A job description template that produces better candidates

In our own screenings the median candidate matched 38% of what the job description asked for, and 62% matched under half. Some of that is people applying widely. A lot of it is postings that ask for nine tools when the work needs three. This is the template we would write, and the reasoning behind each part.
First, what a job description is for
It has exactly two jobs: make the right people apply, and make the wrong people not bother. Most postings do neither, because they are written as an internal document — a list of everything the role might ever touch, approved by three people, with nothing cut.
The cost of that is measurable at your end. Every unnecessary requirement widens the pool and every extra applicant costs you reading time, which is the most expensive part of hiring. It costs you at the other end too: most mediocre hires cleared a requirements list that was never really the job.
The template
1. A title people search for
Not "Growth Ninja". Not "Officer, Commercial Excellence". The words a candidate would type. If two titles are both honest, use the more common one.
2. One sentence on what the person will actually do
Before any company boilerplate. "You will own the monthly management reporting for three business units." A candidate should be able to tell from this line alone whether to keep reading.
3. Three to five real requirements
This is the part that does the work. The test for each one: would you reject an otherwise excellent candidate who lacked it? If no, it is not a requirement. Move it down.
4. A separate "nice to have" list
Where the other six tools go. Candidates read these very differently — a long requirements list reads as a wall, the same items under "nice to have" read as an invitation.
5. What the first three months look like
Two or three lines. This is the single most useful thing you can add and almost nobody does. It tells a strong candidate whether the role is a step up or sideways, which is the question they are actually asking.
6. Who they work with
Who they report to, who they depend on, whether anyone reports to them. Roles sitting between two departments should say so — the people who are good at that will self-select in.
7. Salary range, or say why not
If you cannot publish a range, say that the range is set and will be shared at the first call. Silence reads as "we will pay as little as you accept", and the best candidates are the ones who can afford to skip that.
8. Location and pattern, in plain words
Office, hybrid or remote; which days; which city. "Flexible" is not an answer and generates exactly the applications you do not want.
9. What happens after they apply
One line. "You will hear from us within two weeks, either way." Then keep it — this is the cheapest promise in hiring and the most commonly broken.
Once the posting says what you actually need, the next question is how much each of those things is worth relative to the others. That is a weighting, and it is worth writing down before anyone applies.
A worked example of the part that matters
The requirements section, before:
| As usually written |
|---|
| Bachelor's degree in Business, Finance, Economics, Statistics or a related field. 3–5 years of experience. Advanced Excel, SQL, Power BI, Power Query, DAX, Tableau, Python preferred. Strong communication skills. Excellent attention to detail. Ability to work under pressure. Team player with a can-do attitude. |
And after, for the same role:
| Required | Three years building reports someone else depended on · SQL good enough to write your own queries · Advanced Excel, including the parts people claim and cannot do |
|---|---|
| Nice to have | Power BI or Tableau · Power Query or DAX · Python |
| Not listed | The degree field, unless it genuinely disqualifies · "attention to detail" · "team player" |
The second version filters harder on the things that matter and stops filtering on things that only sound like requirements. "Strong communication skills" excludes nobody — everyone claims it — while "SQL good enough to write your own queries" is a sentence a candidate can honestly fail.
What we noticed, and will not claim
Across our five real postings, the two longest job descriptions — both over 4,300 characters — produced the lowest median skills coverage of the set, and the shortest produced the highest. That is the direction this article argues for.
It is also five job descriptions at one company, and we are not presenting it as a finding. Five is not enough to tell a pattern from a coincidence, and the roles were different enough that the comparison is unfair. We mention it because it is what our data shows, not because it proves anything. When we have enough postings to say something real, we will.
How to tell whether yours is working
You do not need software for this. After your next posting closes, take the applications and answer three questions:
- What share of applicants met the requirements you listed? If it is very low, either the posting attracted the wrong people or the list was aspirational.
- Which requirement did most people fail? If it is one you would not actually reject over, delete it before the next round.
- Did the people you interviewed look like the posting? If your shortlist keeps containing people who fail a stated requirement, the requirement is fiction — and it has been quietly filtering out others who took it seriously.
That third one catches more bad postings than anything else we know of. A scoring model reads your requirements literally, which is uncomfortable but useful: it shows you what your posting actually said rather than what you meant.
Questions we get asked
How many requirements should a job description list?
Three to five real ones, with everything else under "nice to have". The test for each: would you reject an otherwise excellent candidate who lacked it? If not, it is not a requirement.
Why do so few applicants match the job description?
In our screenings the median candidate matched 38% of what was asked and 62% matched under half. Some of that is people applying widely; a lot of it is postings listing nine tools when the work needs three.
Read next
What it actually costs to hire one person in Bangladesh
Job posting prices, measured screening costs, and the one line item only you can price. A worked breakdown rather than a borrowed average.
AI in hiringWhat AI screening can and cannot judge from a CV
Six dimensions, what each one reads, which actually separate candidates, and the three things no parser sees. With our own numbers.
PracticalTurning "we want a good analyst" into scoring weights
Six dimensions, three real weightings side by side, and why weighting achievement heavily can measure CV writing instead of ability.
Cost of hiringThe hidden cost of a bad hire
Why the famous multipliers do not transfer, which two components you can price from your own records, and the one-hour self-audit.