How Payroll Actually Works When You Have to Build It From Scratch
Last year I had to figure out salary bands for a team of 40 people across three time zones. The template from our old HR system was a mess of broken links and outdated tax tables. I spent a week building something that actually reflected what we were paying versus what the market was offering at the time. It turned out nobody had updated the guide since 2019, and the mid-level engineer salaries were sitting $12,000 below what a candidate would actually accept. The first thing I learned is that any document called a Staff Salary Guide 2022 needs to account for regional differences, not just title-level data. A senior developer in Austin and a senior developer in Columbus are not the same number. I found that out the hard way when two offers for the same role ended up creating internal equity issues within three weeks of hiring both people.
Using a Staff Salary Guide 2022 for Band Setting
Start with the data you already have in your payroll system. Pull the last 24 months of salary history by job family, not by individual employee. Aggregating at the percentile level hides problems that don't show up in averages. If your 50th percentile is clean but your 75th percentile is bleeding into the next band, you have a compression problem and nobody will notice until someone leaves. I use a simple spread model. Take the median salary for each level and apply a fixed range percentage around it. Most companies I talk to use plus or minus 15 percent. That works fine for entry-level roles where the variance is naturally smaller. For senior positions the range should be wider, closer to 20 to 25 percent. The higher the level, the more variable the market pricing becomes and the tighter your band will feel if you keep it narrow. Here is what most people miss. The guide needs a separate column for location cost adjustment. If you ignore geography entirely, you will either overpay in lower-cost markets or fail to hire in expensive ones. I built a multiplier table using Bureau of Labor Statistics data and applied it directly to the base numbers. It is not perfect but it is defensible when someone asks why two people with the same title are paid differently.
The real test comes when you try to use the guide for actual offers. I ran into a situation where a candidate's expectations sat at the 90th percentile of our band. The math said we could not justify it internally without creating a precedent that would drag the whole level up. I ended up moving the offer to a different title with a higher band rather than breaking the existing structure. That is a messy solution but it kept the comp system from unraveling. If you are starting fresh and need a reference point, look at Radford, Willis Towers Watson, or Mercer surveys for your industry. They are expensive but they give you a baseline that holds up under scrutiny. Free resources like Glassdoor and Payscale are okay for rough estimates but they lag behind actual offer data by six to nine months. In a fast-moving market that gap matters more than people admit. I also keep a running log of every offer I make against the guide. After about 15 offers you start seeing patterns. There was a stretch where my senior product managers consistently came in below the 25th percentile of the band. The guide was wrong, not the market. I adjusted the band upward by about 8 percent and stopped losing candidates at the offer stage within two quarters.
Get the Full Details
Build the spreadsheet. Update it annually. Treat it as a living document, not a PDF you send to the board and forget. That is the part nobody warns you about until you are explaining to a senior hire why their counter was rejected on paper but accepted in practice.
