There are all kinds of reasons to generate benchmark data. It might be for a specific job that you are struggling to hire for, it may be for a new region or country where there are no structures in place, or for a hiring manager that wants to look more closely at a particular position.
Whatever the reason, on the face of it, benchmarking a role is relatively straightforward but the reality is that it can be a daunting task, particularly if you are unfamiliar with the data, and many different factors will impact the final recommendation.
For the purposes of discussion, I am going to assume that we have settled on the job match, level, and data source and talk about some approaches to triangulate benchmark data in situations where there is limited or questionable data availability. What if you are asked to provide data for a role in a specific region, or for a level that doesn’t exist in the survey, or for a peer group that doesn’t yield sufficient data?
First of all, I want to reinforce that a benchmark is not just survey data. Market data will inform the final benchmark but requires more thought than just copying and pasting market data.
So let’s look at a few techniques that can be used where market data is of questionable reliability or non-existent:
In certain locations, you may be aware that a premium or discount should be applied however there is insufficient data to create regional cut of survey data.
Regional differentials are a reliable way to take national data and slide it up or down to arrive at the appropriate level for a given region. “Geo-differentials” with respect to pay can be obtained from a variety of sources and most survey providers provide geo-differentials as part of subscriptions, derived from the same source as the compensation data.
To do this, start with national data for the role in question, and simply multiply base salary data according to the appropriate geo-differential.
A couple of caveats with this approach – in practice, geo-differentials apply differently to roles depending on seniority. For example, executives are typically benchmarked nationally (or even internationally) in accordance with the talent pool, however pay mid-level individual contributors, and support-level staff will vary by region. The second caveat is to be aware of the distinction between cost-of-living differences and pay differences. While the two will move in similar directions, cost-of-living differentials (driven by rent etc.) will be far more pronounced than pay differentials (driven by company policies).
Peer group differentials
What about when you are asked to produce a benchmark for that brilliant 15-company-strong peer group that leadership has signed-off on but turns out produces almost no data in key locations?
This is where peer group premiums can be used.
Say for example, you have sufficient data in a given country to generate some roles but not all roles. What you can do is compare the data for the roles that can be reported to a broader cut of data to calculate the average premium/discount of the peer group data.
Once you have calculated the peer group differential, all you need to do is take available market data from the broader cut and multiply it through by the calculated premium to arrive at an appropriate benchmark. You may find that taking this approach produces much smoother, more usable data, in general than by looking at the peer group data in the first place.
This is where things can get a little more complicated. Let’s say you have gotten as close as you can to a benchmark but you still have limited confidence or no data at all – now what?
Now we have to start looking for patterns or trends in the data that can be applied to the data that we do have, in order to arrive at a benchmark. One would usually do this in multiple ways before deciding on the benchmark range.
For example, we might be missing data for Role A in Portugal but we do have data for a Role B, which is similar. In this case, we could look at the differential between Roles A and B in Spain, where we do have data, and apply that differential to extrapolate Role A in Portugal.
Another example is that you have data for levels 1-3 for a given position but you need level 4. Here you might take the average progression between levels 1 to 2, and 2 to 3, and apply that to arrive at level 4.
So there you have it – This certainly isn’t exhaustive and it takes a bit of time to gain confidence using these techniques. Hopefully this takes some of the pain out of your next problem benchmark.