"Seemed strong" is not a screening result. If two recruiters screen the same developer and one says "strong" while the other says "maybe", the problem isn't the candidate. It's the lack of a shared rubric.
Here's a simple rubric recruiters and hiring managers can use for first-round technical screening, with worked examples.
The three-level scale
Keep it small. Five or ten levels invite hair-splitting. Three is enough for a first round:
| Score | Label | What it means |
|---|---|---|
| 1 | Not shown | Generic or memorised answer. No real example, or the example doesn't hold up to a follow-up. |
| 3 | Credible | A real example with some detail, a sensible approach, and an honest account of what happened. |
| 5 | Strong | Clear reasoning, trade-offs considered, a measurable outcome, and lessons learned. |
Use 2 and 4 only when an answer genuinely sits between two levels, and always write why.
Score each must-have separately
Don't give one overall score. Score each of the role's three to five must-haves on its own line. A candidate can be a 5 on debugging and a 1 on system design, and the hiring manager needs to see that.
Worked example: "Tell me about a slow page you fixed"
Score 1 answer:
"I'd use caching and optimise the database queries, and maybe add a CDN."
No example, just a list of techniques anyone could recite.
Score 3 answer:
"Our dashboard took about eight seconds to load. I profiled it and found an N+1 query on the orders table, added eager loading, and got it under two seconds."
A real problem, a method and a result.
Score 5 answer:
"Same dashboard, but I also explained why we didn't just add caching: the data changed every few minutes and stale numbers would have confused the finance team. We fixed the query, added an index, and set up an alert if load time went above three seconds. That alert caught a regression a month later."
Reasoning, a trade-off, an outcome and a lesson.
Always write a one-line reason
A score without a reason can't be checked or defended. Good reasons look like:
- "3: real example, but couldn't explain why the fix worked."
- "5: explained trade-off between caching and data freshness; gave numbers."
- "1: listed techniques, no example even after a follow-up."
Calibrate once a quarter
Pick three recorded answers. Have two or three recruiters score them independently, then compare. Where scores differ by two or more, discuss and update the level descriptions. Fifteen minutes a quarter keeps the whole team consistent.
Don't forget integrity
A 5-level answer is only worth something if it's the candidate's own. Note any proctoring flags (tab switches, copy-paste, a second voice) next to the score, and treat them as a reason to look closer, not an automatic rejection.
Where SimpleHire AI fits
SimpleHire AI helps recruiters screen candidates and review their responses along with proctoring evidence, so they can make a better and stronger client submission. Structured questions and consistent scoring are built in, and the recruiter still makes the call. If you're building question sets, the free screening question guides are a good place to start.
Visit SimpleHire AI: https://simplehire.ai/
