How to run a salary benchmarking exercise for the first time

"I want to know how competitively we pay in the market for all our roles”.

The question that all People teams dread. Whether you’re a seasoned VP of People/Talent or People Ops coordinator fresh into the world of startups, a company wide salary benchmarking is not something anyone looks forward to, however it doesn’t have to be this way.

The first time I carried out a salary benchmarking exercise was back in 2014 at Qubit and I remember it fondly. It consisted of me trawling through LinkedIn, Glassdoor, Payscal and huge 100 page PDFs from recruiters like Hays.

I had a blank spreadsheet open and was painstakingly trying to find the junior, mid, senior and exec salaries for all the roles we had at the time from the various sources. It was not fun. It took ages and the data from each of the sources differed widely. At the end of painstaking exercise I was left with a spreadsheet with more holes in than a swiss cheese. It didn’t really tell me much about whether we were paying ‘competitively’ or not.

That’s when someone said ‘why not throw some money at the problem and buy a data set from one of the big salary benchmarking companies like Radford, Willis Towers Watson, Mercer?’ I’d never even heard of these but given that’s how traditionally a lot of the larger companies carry out salary benchmarking then surely this will give us the information we need? Well, I’m here to tell you that there is no silver bullet in salary benchmarking.

I’ve carried out 40+ salary benchmarking projects in the last 2 years and want to share some of our learnings and answer the question:

How do you carry out a salary benchmarking exercise for the first time?

I’ll get the bad news out the way first. You will need to invest some time into building out a compensation model. There is no quick fix, but spending time on it now and getting it right, will pay dividends and result in something that you only need to revisit 1–2 times a year, not once or twice a week. So what do you need to do?

Step 1: Assign everyone in the business a role and level

In order to map employees to the right benchmarks in an objective manner, they need to be organised and assigned the following pieces of information:

1. Job Family - this is the function a role is a part of, often discernible by the department or team this role falls in at your company.
2. Track - Are they managing people or processes?
3. Level - this is how senior / experienced the person is and the scope of responsibility that their role covers.
4. Sub-level - how established in their role are they
5. Location - where in the world are they performing their work - country or city

Why should I go through this process?

• 💰 Accurately mapping the levels and job families within your organisation is how we’re able to match your roles back to industry salary benchmarking data.
• 🍎 Getting this right is the difference between comparing apples to apples vs. apples to oranges, so is very important.
• 🔑 It’s worth it because it’s the foundation supports many strategic initiatives that go beyond just salary benchmarking (i.e. performance management, personal development and aligning of job titles, hiring etc)

Step 2: Find some salary benchmarks

Now the fun part. How to turn all those £0 into something that is equitable, sustainable for your business and something that people believe in. Here are some questions you need to think about:

• Where should we get the base salary data from?
• The data I’ve pulled looks wonky or doesn’t even exist for some roles. Help!?
• There is no job category for the role I’m benchmarking against. What should I do?
• Who should we benchmark against?
• What is the competitive rate? What percentile should we pay in? Should we even be talking about percentiles with our employees/candidates?
• How do we handle different geographical locations?

I’m not going to tackle all of these questions now but have picked a few that come up every time I carry these projects out.

Where should we get the base salary data from?

If this is your first go at salary benchmarking then I recommend easing yourself in and using Option Impact. It has over 3,000 participants and is the largest database of private, venture-backed companies. It’s also FREE (as long as you upload your data once every 6 months).

You could also go all in and pay for something like Radford, Willis Tower Watson. From our experience, there isn’t much difference between them. Getting set up and onboarded with them can be quite a time consuming process compared to Option Impact so bear that in mind.

Challenges with the data

Having worked with salary data from various sources, time and time again I hit the same two challenges:

1. Not enough data for a role

• This becomes a problem when you apply too many filters. The more granular you want to get = the less data becomes available i.e. drilling down to look at data for companies that have only raised Series A or just looking at London level data.
• You may apply some generous filters but still the benchmarks come back empty.

2. Pay not progressing

• The data is telling you that for an IC4 the benchmark is £69,000 but an IC5 = £58,300. Eh, that doesn’t make sense?!
• I’m not going into reasons as to why this happens, however Radford has written an article on this — The Art & Science of Benchmarking: What Happens When Pay Doesn’t Progress?

So what options do you have to solve these challenges?

• Apply generous filters — While it would be great if you could compare yourself to companies of a similar funding round for each role, explore using fewer filters and see how that helps. In the example above we can see that looking at the whole market i.e. Seed to $500m, results in a more comprehensive data set.
• Use linear regression — you can use linear regression on the data to calculate what a change in salary should look like between two points when the data doesn’t progress or if you have no data for a particular level.
• Look at different percentiles — Have a look across the different percentiles to see if data exists. The above example shows only the 50th percentile.

Take your time to figure out which is the best option for each team. I work with clients on a team by team basis, looking at different filters and sources to arrive at a final recommended salary for each level.

Make sure you clearly document what approach you took to arrive at the final benchmark so that when you come to refresh the data in 6 /12 months time, you know what methodology to use for each role. If you don’t do this, you’ll be in a world of pain. Trust me, I’ve been there and it’s not pretty.

What is the competitive rate? What percentile should we pay in? 

• In order to remain competitive and attract and retain talent in a given market, I recommend you start with pulling data in the 50th percentile i.e. the median or mid point of the market. You’ll want to pull this for the country or city where the majority of you employees are based — ‘base market rate’ — for most of our clients this is usually data for the whole of the UK (we can adjust for different locations, both cities and countries in a bit).
• That does not mean that you then have to pay in the 50th percentile! You may want to pay in lower/higher percentiles for different roles depending on how strategic you want to be in competing with other companies on talent.

How do we handle different geographical locations?

• In Option Impact, you can filter data based on different countries (at present there is no option to get to city level).
• I recommend you pull data from the different countries you operate in and then interrogate it. Are there enough data points? Does it make sense? If not then you may want to look at what the location differential is between these patchy markets and markets where you have robust data. By looking at these differentials, you can create a list of 'location indexes' for the markets where the data is light and then pivot off of the strong markets. It'll look something like this:

• You can create location indexes for cities and countries by gathering data from local markets, including recruitment reports, cost of labor tools, candidate feedback, and cost of living and rent indexes to build up a picture of the what the differences are between a location and the baseline market.
• If you opted for something like Radford you can drill down to city level. Again, I recommend pulling data for the different cities you operate in and then taking a deep look at this. Are there notable differences between cities, especially those in the same country? More importantly does it make sense to pay differently between two cities in the same country i.e. would you pay two people differently just because one lives in Berlin and the other in Munich? There are no right or wrong answers here!

Due to the recent pandemic, and the move to remote working practices for many, there are questions around the traditional cost of living adjustments that companies have taken in the past with global compensation models. How should you remunerate employees based on their location? What do you do when someone chooses to move from London to a remote part of Portugal or vice versa? How do personal life choices fit into this? Should you even adjust compensation based on where someone lives?

What about pay bands?

The final step you need to take is to create pay bands i.e. your min, mid and max points for a level, and then adjust for things like the pay range spread and whether you want overlapping bands or not. Here is an example of what this could like below (screenshot from my compensation calculator).

Step 3: Benchmarking existing employees

Now you have your data, you need to benchmark existing employees against it and see where you are under/over paying people.

I've built a handy spreadsheet that allows you to benchmark employees at scale against different percentiles for each team and then understand where adjustments need to be made and what that will cost the business. Budgets can be built and the true cost of raises realised + accurate costs of future hires can also be calculated in the 'Open roles' sheet.

That being said, whether you need to make adjustments at all is the big question. A thorough analysis of the data and percentiles will inform you as to where your salaries sit in comparison to your preferred position and the levels paid by your comparator groups, however that does not mean that’s what you have to pay. Do compensation surveys actually tell you how much you have to pay? The answer is no. To quote Radford:

“Market data can tell you what the salary for a junior software engineer is in Austin, Texas, but that doesn’t mean that’s what you have to pay.”

As an organisation, your job is to design a rewards package that resonates with your employees and is competitive in attracting and retaining the type of talent your team requires. Yes, base salary is a large part of this, but it shouldn’t form all of it.

So what is the key takeaway from this? Fundamentally, salary benchmarking is neither an art, or a science, it’s both. You can be data driven but the data is only as good as you know what to do with it.

If you’d like help with your next salary benchmarking exercise and learn about how I’ve helped companies like TrueLayer, Bud, Seedrs, Lingokids, CharlieHR, Second Nature and many more solve this gnarly topic then reach out directly to me on alistair@justly.company or you can learn more on my website here.