The salary map

Salary conversations in data are mostly folklore: a friend at a big tech company, a recruiter’s range, a number someone saw on a forum. Here is the same conversation with a few thousand actual responses behind it. Filter by what matches you, then type your salary to see where you sit.

How to read it

Each dot is one person’s reported base salary, converted to US dollars. Rows are job titles, grouped from the raw survey labels so “ML Engineer” and “Machine Learning Engineer” are the same row. The black tick is the median for the current filter. Colour is seniority.

The default view is US only, all levels, any remote arrangement. That is where most of the responses are, and mixing regions into one row hides more than it shows.

Five things the data says

Level and title matter about equally. Switch off everything but Entry, then everything but Senior. Within one title the median rises somewhere between a quarter and four fifths, with analysts seeing the biggest jump. Now hold Senior and read down the rows: the gap from analyst to research scientist is about the same size. There are two levers, and most people only pull one.

Engineers and scientists have converged. Data engineer and data scientist medians sit close together at every level in this sample. The old premium for “scientist” has mostly gone. Analytics engineer, a title that barely existed in 2020, lands between analyst and engineer, which is exactly where the work sits.

The analyst ceiling is real. The data analyst row has the lowest median at every level, and senior analysts in this sample earn less than mid-level engineers. Most people who out-earn it do so by changing title, not by staying an analyst for longer. This is the single most useful thing on the chart for someone starting out.

Remote does not cost you, in the US. Flip between on-site and fully remote with the US selected. For engineers, ML engineers, and research scientists the remote median is actually higher. Analysts and analytics engineers give up a few percent. The discount people expect shows up only when you widen to other regions, where remote roles are often paid at local rates.

Outside the US, the map is a different map. Select Europe or India alone and the whole axis shifts left. That is not a comment on the work. It is a comment on where the survey respondents are and what local markets pay. Compare within a region, not across.

Using it in a negotiation

1. Set the filter to your region, your level, and your remote arrangement. Ignore the rest.
2. Type your current number. If you are below the median for your own row, you have a fact, not a feeling.
3. Look one level up. That distribution is your next target, and the title change is the argument for it.

The point of publishing this is not the specific numbers, which will drift. It is that the shape of the market is knowable, and it is a lot less mysterious than the folklore makes it.