SnipTray logo
Data

Spot-check a dataset

The five queries to run before trusting it.

  • data quality
  • validation
  • sanity checks

The prompt

I have a dataset: {{description}}. I want to use it for {{intendedUse}}.

Give me the checks to run before I trust it — row counts against a known total,
duplicate keys, nulls in fields I am about to group by, dates outside plausible
range, and whatever else this shape of data usually gets wrong.

For each check, say what a bad result looks like and whether it kills the analysis
or just needs a caveat.

What to fill in

{{description}}
What the dataset is — e.g. one row per subscription event, from Stripe
{{intendedUse}}
What you'll use it for — e.g. monthly churn by plan

117 prompts, in the app you already paste from

The whole library, searchable, with your own prompts alongside it. Free to browse here; Pro to keep them a keystroke away on every device.