Building a Tech Salary Guide 2023 from scratch
Most people think a salary guide is just a spreadsheet someone downloaded. It's not. If you're actually putting one together, you need to understand where the numbers come from, why they shift mid-quarter, and how to avoid the traps that make most internal guides useless within six months. I started building a Tech Salary Guide 2023 two years ago because our hiring managers kept citing outdated bands and candidates were walking away after second-round conversations when they realized the comp offered was below market. The first thing you learn is that Glassdoor and levels.fyi are reference points, not sources. They reflect self-selected data—people who got an offer and posted about it, which skews toward higher percentiles and larger metros. For baseline calibration they work. For building actual salary bands, you need something more granular. The practical approach is combining three inputs. First, survey data from Radford or similar frameworks, adjusted for your company size and region. Second, anonymized offer data from your own recruiting system if you have at least fifty hires per year. Third, direct compensation benchmark reports you purchase from firms like Pave or Compt. When you triangulate those three, the variance drops significantly. Using a single source will always leave blind spots.
The methodology that actually works
Here's what most people skip and then regret. You don't build salary bands by averaging current employee pay. You build them by mapping roles to level architectures, then anchoring each level to market percentiles. I spent months doing this backwards because our engineering team had been hired across wildly different band structures after three acquisitions. The fix was building a role ladder with explicit competencies per level, then running every existing employee through that ladder to see where they actually sat relative to market. The formula itself is straightforward. For each role-family and location, take the 25th, 50th, and 75th percentile base salary from your benchmark data. Set the midpoint at the 50th. The range spreads roughly 20 to 25 percent on either side of midpoint for individual contributor roles, wider for management because the variance in team scope creates more comp dispersion. So a 50th percentile at one hundred twenty thousand translates to a band roughly between ninety-six thousand and one hundred forty-four thousand before you layer in equity and bonus. Equity needs its own treatment. Most guides I've seen bury stock options inside total comp without distinguishing them. That's misleading. Base salary bands and equity bands should be published separately. A candidate can negotiate one without automatically triggering the other. I learned this when a senior engineer asked me to justify why her equity grant was below another engineer with similar base comp. The answer was that her previous company had a RSU-heavy package while ours was cash-heavy. Explaining that required two separate data sets, not one combined number.
A specific problem I ran into
During my second quarter of building out the guide, I hit a wall with remote roles in lower-cost metros. The benchmark data was heavily weighted toward SF, NYC, and Seattle. When I applied Seattle-based bands to a remote role in Austin, we lost three offers in a row because candidates had their own expectations based on locally posted ranges. The workaround was creating a location adjustment matrix using COLA indices from the Bureau of Labor Statistics, but only for companies that actually had hybrid or remote policies. Pure remote roles get a blended approach—I used the company's home office market rate minus a capped fifteen percent adjustment rather than fully geolocating, which kept internal equity from breaking. This doesn't solve every edge case. Contract and seasonal roles fall outside standard band structures entirely. Intern programs need their own calculation method. International remote workers require local statutory compliance data, not just market rates. I ended up pulling those roles out of the main guide and publishing them as separate supplements with different update cadences.
Get the Full Details
Common mistakes that derail salary guides
The biggest error I see is setting bands and forgetting to update them. Compensation moves fast. In tech especially, a band built in January will be stale by June if you're in a competitive hiring market. The minimum viable update cycle is quarterly for high-demand roles, annual for everything else. The second mistake is conflating total compensation with base salary. When you present figures like "one hundred eighty thousand total comp" without breaking down base, bonus, and equity, you create negotiation problems downstream. Candidates assume the total is guaranteed when parts of it are variable or vesting over four years. A counter-intuitive point that surprises people: narrower salary bands often create more problems than wider ones. A ten percent spread between minimum and midpoint sounds efficient but it gives hiring managers almost no room to negotiate offers across experience levels within the same band. I've seen leads reject candidates who were under-marketed because the candidate's expected base was five percent above the band minimum, even though their total comp question was well within market. Widening to a fourteen to sixteen percent spread eliminates those false rejections without meaningfully increasing cost.
What a solid Tech Salary Guide 2023 should include
Beyond the raw numbers, the guide needs documentation that explains how each figure was derived, the data cut date, the source methodology, and the revision schedule. Include a section on what's excluded—bonus targets, sign-on bonuses, retention awards, benefits valuation—because those create confusion when people compare your numbers to external published ranges. I added a quick-reference table mapping job family, level, location bucket, and midpoint to base salary so recruiters could answer candidate questions without pulling a full report. For downloading a template or working file, most companies keep these internally, but you can build your own starting structure using publicly available data from levels.fyi salary reports combined with Radford survey summaries. If you need a ready-made framework, Compt and Pave both offer free trial workbooks that map directly to the band-building methodology I described. The cost of a good commercial dataset is real—typically fifteen to forty thousand dollars annually for a company our size—but the alternative is losing offers to comp misalignment, which compounds quickly. The guide isn't a product you ship and walk away from. It's a living document that requires at least an hour of maintenance per quarter per job family. Track every offer made against the band, flag any outlier, and note whether the outlier was justified by negotiation leverage or structural gaps in the data. That feedback loop is what keeps the guide accurate beyond the initial build.
