What is PDQ Similarity Search?
When an employee completes a Position Description Questionnaire, someone has to answer the question it raises: which classification does this work actually belong to? Reading a PDQ against specs one at a time is slow, and it quietly biases the review toward the classifications the analyst already knows well.
The PDQ Similarity Search compares the PDQ's content against every class specification in your catalog at once and scores how closely each one matches the work described. The result is a ranked list of your classifications by fit, which turns "does this position match its current spec?" into a question you can answer with the whole catalog in view. That makes it a core tool for spotting positions that may be working out of class.
You don't need a formal PDQ to use this tool. Any detailed description of duties works, so the report also answers "I have this list of duties, which of our specs fits best?"
Common Use Cases
Reviewing a Reclassification Request
An employee submits a PDQ asserting their work matches a higher classification. Run the report to see where their described duties actually land across the full catalog: how strongly they match the requested classification, the current one, and any classifications nobody had considered.
Placing a New or Changed Body of Work
A department describes a role that's evolved past its old spec, or a new function with no obvious home. Paste the duties into the report to identify the closest existing classifications before deciding whether one fits or a new classification is warranted.
Running This Report
PDQ Similarity Search runs from the Reports section: select PDQ as the report type, enter a report title, and provide the PDQ content in the PDQ response field, either pasted as text or uploaded as files (PDF, Word .docx, or Excel .xlsx). You can attach up to 5 files, so supporting documents like a desk audit or a 10-day duty tracker can be analyzed alongside the PDQ form itself. For the steps, see Running Reports for Bulk Analysis.
Note: Your agency's configuration determines which report types are included for your account, so this report may not be available to you. Questions about your configuration? Contact Holly Support at [email protected].
How the Matching Works
The report evaluates the work described in the PDQ against the content of every spec in your catalog and scores each comparison from 0 to 100%. The evaluation uses the same comparison approach as Holly's internal similarity analysis, which weighs core duties and responsibilities most heavily, followed by seniority and supervisory level, required knowledge and skills, and work environment. For the full breakdown of those criteria, see Understanding Similarity Analysis.
Two properties of the process are worth understanding:
Every spec gets a score. The report doesn't apply a cutoff, so the ranking is complete. If the best match is only 55%, that's visible, and it's a finding in itself: the work may not fit cleanly anywhere in your current structure.
Each score comes with an explanation. Alongside every score, Holly provides a short explanation of what drove the match or mismatch, so you can evaluate the reasoning rather than trusting a bare number.
Objectivity and Equity Considerations
Classification reviews are vulnerable to familiarity bias: analysts match faster to classifications and people they know. The PDQ Similarity Search evaluates only text. It reads the PDQ's description of the work and each spec's content, and it knows nothing about the employee, their history, or anyone's prior relationship with the position. That separates the position from the person, which is the standard classification analysis aims for.
The similarity score measures how well the described work aligns with a classification's content. It is not a determination that the employee meets that classification's minimum qualifications. Because duties carry the most weight in the evaluation, a classification can surface as a strong match even where its stated MQs differ from the employee's background. Treat that as useful information in both directions: it keeps restrictive MQs from hiding the right classification during the matching step, and it means MQ eligibility remains a separate question for your team to resolve.
Note: Match scores and explanations are AI-generated to support your classification judgment, not replace it. Classification decisions should follow your agency's established review process.
Understanding Your Results
The results page opens with summary counts of how your catalog distributed across three match bands: High match (above 70%), Medium match (50 to 70%), and Low match (below 50%). Below, the results table lists every spec with its similarity percentage, sorted highest first.
Compare the top candidates against each other. When two or three classifications score closely, the hover explanations usually identify the distinction, such as supervisory level or a specialized duty, that your review should focus on.
Check where the current classification landed. For a reclassification review, the gap between the position's current classification and the top match is the finding: a current classification scoring well below another spec supports a closer working-out-of-class review.
When your review needs to happen outside Holly, export the results; the ranked list and explanations travel with the file.
Filtering Your Results
The results table filters by similarity band, the same three bands as the summary counts. Click the filter control on the Similarity score column and select any combination of below 50%, 50% to 70%, and above 70%, then confirm.
On a large catalog, filtering to the above-70% band is the fastest way to isolate the handful of classifications worth serious review, without scrolling past hundreds of low matches. Combine it with the column sort to order that shortlist highest first.