recent offers, not what the survey median was.

Watch-outs

  • Survey data is old by the time you price an offer. Submissions are snapshots, and results publish on a calendar, so a hot role can move faster than the cut. Guard: age the data forward to the offer date with the salary-increase budget figures from Aon’s own study, and check offer acceptance rate by job family. If one family keeps declining, add a live source for it.
  • Bad job matches produce confident wrong numbers. A Radford benchmark is only as good as the match between your role and its job code and level. Guard: have the compensation analyst document every match with the job description used, re-review matches for roles that changed scope, and use the new job-matching agent as a first pass, not the final answer.
  • Participation is a recurring workload. Missing a submission window can cost you access or the participant terms. Guard: put the submission deadline in the compensation calendar with a named owner, and connect the HRIS through the API so the file is generated, not rebuilt by hand.
  • Thin cells for niche roles and small countries. With 115 countries, coverage is uneven outside the core tech and finance markets. Guard: read the incumbent and company counts behind each cell before publishing a candidate-facing range, and fall back to a broader geography or level when the count is low.