Free salary estimates, and a loop that makes them smarter

Your resume already carries a lot of signal about what you are worth: your title, how senior you are, how many years you have, your industry, where you work. We added a small feature to the report card that reads those signals and gives you a likely salary range. It is free, and it shows up right next to your grade.

Estimating from resume signal

The estimate does not read your whole resume as a wall of text. It works from a short, structured summary of the things that actually move pay: seniority, years of experience, industry, and location. From those, a simple and transparent engine produces a range. We made the range deliberately wide. A salary estimate that pretends to know your exact number to the dollar is lying to you. A believable range is more useful, and more honest.

This is a rebuilt version of an idea we ran years ago. The old one was a small classifier that got somewhere around eighty percent of the way there. The bar to beat is not high, and a free engine clears it comfortably.

A teacher that nudges it

On top of the free engine, we run a second pass that reviews each estimate and suggests a correction when the first number looks off. Think of it as a teacher checking the work. When the two agree, we are confident. When they disagree, that disagreement is information we keep. The free engine stays the product, the review just sharpens it.

The part that makes it smarter: your corrections

Here is the loop. Under the estimate is a slider. If our range is wrong, you can drag it to where you think it should be and tell us. We store every one of those corrections.

One thing we were careful about: when you move the slider, you are telling us your target, the number you are aiming for, not necessarily what you make today. We record it as exactly that. Mixing up “what I want” with “what I have” would quietly poison the data, so we keep those two ideas separate from the start.

Every honest correction is a real-world label. Over time, a pile of those labels lets the estimate calibrate against what people actually report, instead of guessing in a vacuum. That is the flywheel: the more people use it, the better it gets, and it improves without us having to hand-tune anything.

A small confession

Building the thing was the easy part. Making it actually appear on the page took some debugging. The widget was quietly tied to an unrelated piece of the report card, so it loaded only some of the time and looked broken the rest. We gave it its own trigger, and now it shows up every time. A good reminder that a new feature should stand on its own and not borrow another feature’s plumbing.

The estimate is live on the report card now. Try it, and if we get your number wrong, tell us. That is the whole point.