Salary Benchmarking in Practice: Mercer vs Aon Radford and How to Use Market Data Properly

Salary benchmarking sounds straightforward, but in practice, it is much more nuanced.

Two reputable compensation surveys can produce different market numbers for what appears to be the same role. A company may benchmark a Software Engineer at one level in Mercer and receive a noticeably different result when the same employee is matched to Aon Radford.

That does not necessarily mean one survey is wrong.

Usually, the difference comes down to job matching, leveling methodology, participant populations, industry mix, geography, company size and the way each provider structures its compensation data.

Understanding those differences is one of the most important skills in salary benchmarking.

What salary benchmarking actually involves

Salary benchmarking is the process of comparing an organisation's compensation against relevant external labour-market data.

The objective is not simply to answer:

"What is the average salary for this job?"

A better question is:

"What does the relevant market pay for jobs of comparable scope, responsibility and seniority?"

That distinction matters.

Consider three employees who all have the title Product Manager.

One manages a relatively mature internal product.

Another owns a major customer-facing product and its commercial roadmap.

A third is effectively operating as a Group Product Manager but retains the Product Manager title.

Benchmarking all three employees against the same "Product Manager" market figure would produce misleading results.

This is why good benchmarking starts with the job, not the job title.

The practical benchmarking process

A typical benchmarking exercise looks something like this:

1. Understand the job

Review the actual responsibilities, scope, decision-making authority, technical or professional complexity and organisational impact.

2. Match the job to the survey

Identify the closest survey job rather than relying purely on title.

3. Determine the appropriate level

This is often the most important step. A correct job family with the wrong career level can produce a significantly distorted benchmark.

4. Select the relevant market

Decide which geography, industry, company-size segment or other peer group is appropriate.

5. Select the market percentile

Companies commonly reference the 50th percentile, but some target the 60th, 65th or 75th percentile for particular talent markets or roles.

6. Compare internal compensation with the market

For example:

The calculation is simple:

Market position = Employee salary ÷ Market reference

But interpreting the result requires considerably more judgement.

An employee at 90% of the market median is not automatically underpaid. They may be relatively new to the level, developing in the role or positioned appropriately within an internal salary range.

Likewise, an employee at 110% of market is not automatically overpaid. They may be highly experienced, have scarce skills or be approaching promotion.

Mercer vs Aon Radford: what is the difference?

Mercer and Aon are both major compensation-data providers, but organisations often encounter Mercer and Aon Radford in somewhat different contexts.

Mercer

Mercer has broad compensation coverage across industries, functions and countries.

It is particularly useful for organisations that need to benchmark a wide variety of roles, for example:

• Finance
• HR
• Legal
• Sales
• Operations
• Engineering
• Corporate functions
• Executive and management positions

For a large organisation with many different job families, Mercer's breadth can be a significant advantage.

Aon Radford

Radford, part of Aon, is particularly well known for benchmarking talent in technology and life-sciences markets.

It is commonly used for roles such as:

• Software engineering
• Product management
• Data and analytics
• Cybersecurity
• Technology leadership
• Sales and go-to-market roles
• Biotechnology and life-sciences positions

For a technology company competing with other technology companies for specialist talent, Radford can therefore provide a particularly relevant comparison market.

The important point is that this is not simply a question of "Which survey is better?"

The better question is:

"Which dataset provides the most relevant labour market for the jobs we are trying to benchmark?"

Why Mercer and Radford can give different answers

Imagine that a company benchmarks a Senior Software Engineer.

Mercer indicates a market median base salary of $150,000.

Radford indicates $165,000.

It is tempting to conclude that Radford "pays higher."

But that is not necessarily what the data is telling us.

There are several possible explanations.

1. Different participant populations

The companies contributing data to each survey may differ.

A broad general-industry dataset may contain employers from financial services, manufacturing, retail, professional services and technology.

A technology-focused dataset may contain a greater concentration of software companies competing aggressively for engineering talent.

The resulting market rates can therefore differ even when the job descriptions appear similar.

2. Different leveling methodologies

A "Senior Software Engineer" in one survey methodology may not represent exactly the same level of responsibility as the apparently equivalent level in another.

For example, one survey level might describe someone who:

• independently handles complex assignments;
• mentors junior employees; and
• has significant technical expertise.

Another survey's comparable level might require:

• organisation-wide technical influence;
• leadership across multiple projects; and
• substantially greater autonomy.

If the levels are not equivalent, comparing their market medians directly is misleading.

3. Different job matching

Job titles are notoriously inconsistent.

A company's "Director" could match:

• a senior individual contributor;
• a people manager;
• a manager of managers; or
• a genuine executive-level position.

The survey methodology therefore matters more than the title.

4. Different company populations

Company size, revenue, funding stage and organisational complexity can affect compensation.

A Software Engineer working for a 100-person SaaS company and one working for a global technology company may technically perform similar work, but the talent markets in which their employers compete may be different.

5. Different compensation definitions

Always check exactly what you are comparing.

Common survey measures include:

Base salary – fixed cash compensation.

Total cash compensation – typically base salary plus cash incentives.

Target total cash – base salary plus target incentive opportunity.

Total direct compensation – may incorporate longer-term incentives or equity depending on the survey definition.

Comparing Mercer's base salary figure with Radford's total cash figure will obviously produce a false conclusion.

A practical example

Suppose a technology company employs 200 people and needs to benchmark three positions:

• HR Business Partner
• Senior Software Engineer
• VP Engineering

Using a single survey for every role may be convenient, but it may not always produce the strongest benchmark.

For the HR Business Partner, Mercer might provide a robust general-market comparison because HR talent moves across many industries.

For the Senior Software Engineer, Radford might provide a more relevant technology labour market.

For the VP Engineering, the company might review both sources, particularly because executive scope, company size and organisational complexity become increasingly important at senior levels.

This leads to an important principle:

A compensation survey should be selected based on the labour market for the job, not simply because the organisation has purchased access to it.

Should you use multiple surveys?

Where resources allow, using more than one source can be valuable.

But averaging two survey numbers without understanding them is not good benchmarking.

Suppose:

• Mercer P50 = $150,000
• Radford P50 = $165,000

Simply calculating a blended market rate of $157,500 creates an apparently precise answer without explaining why the underlying data differs.

A better process is to investigate:

1. Are the job matches genuinely equivalent?
2. Are the career levels comparable?
3. Are both figures based on the same geography?
4. Are the compensation definitions identical?
5. What companies contribute to each comparison market?
6. Which dataset better represents the employers competing for this talent?

Only after answering those questions should the organisation decide whether to use one source, blend multiple sources or apply different sources to different job families.

Market data is not the same as a salary range

Another common mistake is treating a survey median as the salary employees "should" receive.

If the market median for a role is $150,000, that does not mean every competent employee should earn exactly $150,000.

Organisations normally translate market data into salary structures.

For example, a company could establish:

The midpoint may be anchored to the selected market reference.

Employees can then progress through the range based on factors such as experience, sustained performance, proficiency and time developing within the level.

This separation between market benchmark and individual pay decision is critical.

Market data helps price the job.

It does not, by itself, determine what a particular person should be paid.

Four questions to ask when Mercer and Radford disagree

When two surveys produce materially different results, resist the temptation to immediately choose the higher or lower number.

Instead, work through four questions.

First: Are we matching the same job?

Compare the actual survey descriptions.

Second: Are we matching the same level?

Look carefully at scope, complexity, autonomy, leadership and organisational impact.

Third: Are we comparing the same market?

Check geography, industry, organisation size and participant population.

Fourth: Are we comparing the same compensation measure?

Base should be compared with base, target cash with target cash, and total compensation with equivalent total compensation.

Most unexplained survey differences become much easier to understand after working through these questions.

Common salary benchmarking mistakes

Several mistakes appear repeatedly in benchmarking exercises.

Matching on title alone.

Titles are an unreliable indicator of job size.

Choosing the survey that produces the desired answer.

Benchmarking should inform compensation decisions, not justify a decision that has already been made.

Using the highest available market number.

More generous data is not necessarily more relevant data.

Combining surveys without normalising them.

Two numbers should not be averaged simply because both are available.

Ignoring survey methodology.

Career architecture and leveling definitions can materially change the result.

Treating P50 as the "correct salary."

A percentile is a market reference point, not an individual salary recommendation.

Over-segmenting the data.

Filtering by location, industry, company size and other variables can create a comparison group so narrow that the sample becomes unreliable.

Using market data without internal context.

External competitiveness matters, but so do internal equity, salary structures, performance, experience and career progression.

A simple rule for choosing between Mercer and Radford

If you are deciding which source to use, start with the talent market rather than the provider.

Ask:

Who are we genuinely competing with for this employee?

If the role competes broadly across industries, a broad market source such as Mercer may be highly appropriate.

If the role competes primarily within technology or another market strongly represented in Radford, Radford may provide a more relevant benchmark.

For some positions, both datasets can provide useful evidence.

The objective is not to prove that one provider has the "right" number.

The objective is to build the most defensible view of the external market.

Final takeaway

Good salary benchmarking is not about finding a number in a database.

It is about making a series of defensible decisions:

What is the job?What level is it?What labour market competes for it?Which survey best represents that market?Which compensation measure should we compare?How should that market information translate into our internal salary structure?

Mercer and Aon Radford are both valuable sources of compensation data. Their numbers can differ because they are looking at the market through different datasets, job architectures and participant populations.

The skill of compensation benchmarking is therefore not knowing which database to trust blindly.

It is knowing why the numbers differ and which number is most relevant for the decision you are making.