Inspect before changing

Module 1 quiz

Practice questions: choose an answer and check it to see the explanation. Nothing here is timed or graded.

1. 1. Why profile every row instead of the first 100?

   Profiling is faster on the whole fileRare defects in the tail (one bad timestamp, one padded key) still break loads and joinsTools refuse to profile samplesThe first rows are always clean
   Check
2. 2. The data dictionary says customer_code is required, but 3% of rows have it blank. What do you do first?

   Fill the blanks with "UNKNOWN"Drop the rows silentlyRecord the conflict, count the rows and confirm the rule with the data ownerChange the dictionary to say optional
   Check
3. 3. An analyst profiles order_id and reports "unique: yes" after checking the first 1,000 rows of a 2-million-row file. What is wrong?

   Nothing; 1,000 rows is a large sampleUniqueness must be checked on all rows; a duplicate can sit anywhereorder_id should be converted to a number firstProfiling should only count missing values
   Check
4. 4. A column has 1,000 rows: 940 values, 40 empty strings and 20 missing. What share of rows has no usable value?

   2%4%6%94%
   Check
5. 5. The brief asks for "active customers" but doesn't define active. What do you do?

   Pick a definition silentlyPropose a definition, confirm it with the requester and record itUse every customerUse customers who bought today
   Check
6. 6. What is the main purpose of listing assumptions before cleaning?

   To make the report longerTo make every decision visible so it can be confirmed or challengedTo avoid writing codeTo satisfy the tool
   Check
7. 7. A profile shows a status column with 12 distinct values where the dictionary lists 3. What should you conclude?

   The dictionary is wrongThere are variants (case, spacing, typos) or new statuses to investigate before mappingDelete the extra 9Nothing
   Check