Learning Data

Topics › Master and reference data › The same one, everywhere

The same one, everywhere

A series on master and reference data. Linh graduated from one university, and now studies, tutors and does research there. She is five records in four systems, and nobody sees they are one person. Film by film, the university learns to keep the person apart from her roles, decide once who is who and which value wins, date the roles, own the lists that sort people, use each role's data only for its purpose, share one person across its systems, and keep it right.

The films

Watch them in order: each film builds on the one before.

Film 1 of 9 · Master data, and the person behind the role

One person, many roles

One person, four systems, five records. Keep the person apart from the roles, and ask four questions to know what needs mastering.

6 min · Pause and think · The same one, everywhere

Film 2 of 9 · Master or reference data, the golden record, and deciding a match

Who is it?

Two records, one person? A check proposes pairs, weights make a score, and a steward decides the hard ones once, with a reason.

5½ min · Pause and think · The same one, everywhere

Film 3 of 9 · Survivorship: a rule declared for each field

Which value wins

Five linked records, and the values still disagree. Declare a rule for each field, not one order hidden in the code.

6 min · Pause and think · The same one, everywhere

Film 4 of 9 · Roles and relationships, with dates

Roles come and go

A status is true for a day. Give every role a start, an end and an owner, and ask any question as at a date.

5½ min · Pause and think · The same one, everywhere

Film 5 of 9 · Reference data, with an owner and a date

Lists everyone uses

Three names for one role, and a country counted three times. Give each list an owner, a date and one home.

4½ min · Pause and think · The same one, everywhere

Film 6 of 9 · Purpose and consent, on each role and on the person

For this purpose only

Data given for one role isn't free for every role. Write each role's purposes beside its data, and keep the person's wishes with her.

5 min · Pause and think · The same one, everywhere

Film 7 of 9 · Master data, shared across systems

One person, many systems

Four ways to give many systems one person. Search before you create, and choose the lightest way that solves the problem.

5 min · Pause and think · The same one, everywhere

Film 8 of 9 · A master data hub, at work every day

A hub at work

Rules act, AI suggests, people decide: a week in a master data hub, and five questions to ask any hub.

4½ min · Pause and think · The same one, everywhere

Film 9 of 9 · Master data, looked after every day

Kept, not cleaned

A clean-up merged two thousand duplicates, and a year later there were eleven hundred new ones. Check at entry, measure every day, and keep it.

5½ min · Pause and think · The same one, everywhere

Take it apart, then make the call

29 hands-on labs, film by film, then 72 situations from every film. Your progress stays in this browser.

Take it apart: 29 hands-on labs →Make the call: 72 situations

One thread: one person, Linh

Linh graduated from the university in 2022 and came back to study health informatics, under a new student number. She tutors a first-year unit, and this year she joined a research project. Each system knows one part of her, by its own number. Film by film, the university gives her one record that every system can read: roles that start and end, each value labelled with where it came from, matches decided once by the people who look after them, the lists that sort people owned and dated, and each role's data used only for its purpose.

Most films open somewhere else, at another time

FilmOpens withStill true today
One person, many rolesSweden's parish books, 1686, and a number for life, 1947One person, written down once, and followed wherever they go.
Who is it?Martin Guerre's village, Artigat, 1548 to 1560A man taken for another, because so much matched.
Which value winsHMS Beagle's twenty-two chronometers, 1831 to 1836Which clock to trust is decided before the clocks disagree.
Roles come and goRome, 204 to 184 BC: Cato's ladder of offices, each held for a year, and the censusEach role has its start and its end, and the person outlives them all.
Lists everyone usesAirport codes: Toronto's YYZ, after Malton's old station code, and SGN, which still says SaigonA code outlives its reason, because changing it costs everyone who uses it.
For this purpose onlyIndia's Charaka Samhita, about two thousand years agoWhat a healer sees in a home stays in that home.
One person, many systemsThe treaty of Bern, 1874, and one postal territoryEach keeps its own, and they share the rules.
A hub at workToday, not the past: a Monday morning at the hubWhat the earlier films decided, at work every day.
Kept, not cleanedDujiangyan, in Sichuan, around 256 BC: a river split without a dam, and dug out to its mark in the dry season, year after yearWhat's kept every year lasts. What's cleaned out once decays.

Who knows the person

Cyan outlines are technical; gold outlines are business.

The person

Linh Nguyen, a master's student, alumna, tutor and researcher, held as five records in four systems

Draws the model

Noor, the university's data architect

Looks after the person

Mei, whose office decides who someone is, and whether two records are one

Builds the model

Jun, the analytics engineer, who turns what's decided into code, with tests