Study guide

Study guide for chapter 7 of the Data 360 Architect course. Read the six 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 7 is reshaping data inside Data 360: when you need a transform, how to build one node by node, how layers build on each other, the key every output needs, streaming transforms, and what goes beyond the builder.

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.

ING-1: What data transforms can do

• I know a transform reads data already in Data 360, reshapes it (filter, change columns, join, aggregate) and writes another object. (What a transform does)
• I can choose batch (a whole data set, on demand or a schedule) or streaming (one record in as it arrives, reshaped, out). (Batch or streaming)
• I can build a batch transform from its nodes: input, filter, transform, aggregate, output. (A Batch Transform, Node by Node)
• I know a batch transform’s formulas use API names, such as FlightDate__c, not labels. (3. Transform)
• I know an aggregate keeps only what you counted and grouped by, so a key built from other columns comes after it. (5. Transform: the keys)
• I know the rules: lake in, lake out, or model in, model out; and a batch transform’s output can’t be written by another batch transform. (The rules to know)
• I know a streaming transform is one SQL statement, active as soon as it’s saved, and can’t be changed after; it can UNION one record into several and join two streams. (Streaming Transforms)
• I know batch transforms can flatten JSON, pivot, run incrementally or in full, and write by Replace or Upsert. (Beyond the Builder)
• I know when Code Extension fits: custom Python, deployed and then run as a batch transform. (When the builder isn’t enough)

SA-4: Diagnosing and troubleshooting

• I know a transform won’t save until it has an output node. (A warning before you save)
• I know the builder’s preview is a sample, not the whole data set, so the counts while building can be lower than the output’s. (4. Aggregate)
• I know a new output object isn’t in any data space until it’s added to one, so Data Explorer doesn’t list it. (Run it and check it)
• I know a run is skipped when neither the source data nor the transform has changed. (Rule 4)
• I can check an output’s key: count the rows, look for repeated keys, run it again. (Check the key before you trust the output)

As an Architect

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

• First ask whether the data should arrive in the right shape from the source, before reshaping it in Data 360. (When the Shape Is Wrong)
• Every run costs credits. Choose how often each batch transform runs, and prefer the cheap incremental run where it applies. (Transforms cost credits)
• Define the words in the data before the filter: “arrived” is a destination, not cancelled, not diverted. (2. Filter)
• An enriched layer is an option, not a rule: an interface between source and model, at a cost. (Why an enriched layer)
• Join only to the layer above, and key each layer on a value that never changes and means the same in every org. (Rule 1, Rule 2)
• Derive the labels once, in the transform, so every segment and agent gets the same answer. (Rule 3)
• The row decides the key: for an aggregate, build it from every group-by column, with a separator, a fixed date format, as text. (Building it so it holds)
• A key qualifier holds its API name. Give a source column like key_source the same text, so the two never disagree. (When the output is a data lake object)
• Test streaming SQL before you save it: it can’t be changed afterward. (Setting one up)

Words to Know

Term	In one line
Batch data transform	Reshapes a whole data set, built from nodes, run on demand or on a schedule
Streaming data transform	One SQL statement that reshapes each record as it arrives
Aggregate node	Counts, sums or averages rows, grouped by the columns you choose
Enriched layer	Lake objects written by transforms, between the raw data and the model
Grain	What one row of an output stands for: here, one airline on one day
Composite key	A key built from two or more columns, with a separator
Incremental run	A run that processes only what was added, changed or deleted
Write mode	Replace or Upsert: how a batch run writes to its output
Code Extension	Custom Python, run as a batch transform

Check Yourself

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

1. A formula in a transform node says the field doesn't exist, though you can see it in the list. What's wrong?

The formula used the label. Batch transform formulas use API names, such as FlightDate__c. Reread 3. Transform.

2. You build the key from the airline and a formatted date before the aggregate, and it disappears from the output. Why?

An aggregate keeps only what you counted and grouped by. Build the key in a transform node after the aggregate. Reread 5. Transform: the keys.

3. An arrivals output keyed on the date alone shows far fewer rows than expected. What happened?

The key is too narrow: two airlines landing on the same day get the same key, and one overwrites the other. The key needs every group-by column. Reread Building it so it holds.

4. The transform ran, but the new object isn't in Data Explorer. What's missing?

A data space. A new output object isn’t in one until you add it. Reread Run it and check it.

5. The airport transform joins to the regions and the countries. What would you change, and why?

Join only to the layer above: the regions already carry the country. One join is one place to get wrong, and one input to wait for. Reread Rule 1.

6. A team wants to fix a typo in a saved streaming transform. How?

You can’t change it. Delete it and create another, and test the SQL first next time. Reread Setting one up.

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.