Screen and score applicants with a pre-screening form

Most application forms collect answers and leave a person to read them. This recipe builds one that decides. Criteria rules out the applicants who cannot do the job, Scoring ranks the ones who can, and the summary gives your recruiter the response in three sentences before they open it.

​

What you will build

One form with four parts:

1. Four questions, each of which decides something
2. Criteria that rejects an applicant who fails a hard requirement
3. Scoring that ranks everyone who passes
4. A summary and a generated field your automation can act on

Every setting used here is covered in the evaluation reference. This page walks one form from blank to published.

​

Before you start

Decide two things with whoever owns the vacancy.

• What disqualifies someone. These become must-have criteria. Be strict: a must-have that is really a preference rejects people you wanted to talk to.
• What makes one qualified applicant better than another. This becomes the scoring, and it is usually one or two questions rather than all of them.

Write both down before you open the builder. The form is quick. Agreeing on the rules is not.

​

Build the form

​

1. Ask the fewest questions that decide something

Open Forms, click New Form, and pick Blank Form. A pre-screening form earns its keep by being short, so add only the questions whose answers change what happens next. Finish with an ending screen so applicants know the form submitted.

​

2. Switch the evaluators on for the form

Open the Evaluation tab and switch on Criteria, Scoring, and Summary. Leave Skills off: it sorts qualified people by strength, which is a different job from this one. Do this before you configure anything. A question’s evaluator switch does nothing while the form-level switch is off, and it does not even appear on the question card. See the enable order.

​

3. Set the knock-outs with Criteria

Click Criteria to open the list of questions, then switch on the three that matter.

• Do you have a driving licence? Set Must-have, keep Use answer rules, and expect is equal to Yes.
• When can you start? Set Must-have, keep Use answer rules, and expect is less than or equal to the date the client needs someone. Logic cannot compare dates, but a criterion can.
• Why do you want this job? Set Nice-to-have and switch to Evaluate with AI. In What should AI look for?, say what a real answer contains, such as “Look for a specific reason connected to this work. Generic enthusiasm does not count.”

Leave the experience question out of Criteria for now. It is the ranking question, and putting it in both places means one answer can reject and rank at the same time, which makes a surprising result hard to read. The verdict is Does not meet criteria as soon as one answered must-have fails. An applicant who never reached a must-have question comes back Incomplete instead, so check that no logic jump can step over one.

​

4. Rank the rest with Scoring

Click Scoring. Keep the three ranges it starts with, Low at 0, Medium at 50, and High at 80, then switch on the experience question.

• Set How many points is this question worth? to 20.
• Add a rule: is greater than or equal to 5, worth 20 points.
• Add a second rule: is greater than or equal to 3, worth 10 points.

Both rules match an applicant with five years, and the total is capped at the question’s maximum, so they score 20 rather than 30. That cap is what lets you write overlapping bands and stop worrying about the gaps between them. One 20-point question means the score is 20 out of 20 for five years, which is 100% and High. Add a second scored question and the maths spreads across both. See how the percentage is worked out.

​

5. Let the AI write the handover

Summary needs no configuration. It writes the finished response up in a few sentences for whoever picks it up. For the one decision your automation has to make, add a generated field. Switch Generated fields on, click Create custom, and build a List output:

• Name: Next step
• Reference as: next_step
• What is it for?: What should happen with this applicant now
• Options: Book an interview, Ask for more detail, Reject politely

A list output can only return one of your options, which is what makes it safe to branch on. A free-text output would give you a different sentence every time. In Hints for the AI (optional), add “Leave a field empty when the answers do not say.” Without it, the AI guesses at an output when the response is thin.

​

6. Test it before you publish

Open Test & Evaluate and run all four scenarios with Autofill with AI. Each one has a result you should be able to predict. Open each result panel and check the per-question detail, not only the headline. A right verdict reached for the wrong reason breaks as soon as a real applicant answers differently. Remember that running an evaluation saves the form, so anything you were still deciding about gets committed.

​

7. Act on the result in an automation

Publish the form, then build the automation that does something with the result. Automation variables at the bottom of Test & Evaluate lists every value with the result from your test run beside it, which saves guessing at the exact wording. Start with one condition on the verdict, such as rejections going to a polite decline and everything else creating a task. Add the score once you trust it. See Automations.

​

Review and improve the form

After twenty or thirty real responses, compare the verdicts against what your recruiters would have decided.

• Are you rejecting people the team would have called? A must-have is too strict, or it belongs in Scoring.
• Is everyone landing in the same score range? The minimums need moving, or the scored question does not separate applicants.
• Does the AI criterion agree with a person reading the same answer? If not, the note is too vague. Say what a good answer contains, not that it should be good.
• Are the summaries worth reading? If they only repeat the answers, the form is asking closed questions and the summary has nothing to add.

Change one thing at a time and re-run the scenarios in Test & Evaluate before you publish again.

​

Need more help?

If the evaluators return nothing, the usual cause is the enable order. Check the form-level switch as well as the per-question one, on the Evaluation tab. If a criterion never fails, open the question in Test & Evaluate and read the expected answer under it. A criterion with no conditions is skipped rather than failed. If you still need help, contact support@hiredata.com and include:

• The form name
• The question and the rule that behaved unexpectedly
• The answer you tested with, and the result you expected