Screening 500 resumes: the full walkthrough

The whole run, start to finish, on all 500 CVs. Allow about forty minutes the first time and fifteen once you have done it before.

Before you start

Have three things in front of you: the job blueprint from Module 2, your three filters, and your scores with their proof notes.

If you are inventing rules while looking at candidates, you are not screening. You are justifying choices you have already made, and the record will show it.

Step 1 — Check everything arrived (2 min)

Compare the number of candidates in the system against the number of files you uploaded.

If they do not match, some files were rejected — usually password-protected PDFs, files that are too big, or formats the tool does not read. Those people are not low-scoring. They are missing, and missing looks exactly the same as never having applied.

Find them in the log and deal with them separately. Never let a file format decide who gets read.

Step 2 — Spot-check the reading (5 min)

Open five candidates: two ordinary layouts, two with sidebars or columns, one scan. Look for the problems from Lesson 2, especially dates.

If dates look unreliable in this batch, change experience from a filter to a score before going further. Doing that now takes thirty seconds. Discovering it afterwards costs you the whole run.

Step 3 — Enter your rules (8 min)

Three filters. Everything else as scores, with weights that add up to something you can reason about.

Write the proof note for each score. The tool uses it, and more importantly you will re-read it in Step 6 when you are deciding whether you agree with the result.

Step 4 — Run it (3 min)

Screen the whole set. While it runs, write down a guess: how many will get past the filters?

Comparing your guess with the answer tells you whether you understand your own pool. If you guessed much too high, your filters are stricter than they sound. Much too low usually means one of them is not filtering anything at all.

Step 5 — Look at the shape, not the list (5 min)

Before opening anyone, look at how the scores are spread out.

Step 6 — Read the ones you disagree with (12 min)

This is the step that separates screening from just handing the job over. Open six people:

1. The three highest scorers who were rejected — people who failed a filter but did well on everything else. This is where over-filtering shows up.
2. The three lowest scorers who got through — people who only just cleared the bar. This is where your rules are too loose.

For each one, decide whether you agree. Almost always the answer is that a rule was worded badly, not that the software behaved oddly. Fix the rule, not the person. Moving one candidate by hand fixes one candidate; fixing the rule fixes everyone it was doing this to.

If you agree with all six, that tells you something too. Your rules are probably too blunt to be making real distinctions.

Step 7 — Draw the line and write it down (5 min)

Choose your shortlist by how many interviews you can actually run, not by a score. "The top 20" is an honest statement about your calendar. "Everyone above 78" pretends the score is more precise than it is.

Save your rules alongside the shortlist. When someone asks in three months why a candidate was not progressed, that record is the whole answer — and you cannot put it together afterwards.

Five mistakes people make first time

• Setting rules before checking how well the CVs were read, then blaming the rules.
• Ignoring it when the file count does not match.
• Reading the top ten and stopping.
• Adjusting weights until a favourite candidate rises, without writing down that you did.
• Picking a score threshold and reusing it on the next job, where it means nothing.