Most candidates undershoot their market rate — not because they fail at negotiation, but because they fail at research. The default move is to look at one or two data points — a Glassdoor listing, a friend's salary from years ago — and anchor on a number that is below where they could realistically negotiate. Underprepared negotiators ask for less because they do not know what is possible. The fix is not a better negotiation script. It is better data.
How Levels.fyi changed salary transparency for tech
Before Levels, tech compensation was opaque and companies had a significant information advantage. Levels aggregated real total compensation data — base salary, bonus, equity — from thousands of engineers, PMs, and designers at major tech companies. Filtering by company, role, and experience gives you a distribution, not just a range, which tells you where the 25th, 50th, and 75th percentiles are. Most initial offers land near the 25th percentile, which means the 75th percentile is a realistic target for negotiation — if you know to ask for it.
The triangulation method that catches systematic biases
Each salary database has biases. Levels skews toward engineering at large companies. Glassdoor has response bias toward people who had strong opinions about their pay. LinkedIn Salary is broad but less granular. Using three sources and taking the median of their 75th percentiles gives you a number that is both defensible and realistic. When a recruiter asks for your range, you can say exactly where it comes from — which signals that you have done your homework rather than guessed.
What to do when the data is not there
Smaller companies and non-engineering roles are underrepresented in public salary databases. In these cases, the network is the best data source. A 20-minute conversation with someone at a similar level at the company, or at a comparable company, reveals the actual ranges in a way that no database can. This is one of the highest-value uses of informational interviews — not just career advice, but concrete compensation intelligence that changes what you ask for.