learn-data-governance-with-phoebe / Session 1 of 8
Learn Data Governance with Phoebe · Session 1 of 8

Data governance and the DPO

What data governance actually is, why Singapore law makes the Data Protection Officer role mandatory, and where you sit when you're the first DPO a company ever hired.

🟢 Start here DA · DE · DS · AI DPO track Singapore + EU 45 min live + self-study depth
0-3 · Welcome 3-23 · Concepts (live cards) 23-42 · Build-along on Himalaya 42-45 · Q&A
Part 0

Why this course exists

Your data team can build pipelines, models, and dashboards. What most data teams cannot do yet is answer a regulator's simplest question: "What personal data do you hold, why, and who said you could?" This 8-session course turns a data practitioner into someone who can run as a Data Protection Officer - protect the people in the data, and protect the company from the fine. We learn it by governing one company, Himalaya, from its messy first day to a board-ready program.

Live - presented in session Self-study - read after class ★ Do it now on Himalaya PDPA-primary · GDPR mirror
★ What you walk out with today A precise definition of data governance you can defend to an engineer and a board member, the DAMA map of everything governance spans, a clear read on why the DPO role is legally mandatory in Singapore (and only sometimes in the EU), and a first draft of Himalaya's data landscape with your seat on it. Cards marked "Live" are what we do together; "Self-study" cards give the full depth at your own pace.
Part 1 · covers DAMA-DMBOK2 - the governance function

What data governance actually is 7 min live

Governance is not a tool, a team, or a folder of policies nobody reads. It is the exercise of authority and decision-making over data: who gets to decide, how decisions are enforced, and who is accountable when it goes wrong. Everything else - quality, security, catalogs - is a capability governance directs.

1 Decision rights Who is allowed to decide what happens to a dataset - named, not assumed. 2 Enforcement Policies and standards with teeth - controls, reviews, escalation when broken. 3 Accountability Someone answers for the data - to the business, the regulator, and the person. No named decider, no enforcement, no one accountable = no governance, however many tools you bought.
🔍 Click to zoom - the three things every governance program must establish
LiveGovernance vs management vs "the data team"3 min

People blur three different things. Data management is the doing - building pipelines, storing, modeling, serving. Data governance is the deciding - setting the rules, rights, and accountability the doing must follow. The data team is a set of people who mostly do management and, until someone insists otherwise, govern by accident.

The tell that a company has no real governance: ask "who decided we could keep this?" and you get a shrug, a Slack thread, or "that's just how the pipeline was built." Governance replaces the shrug with a name, a policy, and a paper trail.

Why this matters to you

  • As a DPO you will spend far more time on decision rights and accountability than on technology. The hard part is organisational, not technical.
  • Your credibility comes from being able to say who is accountable for each dataset and on what basis it is held - not from owning a fancy catalog.
Real world

Himalaya, day one. Raj (Head of Platform) says "we keep everything in Redshift, it's all governed - we have access controls." That's management, not governance. Nobody has decided what should be kept, for how long, or why. Your first job is to turn that "it's handled" into a real, defensible answer.

LiveThe DAMA wheel - governance at the hub2 min

The industry's standard map of data management is the DAMA-DMBOK2 wheel: eleven knowledge areas. Ten are capabilities - the work. The eleventh, Data Governance, sits at the centre because it directs and connects all the others. You don't need to master all ten to be a DPO, but you must know they exist and how governance steers each one.

Data Governance the hub Architecture Modeling & Design Storage & Operations Data Security Integration & Interop Document & Content Reference & Master Data Warehousing & BI Metadata Data Quality Teal boxes = the four spokes a DPO touches most: Security, Metadata, Quality, and Architecture.
🔍 Click to zoom - the DAMA-DMBOK2 wheel, governance at the centre (source: dama.org)
Say it right Governance is not a twelfth capability bolted on. It is the hub that decides how the other ten behave. When someone proposes "a governance tool," ask which spoke it actually serves - the answer is usually Metadata (a catalog) or Quality.
Self-studyThe ten spokes, in one line each2 min read
Knowledge areaWhat it covers · why a DPO cares
ArchitectureThe blueprint of how data flows. Your data maps (Session 3) live here.
Modeling & DesignHow data is structured. Where personal fields get defined - or hidden.
Storage & OperationsDatabases and their running. Where retention and disposal actually happen.
Data SecurityAccess, encryption, controls. Your Protection Obligation leans on this (Session 5).
Integration & InteroperabilityMoving data between systems - every move is a potential transfer risk.
Document & ContentUnstructured data (PDFs, emails). Personal data hides here too.
Reference & Master DataThe single source of truth for "who is this customer" - key for access requests.
Warehousing & BIAnalytics. Where personal data quietly gets copied for reporting.
MetadataData about data - the catalog, glossary, lineage. Your visibility layer (Session 3).
Data QualityAccuracy, completeness. Backs your Accuracy Obligation (Session 6).
Part 2 · covers PDPA Accountability Obligation + GDPR Arts. 37-39

The DPO role, and why it's the law 7 min live

Here is the fact that surprises most data people: in Singapore, appointing a Data Protection Officer is not optional and not size-dependent. Every organisation must designate at least one DPO. In the EU it's the opposite - mandatory only in three defined situations. Knowing both, and which applies to your company, is DPO literacy step one.

PDPA (Singapore) - your home law Every organisation must appoint a DPO No exemption for size, sector, or data volume. DPO business contact must be made public. Part of the Accountability Obligation. GDPR (EU) - the mirror Mandatory only in 3 cases (Art. 37) 1 · Public authority / body 2 · Large-scale regular & systematic monitoring 3 · Large-scale special-category / criminal data Himalaya is Singapore-based with EU consumers - so it needs a DPO under PDPA AND likely qualifies under GDPR case 2.
🔍 Click to zoom - two rulebooks, two different DPO triggers
LiveWhat the DPO actually does2 min

GDPR Article 39 gives the cleanest task list, and it maps well onto the PDPA role. A DPO's job is five verbs:

  • Advise - tell the business and its staff what the law requires, before they act, not after.
  • Monitor - check compliance with the law and internal policies; run awareness, training, and audits.
  • Assess - advise on and track data protection impact assessments (DPIAs) for risky processing (Session 8).
  • Cooperate - be the working contact for the regulator (PDPC in Singapore).
  • Point - be the named contact both regulators and data subjects can reach.
Real world

When Ilsa (CRO) plans a new 8M-consumer SMS campaign, the DPO doesn't run the campaign or block it on instinct. The DPO advises ("here's the consent and DNC check you need first"), monitors that it happened, and is the person the PDPC calls if a complaint lands. Advisor and watchdog, not gatekeeper.

LiveIndependence - the part companies get wrong2 min

The DPO role only works if it has teeth, so the law protects its position. Under GDPR Article 38 - and as good practice everywhere - the DPO:

  • Reports to the highest level of management - not buried three layers down under a manager whose project you might have to flag.
  • Acts independently - no one instructs the DPO on how to do the role, and the DPO cannot be dismissed or penalised for doing it.
  • Has no conflict of interest - you can't be DPO and also the person who owns the marketing database whose lawfulness you're meant to judge.
  • Is properly resourced - given time, budget, and access to carry the role, not handed it as a title on top of a full-time job.
The independence test If the DPO cannot safely say "no, not like that" to the CEO, the role is decorative. When you take a DPO seat, negotiate the reporting line and the "cannot be penalised" protection on day one - that is the job's foundation, not a perk.
Self-studyThe DPO and the certifications1 min read

This course teaches the DPO body of knowledge. If you later want the paper credentials, the recognised ones are from the IAPP:

  • CIPP/E - knowing the privacy law (GDPR and European data protection).
  • CIPM - operationalising a privacy programme (the management side of the DPO job).
  • CIPT - engineering privacy into products and systems (privacy by design).

IAPP positions CIPP/E + CIPM together as the standard training combination for the DPO role. In Singapore, the PDPC also runs its own practitioner courses. The exams and certificates stay with those official bodies - this course gets you fluent in what they test.

Part 3 · covers PDPA + GDPR - what counts as personal data

Personal data, PII, and the sensitive tier 6 min live

You cannot protect a category you can't define. Data people say "PII" loosely; the law is more precise, and one tier - sensitive or "prescribed" data - carries far heavier duties. Getting this ladder right is what makes your data map (Session 3) actually mean something.

Non-personal data Aggregate metrics, anonymised counts - no duties attach. Personal data Any data about an identifiable person: name, email, phone, address, purchase history, an ID. Sensitive / prescribed data - highest duty NRIC-style national IDs, financial account info, health data, login credentials, biometrics. ← wider category, lighter duties narrower, heavier duties → Himalaya's wellness-vertical clients collect health notes - that pushes Himalaya into the top tier.
🔍 Click to zoom - the classification ladder that drives every duty you owe
LivePersonal data - broader than you think2 min

Under both PDPA and GDPR, personal data is any data about an identifiable individual - on its own or combined with other data the organisation has or is likely to have. That "combined with" clause is the trap: an anonymous-looking table plus one join key becomes personal data.

  • PII (personally identifiable information) is the common US-flavoured term for the obvious identifiers - it is roughly a subset of "personal data," which is the legal term you should use.
  • A device ID, an IP address, or an in-app behaviour trail can be personal data when it can be tied back to a person - which, in a product like Himalaya's, it usually can.
  • Anonymisation (truly irreversible) takes data out of scope; pseudonymisation (reversible with a key) does not - it's still personal data.
Real world

Raj proudly shows you an "anonymised" analytics table - no names, just a consumer_id. But there's a lookup table one schema over that maps consumer_id to email. Combined, it's fully personal data, and it's been feeding a US analytics vendor for a year. That's not anonymised; that's pseudonymised, and it's in scope.

LiveThe sensitive tier - why it changes everything2 min

A subset of personal data is treated as far riskier because misuse causes real harm. GDPR calls it special category data (Art. 9); Singapore's breach rules turn on prescribed personal data. Either way, the practical effect is the same: heavier consent, stronger protection, and a lower bar for when a breach becomes notifiable.

  • Typical members: health data, financial account details, national IDs (NRIC), login credentials, biometrics, and - under GDPR - race, religion, sexual orientation, and more.
  • A breach involving this tier is much more likely to cross Singapore's "significant harm" threshold and trigger mandatory notification (Session 5).
  • Consent expectations rise: you generally need clearer, more specific consent to process it.
DPO instinct The first thing you find in any new company is where the sensitive tier lives - because that's where the biggest fines and the worst headlines come from. Map that before you map anything else.
Self-studyData subjects, controllers, and processors2 min read

Three roles you'll use in every session. Getting them right decides who owes which duty - critical in a B2B2C platform like Himalaya.

RoleMeaning · Himalaya example
Data subjectThe person the data is about. Himalaya's 8M end consumers, and named contacts at the 1,200 business clients.
ControllerDecides why and how data is processed. Often the business client - but Himalaya is a controller for its own platform data.
ProcessorProcesses on a controller's instructions. Himalaya is frequently a processor for its clients; the US analytics vendor is a processor for Himalaya.

The controller/processor split matters because duties and liability follow the role. Part of your Session 3 mapping work is deciding, dataset by dataset, whether Himalaya is acting as controller or processor - because it's both, in different places.

Build-along · everyone builds

Map Himalaya's data landscape and place your seat ★ 19 min · on Himalaya

Your first act as Himalaya's DPO. We don't fix anything yet - we get oriented. By the end you'll have a one-page data landscape, a first-pass sensitivity read, and a defensible answer to "where does the DPO sit?" Follow along on Himalaya; then repeat the exact steps on your own organisation for homework.

Your company

Himalaya - B2B2C SaaS, Singapore-headquartered. 1,200 business clients (the B), 8M end consumers (the C). Runs on AWS Redshift (40+ schemas) plus a CRM, a US-hosted marketing automation tool, and a support tool. Wellness-vertical clients collect health notes. Growth-first, governance-last: a near-miss export exposed consumer records for 6 days, and marketing blasts the 8M base with no DNC check. The board just made you its first DPO.

List the data stores. Write every place personal data could live: Redshift schemas, CRM, marketing tool, support tool, product modules. Don't verify yet - capture the landscape.

Tag the B-side vs C-side. Mark which stores hold business-client data (the B) and which hold end-consumer data (the C). The C-side is where your 8M data subjects live.

Flag the sensitive tier. Circle anything likely to hold health notes, NRIC/national IDs, financial account info, or credentials. For Himalaya, the wellness-client data and payment tokens get circled first.

Mark the cross-border hop. Draw the arrow from EU consumer data → Singapore warehouse → US marketing vendor. This one line is your Session 6 transfer problem; today, just make it visible.

Place the DPO seat. Decide your reporting line (to the CEO / board, per independence), and write one sentence on the conflict you must avoid (you cannot also own the marketing database). This is your position, on paper, day one.

★ Do it now - the landscape promptYou are helping me, the newly appointed Data Protection Officer, get oriented at a company. Company: [paste the Himalaya summary above, or describe your own org in 4-5 lines]. Produce a one-page data landscape: 1. A table of data stores: store name | B-side or C-side | example personal data | likely sensitive tier? (Y/N) | controller or processor role 2. A short list of the 3 highest-risk spots you'd investigate first, and why 3. One sentence on where the DPO should report, and the main conflict of interest to avoid Use ONLY what I gave you. Where you'd need to confirm something, list it as an open question instead of guessing.
Data tip Never paste real personal data into a practice chat. Describe the shape of your data ("a table of consumer emails and purchase history"), not the data itself. That habit - the Protection Obligation applied to your own tools - is the DPO mindset from minute one.
Homework

Try it yourself - this week ◐ 30-45 min total

Source material

What this session covers

This session teaches the working content of the frameworks below. Certification exams, member-only frameworks, and legal advice stay with their official sources - this page makes you fluent in the body of knowledge, honestly flagged where depth lives elsewhere.

DAMA-DMBOK2 - the governance function & the wheelPart 1 · governance at the hub + all 10 spokes named; full depth in the book
PDPA - Accountability & the mandatory DPOPart 2 · every SG org must appoint a DPO (pdpc.gov.sg)
GDPR Arts. 37-39 - DPO role, tasks, independencePart 2 · the 3 mandatory cases + Art. 39 task list (gdpr-info.eu)
Personal data, sensitive/prescribed, the three rolesPart 3 · PDPA + GDPR definitions
IAPP CIPP/E · CIPM · CIPTPart 2 · named and mapped; exams stay official (iapp.org)
Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · A Singapore company has 4 employees and holds a little customer data. Does it need a DPO?

PDPA's DPO mandate has no size exemption. The "large-scale monitoring" test is GDPR's - a different rulebook.

2 · Where does Data Governance sit on the DAMA wheel?

Governance is the centre. It's the deciding function that steers the ten management capabilities around it.

3 · An "anonymised" table uses a consumer_id, and another table maps that id to an email. The table is...

Reversible with a key = pseudonymised = still personal data. Only truly irreversible anonymisation leaves scope.

Session 1 cheat sheet · pin this

Governance = deciding, not doingDecision rights + enforcement + accountability. Management does; governance decides.
DAMA wheelGovernance at the hub, 10 capability spokes around it. DPO touches Security, Metadata, Quality, Architecture most.
DPO mandatePDPA: every SG org, no exemption, contact public. GDPR: only 3 cases (Art. 37).
DPO's five verbsAdvise · Monitor · Assess (DPIAs) · Cooperate with the regulator · be the Point of contact.
IndependenceReports to the top, can't be penalised, no conflict of interest, properly resourced.
Data ladderNon-personal → personal → sensitive/prescribed (health, NRIC, financial, credentials). Heavier duties up top.
Pseudonymised ≠ anonymisedReversible with a key = still personal data. Only irreversible anonymisation leaves scope.
Three rolesData subject · controller (decides) · processor (acts on instructions). Duties follow the role.