Criteria

Criteria answers one question about a response: does this person meet what the role needs? You mark each question you care about as a must-have or a nice-to-have, say what a good answer looks like, and HireData returns a single verdict. It is the evaluator to reach for when some answers are disqualifying. A candidate without a driving licence cannot do a driving job, and Criteria says so without anyone reading the response.

​

Switch it on

Switch Criteria on for the form in the Evaluation tab, then click Criteria to open it. Every question in the form appears underneath with a switch of its own. A question is only checked once you switch it on here.

​

Settings per question

Switch a question on and its settings open.

• How important is this requirement? with Must-have and Nice-to-have. A failed must-have rejects the response. A failed nice-to-have only downgrades it.
• How to evaluate? with Use answer rules and Evaluate with AI.
• What answer(s) do you expect?, on the rules side, holding the conditions the answer has to satisfy.
• The note box, labelled What should AI look for?. It steers the AI and is ignored by rules. See one note per question.

​

Expected answers

A condition reads <question> <operator> <value>. The question is fixed: a rule on this question can only test this question’s answer. Pick the operator, then the value. Add condition adds a second condition and an All and Any switch above it. All needs every condition to match. Any needs one. Both apply to the whole set, not to one pair. Leave the conditions empty and the question is skipped, whatever its switch says. A question with nothing to compare cannot pass or fail.

​

Operators

The type of the question decides the operators. This is a longer list than logic offers, so a question logic cannot branch on can still be a criterion. Dates are the useful surprise. The Logic tab can only ask whether a date was given, but a criterion can ask whether the start date falls before the day you need someone. A File Upload question offers a second picker above the operator, for which part of the file to read: Answer Name, Answer File Name, Answer Mime Type, Answer Size, and Answer Url. A rule never reads what is inside the file. Only the AI does.

​

Evaluating with AI

Switch to Evaluate with AI and the conditions disappear. The model gets the question, the answer, and your note, and decides whether the answer meets the requirement. It also returns a sentence of reasoning, which Test & Evaluate shows under the answer. Use AI for free text and files, where the wording varies and no rule fits. Keep rules for answers with a fixed set of values. A Yes/No or a choice question needs no interpretation, and a rule is faster, cheaper, and repeatable. A File Upload question set to AI sends the attachment to the model. JPEG, PNG, GIF, WebP, and PDF files are read, up to five files and 10 MB per response. Anything else is judged on its file name alone.

​

The verdict

Every question you switched on and configured votes. HireData then picks the verdict in this order.

1. Does not meet criteria, when an answered must-have failed.
2. Incomplete, when a question you switched on has no answer.
3. Partially meets criteria, when every must-have passed and an answered nice-to-have failed.
4. Meets criteria, when everything passed.

A missing answer is not a failure. A candidate who never reached a must-have question is Incomplete, so a rejection always means the person answered and the answer fell short. Watch for this on questions behind a logic jump, which some respondents never see. The verdict is available to automations as {{form_response.evaluation.screening}}, which stores passed, passed_with_restrictions, rejected, or incomplete.

​

Related

• Form evaluation overview covers the five evaluators and the order you switch them on in.
• Scoring ranks the responses that pass.
• Test & Evaluate shows the verdict per question against the answers that produced it.
• Screen and score applicants with a pre-screening form builds a form around this evaluator.