3. Techniques
1. Fact tables at different levels of granularity
2. Combining correlated role-playing dimensions
3. Country-specific date dimensions
4. Dates and times in multiple time zones
9. Step 0: Requirement
The marketing department wants to analyze:
- what flights the company’s frequent flyers take
- what fare basis they pay, how often they upgrade
- how they earn and redeem their frequent flyer miles
- whether they respond to special fare promotions
- what proportion of these frequent flyers have gold, platinum, aluminum,
or titanium status
10. Step 1: Identify Business Process
*Note :
Reservation or ticketing activity
data that didn’t result in a
passenger boarding a plane is
not considered
11. THA JAP USA
TG 001 TG 001
Leg Leg
Segment
Step 2: Granularity
12. THA JAP USA
TG 001 TG 002
Segment Segment
Trip
Leg Leg
Step 2: Granularity
14. ● Segment level : each flight (same flight No.)
● 1 row in the fact table for each boarding pass collected
What can be determined?
● passenger revenue and mileage credit
Focus on..
● Marketing and Revenue groups
Step 2: Granularity
15. Multiple Fact Table Granularities
Level Use cases Note
Leg (atomic)
take off to landing
- Capacity planning analysis
- Flight scheduling analysis
number of seats, flight
duration and the number
of minutes late at
departure and arrival,
identify on-time arrivals
Segment
single flight number
- passenger revenue and mileage credit
- marketing and revenue groups
lowest level of data with
meaningful revenue
metrics
Trip
One-way trip
May change flight (transfer)
- Get an accurate picture of customer
demand
- Sales and marketing
Itinerary
Whole ticket/reservation
number
- Want to have whole idea of the travel
17. Linking Segments into Trips
Segment can’t answer one of the most important questions about your
frequent flyers: “Where are actually they going?”
Impossible to know the trip starting and end points
SOLUTION : introduce two more airport role-playing dimensions
→ 1. TripOriginAirport 2. TripDestinationAirport
18. Combining correlated role-playing dimensions
1. Class of service
2. Origin and Destination
1. Data volumes are extremely small
2. Attributes depend on combined roles for context and meaning
When to combine
19. Class Of Service Dim
Business requirement: filter report based on whether an upgrade or
downgrade occurred
Class Purchased
Key
- Need to have an upgrade indicator
Row counts are so small
Class Flown
Key
Class Change
Key
20. ( Similar as JUNK DIM )
Combine into
single dimension
Class Of Service Dim
21. Origin and Destination Airport
Data volumes are more
significant
→ Separate tables
But
Business users may need
additional attributes that
depend on the combination of
→ Origin and Destination Table
23. Option
Combine the origin and destination airport attributes, plus the
supplemental city-pair route attributes, into a single dimension
PairAir
portKey
Dpt.
Airport
Name
Dpt. Airport
City
... Arr.
Airport
Name
Arr.
Airport
City
... Directi
onal
Route
Name
Route
Distance
in Miles
Route
Distance
Band
Dom-Intl
Ind
Transocean
Ind
...
1 BOS Boston ... JFK New york ... BOS-J
FK
191 Less than
200 miles
Domestic Non-Oceanic ...
2 JFK New york ... BOS Boston ... JFK-B
OS
191 Less than
200 miles
Domestic Non-Oceanic ...
3 BOS Boston ... NRT Tokyo ... BOS-N
RT
6737 More than
6000
miles
Internatio
nal
Transatlantic ...
4 NRT Tokyo ... BOS Boston ... NRT-B
OS
6737 More than
6000
miles
Internatio
nal
Transatlantic ...
2
24. Normal Date Dimension
Consider business are across 3 countries
with different Calendars
● Need 3 extra duplicate attribute to
represents each specific Holiday.
● Redundant and become increasingly large
as number of Countries increases
Country-specific date dimensions
………………
25. What is the problems
● Different countries and multinational with different calendars
○ require sets of independent attribute of days/months/year in primary
dimension.
● Holidays occurs differently according to various calendars on different
countries or part of the world.
26. Country-specific date dimensions
JOIN to the main calendar dimension as an outrigger or directly to the fact table
Country-specific calendar
“outrigger”
27. Dates and times in multiple time zones
Multiple time zones = Multiple countries
1) Capture the date and time relative to local time
2) Express the time period relative to a standard time (GMT, UTC)
3) Separate date dimension and time-of-day dimension corresponding to
the local and equivalized dates
28.
29. Localization Recap
1. All of database-centric technique fall under the general theme
of “localization”
2. DW/BI system built to serve business users in many countries
need to designed for these localization issues
3. Challenges: How airport control towers and airplane pilots
around the world deal with language incompatibilities when
communicating critical messages → Use one language, unit
measure ex. measure in feet
33. Step 5: Fill in Attributes
Date Dim (multi-role)
Passenger Dim
Passenger Profile Dim
Airport Dim (multi-role)
Aircraft Dim
Fare Basis Dim
Booking Channel Dim
Promotion Dim
Class of Service Dim
Segment Airport Pair Dim
Time Dim (multi-role)