Study guide

Study guide for chapter 3 of the Data 360 Architect course. Read the five modules first; this page helps you check what stuck.

About this study guide. It’s built only from this course’s own content and the published Salesforce exam guide’s list of objectives. It is never based on real exam questions, and it doesn’t claim to know what’s on the exam. If you’ve taken the exam, please don’t send me questions from it: the certification agreement forbids sharing them, and memorizing answers is the wrong way to become an architect.

This course covers the Data 360 Consultant exam, but the exam is just a step on the road. As architects, we go beyond it. Chapter 3 is the work before anything comes in: the shape of the data, what makes each record unique, the dictionary where the design happens, and which fields stay out.

For the Consultant Exam

What this chapter covers of the Data 360 Consultant exam guide (opens in a new tab)’s objectives. Tick them off as you go. Each one links to where it’s taught.

SP-4: Data ethics and governance

• I know a deletion request through the Consent API removes the person’s Individual and what’s related to it by the individual’s ID, but not data in unrelated data model objects. (Personal data has to be erasable)
• I classify every field’s sensitivity in the dictionary, before the field exists. (Classify it at the dictionary)
• I can name Salesforce’s five practices for ethical data use, including “collect and use only what you need.” (Use it ethically, not just legally)

ING-2: Bringing data in from many sources

• I can tell the four shapes of data apart (flat, normalized, semi-structured JSON, unstructured) and what each needs before it comes in. (Flat, Normalized or Unstructured)
• I know Data 360 takes delimited text files such as CSV, and Parquet files from cloud storage. (Flat: one row per thing)
• I check free text for line breaks inside a field before it comes in. (Unstructured: free text and documents)
• I know the primary key decides whether an incoming row adds a record or updates one. (The key decides everything that follows)
• I know that with a record modified field, the newer value wins, not the last row loaded. (The same record, more than once)
• I know Data 360 only “attempts to detect” each field’s type, and that a number it can’t read is stored as null. (Why the type column matters)
• I know an engagement stream needs an event time that never changes, and that a time series without one comes in as Other. (Dates: which ones never change)

HU-2: Modeling data

• I can test a key: unique, unchanging, meaningless, never personal data. (What makes a good key)
• I know how to build a composite key with a formula field when no single field is unique. (When no single field is unique)
• I know a key qualifier keeps two sources’ key values apart. (Keys across sources)
• I know a field’s data type can’t be changed after its lake object or model object is created. (What you can’t change later)

As an Architect

The consultant exam is the floor. This is what an architect takes from this chapter, beyond it.

• Do the thinking outside Data 360. Names, types and keys are set when the stream is built; settle them in the dictionary first. (Do the thinking outside)
• One field chosen in a hurry becomes two, forever. A wrong type means a new field, a new mapping, and the old field kept. (One field, chosen in a hurry)
• Get the structure and a real sample of every feed, and lay one against the other before the design starts. (First, every feed)
• Join on the key that never changes, brought in as text: the runway’s airport_ref, not the airport’s code. (Two keys for one airport)
• Flights are other data in this course: no flight date holds still, because airlines reschedule. (Dates: which ones never change)
• Ask the producer to write the file name and a batch ID into every row, so you can always tell which file a row came from. (Which file did this row come from?)
• Read a dictionary right to left, from what a field is used for back to its source. (Read it right to left)
• Four questions at the door for every field: what it’s for, its type, how sensitive it is, and who owns it. (Four questions at the door)

Words to Know

Term	In one line
Flat data	One row per thing, every value in its own column
Normalized data	Things in separate files that point to each other by key
Semi-structured data	Data with nesting and lists inside it, such as JSON
Unstructured data	Free text and documents, with no rows of their own
Primary key	The field that says whether a row is new or an update
Composite key	A key built from two or more fields, here with a formula field
Key qualifier	A value that says which source a key came from, so sources don’t collide
Record modified field	The date that decides which version of a record is newer
Event time field	An engagement stream’s date of the event, which must never change
Data dictionary	One row per field: its meaning, type, key, sensitivity, where it lands and where it maps

Check Yourself

Short scenarios in the style of the course. Decide your answer before you open it.

1. A stream brought years flying in as text, and a segment for "10 or more years" picks the wrong pilots. Can you change the field to a number?

No. After a lake object or model object is created, a field’s type can’t change. Add a new number field, map it, and keep the old one disabled. Reread What you can’t change later.

2. A team picks the pilot's email as the primary key. What goes wrong?

Emails change. The next load sees a key it has never seen and adds a second pilot, with their history on both. Reread Two ways to get it wrong.

3. A flights file has a date, an airline, a flight number and two airports, but no ID. How do you key it?

Build a composite key with a formula field from the parts that together name one flight: the date, the airline, the flight number and the origin. Reread When no single field is unique.

4. A source sends ages as words in a few rows ("twelve"). Will the stream fail?

No. A value Data 360 can’t read as a number is stored as null. The row loads with a blank, and nobody notices unless they check. Reread Why the type column matters.

5. A person asks to be forgotten. Their details also sit in a model object that isn't related to the Individual by its ID. Are they removed?

No. The deletion request doesn’t reach unrelated model objects. Either keep personal data reachable from the Individual, or don’t bring it in. Reread Personal data has to be erasable.

6. A comments file has 1,000 comments, but counting lines gives 1,466. What's going on, and why does it matter?

Some comments have line breaks inside them. A loader that reads one line as one row will split them. Check and handle them before the stream is built. Reread Unstructured: free text and documents.

Explore

• Salesforce Certified Data 360 Consultant Exam Guide (opens in a new tab), Salesforce Help: the official list of objectives and their weights. Always check the current version before you book the exam.