“Strong CS background”
Not a column. Vibes.
Cannot be answered consistently from the resume. Should be broken into specific competencies: degree-or-equivalent, stack experience, system-design scope.
AI screening you control
AI screening columns that read like recruiter questions.
The leverage point in AI screening is not the model — it’s the column you write. ShortlistTable gives you a column system where every screen is a plain-English question with a typed output: yes/no, enum, number, date. You write the rubric; the AI does the read; every cell carries its source sentence.
No prompt engineering · Reusable templates · Typed outputs
shortlisttable.app / tables / custom-columns-demo6 columns
| Candidate | Kafka in prod | Team size led | On-call ≥ 12mo | Stack maturity | |
|---|---|---|---|---|---|
Alice Chen Ledger · 7y | Yes | 6 | Yes | Mature | Call first |
Marco Silva CloudCo · 12y | Review | ? | Yes | Mature | Review fit |
Nina Patel Indie · 4y | No | Solo | No | Early | Hold |
Citation · Alice Chen · “Led platform team of 6 engineers building event-driven services” · resume p.1Source-backed
Plain English
Column definitions
6 types
yes/no, enum, number, date, list, text
Templates
Reuse per role family
5–10
Recommended columns per role
Why most columns are bad
Most screening criteria are written for the wrong reader.
Recruiters often inherit a list of must-haves from the intake meeting that reads like meeting notes: “strong CS background, good communicator, fast learner.” Those are not screening criteria — they are vibes. They produce inconsistent verdicts across candidates and across reviewers.
A good screening column has four properties: it is answerable from the resume alone, it is one-dimensional, it carries evidence, and it produces a yes / no / needs-review verdict. Those four constraints turn an opinion into a screen.
“5+ years of experience”
Almost never a true must-have.
Replace with seniority indicators that actually matter: team size led, scope of system owned, named on-call ownership, depth of stack experience.
“Cultural fit”
Belongs in the interview, not the screen.
Not screenable from a resume; belongs in the interview rubric. Trying to screen for it produces bias and inconsistency without adding signal.
Column system
What you get when you write a screening column.
01 · Feature
Plain-English questions
“Has the candidate run Kafka in production for ≥ 12 months?” — write it once, the engine reads every resume against it. No system prompt to tune.
Has the candidate built a payments system?
On-call ownership ≥ 12 months?
B2B SaaS background?
02 · Feature
Typed outputs
Yes/no, enum (call first / review / hold), number (years), date (last role start), list (skills). Downstream filtering, sorting, and exports respect the type.
yes/noenumnumberdatelisttext
03 · Feature
Required evidence
Every column produces a verdict and a source sentence. If the resume is silent, the cell is marked needs-review instead of being guessed.
Resume p.1 · Experience“Led platform team of 6 engineers building event-driven services.”
04 · Feature
Column templates
Save a column from one role, apply it to the next. Most agency users build up a library of 30–40 reusable templates over time.
Senior backend packB2B AE packHealthcare RN packFounder pack
05 · Feature
Confidence threshold per column
Each column has its own threshold. The columns that need a recruiter eye land in the review queue; the obvious ones don’t.
0.75
default confidence
threshold
06 · Feature
Recruiter override per cell
Disagree with a verdict on a column? Override it. The audit log shows the AI value and your override side-by-side.
AI: review · 0.62
Override: yes (note)
How it compares
Custom columns vs the alternatives.
| Property | Black-box AI | ATS resume fields | Spreadsheet | ShortlistTable |
|---|---|---|---|---|
| Define columns yourself | –Limited templates | –Schema-bound | ✓Yes — manual | ✓Plain English |
| Typed outputs | ✕Score only | –ATS-defined | –Manual | ✓yes/no, enum, number, date |
| Reusable templates | –Vendor templates | –Tied to ATS schema | –Copy-paste | ✓Save once, apply anywhere |
| Confidence threshold per column | ✕No | ✕No | ✕No | ✓Yes |
| Recruiter override per cell | ✕Limited | –Limited | ✓Yes | ✓Yes — audit trail |
✓ Supported– Partial / manual✕ Not supported
FAQ
Common questions on custom columns.
How many columns should I have per role?+
5–10 works best: 3–5 must-haves, 2–4 nice-to-haves, and one or two contextual columns (current title, years in role) for sorting. More than 12 and the sheet stops being a recruiter tool and starts being a database project.
Do I have to know prompt engineering?+
No. Columns are written as plain-English questions, not prompts. “Has the candidate built a payments system?” is a complete column — there is no system prompt to tune.
Can I save a column template for re-use?+
Yes — templates are first-class. Save a column from one role and apply it to another with one click. Most agency users build up a library of 30–40 templates over time.
What output types are supported?+
Yes/no, enum (custom values you define), number, date, list (e.g. skills), and free text. Downstream filters and exports respect the column type.
Can columns reference each other?+
Not directly — each column reads the source documents independently. If you need a derived column (e.g. “qualifies for shortlist if all must-haves are yes”), set a confidence threshold and a hold queue instead.
How do I handle role-family variations?+
Templates can be cloned and edited. A common pattern is one base template per role family (e.g. “senior backend”) and per-role overrides for the specific must-haves of that search.
More from us
Keep exploring.
Related
Source-backed screening
Citations on every cell.
Related
AI resume shortlisting
How verdicts become priority.
Related
Candidate ranking
Review priority derived from columns.
Free tool
Try the free JD ↔ CV matcher
See semantic column matching live.
Read
Building screening columns
A working guide for hard roles.
Write better columns
Stop screening on vibes. Start screening on questions.
Write your first five columns, drop 25 resumes, get a sheet. No credit card.