highlight · part vi · 1

Trust in your data is a relationship, not a score.

Trust in your data comes from carefully crafting the codebook. This relationship cannot be reduced to a number.


Validation is a relationship you build with your codebook. You learn where it's reliable. You learn where it isn't. You learn which round of refinement made which field trustworthy.

after three rounds of minting, you know:

  • Question 1✓rock solid
  • Question 2✓rock solid
  • Question 3✓rock solid
  • Question 4✓rock solid
  • Question 5✓rock solid
  • Question 6✓rock solid
  • Question 7✓rock solid
  • Question 8✓rock solid
  • Question 9⚠needs a human eye on historical documents
  • Question 10✓works beautifully now — didn't until round 3

Question 9 is not unfinished. Its splits have no policy behind them, so they stay out of the codebook, and a person reads them. A score collapses all this into one dimension. You know a codebook the way you know a colleague — specifically, over time.

Earned through use. Not granted by a metric.

Helping you build a trusting relationship with your data.

@karlrohe