<aside> 🔒
Recreated for this portfolio. I built this interviewing framework at Tandem Care, a home care company, before it had an in-house HR team. The original scorecard belongs to the company, so the example questions and flags below show the same method without confidential details.
</aside>
Two interviewers talk to the same candidate. One says "strong hire," the other says "not sure." Neither is wrong; they were listening for different things. In a small, fast-growing company, that inconsistency means bad hires get through and good ones get lost, and nobody can explain why.
Most interviewer training hands out a question list and stops there. A question list doesn't tell anyone what a good answer sounds like.
The goal: everyone who interviews listens for the same things, scores them the same way, and keeps doing it after training ends.
Who I trained: the managers and coordinators who conducted interviews.
| Stage | What happens | What it proves |
|---|---|---|
| 1. Agree the standard | Go through every interview question together: what does a green flag sound like, and what does a red flag sound like? No one scores anything until this is agreed. | Everyone is listening for the same thing |
| 2. Practice at zero stakes | Watch a recorded interview together, score it independently, then discuss every difference until the group agrees. | They can apply the standard the same way |
| 3. Live sit-ins | I sat in on two live interviews with each interviewer, then debriefed. | They can do it with a real candidate |
| 4. Ongoing checks | Periodic AI scoring of interview transcripts against the rubric I wrote, compared with the interviewer's own scores. | Calibration holds over time instead of drifting |
| Question | Green flag | Red flag |
|---|---|---|
| Tell me about a time a caregiver called out an hour before a client visit. What did you do? | Walks through a specific case: who they called first, how they protected the client, what they documented | Speaks only in generalities, or the answer ends with "I told my manager" |
| How do you handle a family member who is upset about something you didn't cause? | Acknowledges the feeling, owns the next step, keeps the client's care at the center | Explains why it wasn't their fault |
| Walk me through how you keep track of 30 open tasks in a day. | Describes an actual system and how they decide what comes first | "I just stay on top of it" |
<aside> 💬
Scoring rule. A score needs a reason tied to something the candidate actually said. "Good energy" is not evidence.
</aside>
Most calibration is a one-time workshop. People drift back to gut feel within weeks. Periodic AI scoring against a written rubric made drift visible: when an interviewer's scores and the rubric's diverged, that was the cue for a conversation, not a one-off training that everyone forgets.