Panel hygiene auditor
What it does
Panel hygiene is the maintenance work that keeps a first-party participant panel worth recruiting from. Duplicate records split one person's history in two. Contacts go stale as people leave companies and change addresses. A handful of willing regulars end up in study after study, which is participant fatigue, and it quietly narrows who your research actually represents. Nobody schedules this work. It surfaces as a bounce rate or an awkward email from someone who has been asked four times this quarter.
This skill runs the audit against the panel in your Great Question account. It works from what your records carry: opt-in status and the date consent was given, unsubscribe status, when each record was last updated, and participation history across studies. No definition of a stale contact is built in, so you name the threshold in the request, whether that is a consent date older than two years or anyone who has been in more than three studies this year. Findings come back grouped so you can act on them, and it runs from Claude, Cursor or ChatGPT with the MCP connected.
What you get back
- Likely duplicate records, grouped, with what they have in common.
- Contacts that fail the threshold you set, with the signal behind each flag, whether that is an old consent date, an unsubscribe, or a record nobody has touched in months.
- People whose participation history is heavier than the rest of the panel.
- Segment gaps the audit exposes, so you can see who you are running short of before the next study.
- Records updated in place where you ask for it, including marking someone unsubscribed or re-consented, and records deleted when you decide one should go.
How to use it
Example Prompts:
"Audit the panel before we recruit for the reconciliation study."
"Who in the panel has been in more than two studies this year?"
"Show me duplicate records for anyone at our enterprise accounts."
What it won't do
There is no merge operation. Duplicates come back grouped for you to resolve by hand, usually by updating one record and deleting the other. It also reads only what your records contain: a contact who changed jobs without telling you looks fine until an email bounces. Two records that look like one person sometimes are two people, and that call stays with you.
Questions
How does it decide someone has been asked too often?
It doesn't decide on its own. No frequency window is built in and there is no threshold to configure, so you name the number: anyone in more than two studies this year, say. It reads participation history across the studies in your account and works from that.
Will it delete anything?
It can delete a candidate record, and it can update one, including marking a person unsubscribed or re-consented. Nothing changes unless you ask for it, so expect to confirm anything that touches your panel. What it cannot do is merge two records into one.
Should I run this before every study?
Before a recruiting push is the moment it pays. Running it monthly also works if your panel grows steadily, since duplicates arrive with new imports more than with individual signups.
