Market data, salary bands and a realistic request

AI for HR and recruiting · Lesson 2 / 21

What this role actually costs

Half of all stalled vacancies close not by rewriting the ad but by admitting the band sits 20-25% below market. Check that before publishing, not after two months of silence. A model helps here as an analysis accelerator, never as a source of numbers: it does not know your market this week, and it will hand you a global average in a confident tone.

The working sequence

Gather the raw material yourself. 15-25 current ads for the same role, your city or remote setup, with stated bands. Plus what you already pay people in that position.
Hand the analysis to the model. It computes medians well, finds the spread and spots unusual terms.
Inspect the edges. An ad with a band twice the median almost always describes a different role under the same title.
Compare against your internal grid. Hiring 30% above your current staff creates a problem you will be solving in six months.

Prompt for the extract

Below are 20 job ads for role X. For each one extract: the band
(lower and upper bound), work format, mandatory stack, and seniority
as implied by the body text rather than the title. Then compute the median
lower and upper bound, show the three ads that deviate most, and explain
how their content differs. If a band is missing, mark it as missing
and do not invent one.

Check the realism of the request itself as well. A manager often asks for a senior at mid-level money and wants it closed in three weeks. A useful number for that conversation: one accepted offer typically costs 6-10 final interviews, and 5-8 weeks pass between publication and start date for non-executive roles. If the deadline and the band do not add up, you have three moves: raise the money, lower the requirements, extend the timeline. There is no fourth, and this is best settled at the start.

Insight. A model given 20 real ads is useful. A model simply asked what analysts earn returns an internet-wide average for an unknown year.

Common mistake. Publishing without a band so you can discuss it later. You pay for that in applications and in time spent on candidates who walk after the first money conversation.

Pro tip. Price the empty seat, not just the salary. A month of an unfilled role often costs more than the annual gap in the band.

Cheat sheet

You collect the numbers, the model does the analysis.
15-25 ads is the minimum base for a median.
Deadline, band and requirements move together.
A band above your internal grid creates a delayed conflict.

1. Why not ask a model for the market band directly?

2. What do you do when the deadline and the band do not add up?

3. What does an ad with a band twice the median usually indicate?

Task — checked by AI

Take a role you are opening now or closed recently. Collect 15-20 current ads for the same role and run them through the prompt from this lesson. Report: the median lower and upper bound, your current band, the gap between them, which ads were outliers and why. Then state which of the three levers (money, requirements, timeline) you propose to move and how you will argue it to the hiring manager.

← Back

🔒 Answer the question correctly to move on to the next lesson.