Inventory Accumulating Snapshot

Report
INVENTORY CASE STUDY
Introduction
 Optimized inventory levels in stores can have a
major impact on chain profitability:
 minimize out-of-stocks
 reduce overall inventory carrying costs
Value chain example
Value chain
What is the primary objective of most analytic
decision support systems ?
 monitor the performance results of key
business processes
each business process produces unique metrics at
unique time intervals with unique granularity and
dimensionality
 each process typically spawns one or more fact
tables
 value chain provides high-level insight into the
overall enterprisedata warehouse
Some Common Questions related to Inventory
 How did the inventory level changed per product, per warehouse over time?
 How is the profitability of products in our inventory?
 How many times have we placed a product into an inventory bin on the
same day we picked the product from the same bin at a different time?
 How many separate shipments did we receive from a given vendor, and
when did we get them?
 On which products have we had more than one round of inspection failures
that caused return of the product to the vendor?
 … etc.
 BI helps answering these questions
BI Inventory Models
 The three main models discussed:
 Inventory Periodic Snapshot
 Inventory Transactions
 Inventory Accumulating Snapshot
They are complementary models, and provide
different information about the Inventory
Periodic Snapshot
The most common inventory scheme
Example of Retail Store Chain Inventory:
The assumed atomic level of
detail is:
Inventory per product
Per day
Per Store
Basic dimensions:
Product
Day
Store
Fact:
Inventory
Simple Inventory Periodic Snapshot
Usage:
Provide information about inventory levels:
1. Daily Inventory level
2. Average Inventory level over a time period
Problems:
1. Inventory levels are semi-additive (i.e. NOT additive through each dimension)
 Through the Date dimension the quantity on hand is NOT additive
2. Historical Inventory data using daily granularity results in unreasonably huge
amount of data over time
 Suggestion to define distinct atomic time period for short and long term
measures
Enhanced Inventory Periodic Snapshot
Extra recorded facts
Velocity of inventory movement becomes measurable
Key concepts:
 Number of Turns
 Number of days’ supply
 Growth Margin Return on Inventory (GMROI)
Enhanced Inventory Periodic Snapshot
Extra recorded facts
measure
Number of
Turns
Number of
days’ supply
GMROI
daily
Over a period
quantity sold
total quantity sold
quantity on hand
dayily average quantity on hand
quantity on hand
final quantity on hand
quantity sold
average quantity sold
total quantity sold x (value at latest selling price - value at cos t)quantity on hand
daily average quantity on hand x value at the latest selling price
Enhanced Inventory Periodic Snapshot
GMROI - Growth Margin Return on Inventory
total quantity sold x (value at latest selling price - value at cos t)quantity on hand
daily average quantity on hand x value at the latest selling price
Turns
Gross margin
High GMROI
lots of turns
high gross margin
Low GMROI
low turns
low gross margin


GMROI is a standard metric used by inventory analysts to judge a
company’s quality of investment in its inventory.
We do not store GMROI in the fact table because it is not
additive!!!
Inventory Transactions
Record every transaction
that affects inventory:
Receive product
Place product into inspection hold
Release product from inspection hold
Return product to vendor due to inspection failure
Place product in bin
Authorize product for sale
Pick product from bin
Package product for shipment
Ship product to customer
Receive product from customer
Return product to inventory from customer return
Remove product from inventory
Inventory Transactions
Use: Measure the frequency and timing of specific transaction types
Example:
• How many times have we placed a product into an inventory bin on the same day we
picked the product from the same bin at a different time?
• How many separate shipments did we receive from a given vendor, and when did we
get them?
• On which products have we had more than one round of inspection failures that
caused return of the product to the vendor?
Inventory Accumulating Snapshot
In progress!!!
 In a single fact table row we track the disposition of the
product shipment until it has left the warehouse
 only possible if we can reliably distinguish products
delivered in one shipment from those delivered at a
later time
 also appropriate if we are tracking disposition at very
detailed levels, such as by product serial number or lot
number
Inventory Accumulating Snapshot
Fact Table Type Comparison
Periodic Snapshot
Transaction
Accumulating Snapshot
Time period
represented
Regular predictable
intervals
Point in time
Indeterminate time span, typically
short lived
Grain
One row per period
One row per
transaction event
One row per life
Table loads
Insert
Insert
Insert and update
Row updates
Not revisited
Not revisited
Revisited whenever activity
Date
dimension
End-of-period
Transaction date
Multiple dates for standard
milestones
Facts
Performance for
predefined time
interval
Transaction activity
Performance over finite time
Value Chain Integration
 Integrating business processes together benefits:
 Intelligence aspects:


Better understand customer relationships from an end-to-end
perspective
Observe information across business processes
 Technological aspects:
 Reusability
 Less resources used
 Question: How do we properly integrate all business
processes in the enterprise?
 Answer: Data Warehouse Architecture
Data Warehouse Bus Architecture
 Bus:
 “Common structure to which everything can and is
connected”
 Data Warehouse Bus Architecture:
 Defining a standard warehouse architecture (bus
interface) to which different data marts can connect.
 Standardizes dimensions and facts that have uniform
interpretation across the enterprise.
 Architectural framework for the overall design and
separate data marts following the framework.
Data Warehouse Architecture
Kimball vs. Inmon
 Bill Inmon and Ralph Kimball – the co-founders of the
data warehouse concept and their views on data
warehouse architecture
 Dependent Data Mart Structure (Inmon)
 Let everyone build what and when they want and we will
integrate it if we need it.
 Each data mart gets information from the operational data base
and then data is loaded in the data warehouse
 Data Warehouse Bus Structure (Kimball)
 Design everything then build.
 The data warehouse is responsible for loading data in the data
marts from the operational database.
Bus Matrix
 The tool we use to document the Data Warehouse Bus
Architecture
 A part technical, part management, part
communication tool
 Business processes as ROWS
 Common dimensions as Columns
Bus Matrix (cont.)
 Rows :
 Business processes


A business process translates into a First-Level Data Mart
Each Data Mart spanning over multiple business processes translates
into a Consolidated Data Mart (E.G. Profitability)
 Columns:
 Common Dimension used across the enterprise
 Consequences of improper or non-existent bus matrix:
 Isolated data marts blocking the coherent warehouse
environment, narrowing down the scope of information to be
viewed.
 Expansion of the data warehouse is nearly impossible
Conformed Dimensions
 What are conformed dimensions:
 The cornerstone of the Bus Architecture
 A single, coherent view of data across the enterprise that
can be reused across different Data Marts.
 Conformed dimensions have:
 Consistent dimension keys
 Consistent attribute values
 Consistent naming, attribute definitions.
Conformed dimensions (cont.)
 Some characteristics of conformed dimensions
 Each conformed dimension has the same meaning in
each Data Mart
 They are defined at the most granular level possible
Conformed dimensions (cont.)
 Some considerations when defining conformed
dimensions
 Rolled-up dimensions


Rolled-up dimensions – having higher level of granularity
Rolled-up dimensions conform to the base-level atomic
dimension if they are a strict subset of that dimension
Conformed dimensions (cont.) Considerations (cont.)
 Dimension subsetting
 Two dimensions with same level of detail
but representing different subsets of rows
or columns
 Rolled-up dimensions are another
example of dimension subsetting
 Advised Solution – dimension authority
 Has responsibility for defining,
maintaining and publishing dimensions
and their subsets to all Data Marts
Conformed Facts
 Conformed facts are:
 Facts used living in more that one data mart.
 Same rules and characteristics apply in designing and
implementing them as with conformed dimensions
 Few more considerations are:
 Units of measure for the fact
 Identical labeling
 Underlying definitions and equations

similar documents