Budgeting Solution Built with PowerPivotAlberto FerrariBI Architect and ConsultantSQLBI.COM

SESSION CODE: BIU02

Who am I?

alberto.ferrari@sqlbi.com Independent consultant10+ years on SQL Server5+ years on BI & OLAPBook author

Expert Cube Development with Microsoft SQL Server Analysis ServicesPowerPivot for Excel 2010:Give your data a meaning

Founder of SQLBI.COM

Agenda

From the user point of view…Focus on common PowerPivot data models

Ratios and PercentagesBanding (discretization)Simulations (Courier Simulations)Non additive measures (Current account balance)Dynamic Sets

In the meantimeSome insights into the DAX programming language

Ratios and PercentagesHow to compute ratios and percentages

What is a filter context?

Each cell, in a pivot table, defines a filter context

During cell evaluation, the whole database is filtered with the filter context, rows outside it are not considered.

(Australia, 2001)

The background

A ratio need to «exit» from the filter context

Ratio = divide the «part» over the «whole»But… The filter context hides the «whole»…

(Australia, 2001) (2001)

The solution: change the filter context with ALL

We need to define a new measureBased on CALCULATE

We create e new context for a specific dimensionIn order to make it consider all the values for that dimension

Perc = SUM(FactInternetSales[SalesAmount]) / CALCULATE ( SUM (FactInternetSales[SalesAmount]), ALL (DimSalesTerritory[SalesTerritoryCountry]) )

Ratios and PercentagesExcel sheet with ratios and percentages

BandingHow to handle analysis by discretization

Analysis of product sell price

Product price changes over time, due toDiscountsPrice variations

Price is a continuous dimensionLeading to a very high fragmentationClassical solution: BANDING

From 0 to 100 USDFrom 101 to 500…

Very common pattern, even with SSAS

Banding in PowerPivot: the quick and dirty solution= IF ( FactResellerSales[DiscountedPrice] <= 5, "01 LOW", IF ( FactResellerSales[DiscountedPrice] <=30, "02 MEDIUM", IF ( FactResellerSales[DiscountedPrice] <=100, "03 MEDIUM", IF ( FactResellerSales[DiscountedPrice] <= 500, "04 HIGH", "05 VERY HIGH"))))

Even if this works… a better data modelwould be welcome!

Banding: a data driven model

SELECT P.BandName, SUM (S.ExtendedAmount)FROM dbo.FactResellerSales S JOIN PriceBands P ON S.UnitPrice BETWEEN P.FromPrice AND P.ToPriceGROUP BY P.BandName

BandName FromPrice ToPrice

VERY LOW 0 5

LOW 5 30

MEDIUM 30 100

HIGH 100 500

VERY HIGH 500 9999

Banding data model with PowerPivot

We can leverage BAND Expansion with a different data model

SELECT P.BandName, SUM (S.ExtendedAmount)FROM dbo.FactResellerSales S JOIN PriceBands P ON S.UnitPrice = P.PriceGROUP BY P.BandName

Price BandName

1 VERY LOW

2 VERY LOW

3 VERY LOW

4 VERY LOW

5 VERY LOW

6 LOW

7 LOW

… …

PowerPivot first solution: Band Expansion

PowerPivot has no BETWEEN… Thus we need Band ExpansionWe need…

A configuration sheet with band configurationScript to expand the configuration tableCreate a new linked tableConvert prices into integer valuesSetup the relationship

Very simple solution, probably the first one

Band ExpansionBand Expansion, first trial

Banding with DAX

We can leverage DAX to implement BETWEENNo need to perform table expansionUsage of FILTER to find the correct bandUsage of MAXX to convert a table into a value

BandName=MAXX ( FILTER ( PriceBands, FactResellerSales[DiscountedPrice] >= PriceBands[MinPrice] && FactResellerSales[DiscountedPrice] < PriceBands[MaxPrice] ), [PriceBand])

MAXX does not work with strings…

MAXX works only with dates and numbers, no strings!Compatibility with Excel???

Add a new numeric codeUse MAXX to retrieve the numeric codeCreate a calculated column with the codeSet relationship on the code to find the band name

Tricky… yet it works fine and does not need VB scripts

Banding with DAXFirst DAX solution

Banding with CALCULATEThe key is CALCULATEChange the filter contextWorks with any expressionLeverages CALCULATE, VALUES and FILTER together

= CALCULATE( VALUES (PriceBands[PriceBand]), FILTER ( PriceBands, FactResellerSales[DiscountedPrice] >= PriceBands[MinPrice] && FactResellerSales[DiscountedPrice] < PriceBands[MaxPrice] ))

SimulationWhat if we want to choose a new courier?

The background

We have some courier proposalsParameters are

WeightDestination

Result is «Freight»We need to choose the best oneWe carry on a simulation based on the past shipmentsTwo problems

Determine the final weight of a shipmentCreate a good simulation environment

Analysis of the dataLet us take a brief tour on the source data

Compute the shipment weight

For each shipmentSum up the detail rowsUsing a standard many to one relationshipAnd the function RELATEDTABLE

= SUMX(RELATEDTABLE(OrderDetails),OrderDetails[TotalWeight])

Freight for each courier

For each courierNew calculated columnLeverage of FILTER to find the right freightAnd of MAXX to convert a table into a scalar value(yes, it works with numbers! )

=MAXX( FILTER( 'Couriers', [Country] = 'Orders'[Country] && [MinWeight] <= 'Orders'[Weight] &&'Orders'[Weight] < [MaxWeight] && [Courier] = "SpeedyMail"), 'Couriers'[Freight])

Using CALCULATE we get a more elegant solution

Using CALCULATEIf configuration sheet contains duplicates We get an errorCleaner solution, yet more complicated to understand

=CALCULATE( VALUES( 'Couriers'[Freight] ), FILTER( 'Couriers', [Country] = 'Orders'[Country] && [MinWeight] <= 'Orders'[Weight] &&'Orders'[Weight] < [MaxWeight] && [Courier] = "Blu Express") )

SimulationFirst solution to the simulation pattern

What if we have many couriers?

One calculated column for each courierMany columnsData model needs to be updated based on dataSoon we will introduce errors

Let us create a new solutionNew data modelBased on many to many relationship

ProblemsPowerPivot does not handle many to many relationshipsLess intuitive, yet very powerful

Data model with Many to ManyBridge table

With the calculation parametersCountryWeight

Two relationshipsWith OrdersWith PriceList

ProblemsPowerPivot does not handle many to many relationshipsRelationship based on two columnsNeed to create the bridgeAnd to expand the configuration

Bridge creation and expansion

Similar to the banding solutionEasily solved with some VbScript codeNo support in DAX or PowerPivot for this

Relationship with two columns

PowerPivot handles relationships based on one column onlyEven if we live in 2010…And the relational model is 41 years old

SolutionCreate a new calculated column to hold the relationshipOn all tables.

The DAX FormulaFar from being easyUses evaluation contexts:

Filter ContextRow Context

Uses CALCULATE to transform a ROW Context into a FILTER Context

=SUMX( Orders, CALCULATE ( IF( COUNT (PriceList[Freight]) = 1, VALUES (PriceList[Freight]), BLANK () ) ) )

Simulation with many to manySolution of the simulation pattern with Many To Many relationships

Semi Additive MeasuresHow to handle semi additive measures with PowerPivot

Semi Additive Measure - DefinitionAdditive Measure

Aggregates with SUM over all dimensionsNon Additive Measure

Aggregates with different function over all dimensionsExample: average of the sale price

Semi Additive MeasureAggregates with SUM over some dimensionsAggregates with different function over other dimensionsTime is the standard exception for aggregationsExamples

Warehouse stockingCurrent account balance

Example: current account balance

Month level correctQuarter level wrongYear level wrong

Non additive measures

Change the aggregation functionDirectly inside PowerPivotWe can do it

With the user interfaceDefining a new measure

We cannot model the aggregation function inside the database

Semi additive measuresAggregation function depend on the slicerIn the example

We should aggregate for LastChild over timeBut we should use SUM for the other dimensions

Semi Additive Measures: Data ModelWe need to define a time dimension

It is not required by defaultEven if attribute consolidation comes from its adoptionIn this situation it is mandatory

New measure:CALCULATE: to set the filterLASTDATE: to find the last child

LastBalance = CALCULATE ( SUM(Balances[Balance]), LASTDATE(Date[Date]))

Semi Additive MeasuresExcel sheet implementing semi additive measures

Dynamic SetsHere we discover the underlying SSAS cube in PowerPivot.

What is a set

Handled by ExcelStatic Set

Set of tuples (i.e. filter contexts)Uses standard MDX code or User InterfaceUser interface… works, but it could have been more user friendly

Dynamic SetNeed to be written with MDXAnd doing it…You will discover the SSAS cube inside PowerPivot

Dynamic SetsExcel sheet working with static and dynamic sets

LinksPowerPivothttp://www.powerpivot.com

DAX Introductionhttp://blogs.msdn.com/powerpivot/archive/2010/04/05/dax-data-analysis-expressions-measures-in-powerpivot.aspx

Marco Russo’s Bloghttp://www.sqlblog.com/Blogs/marco_russo/default.aspx

Resources

www.microsoft.com/teched

Sessions On-Demand & Community Microsoft Certification & Training Resources

Resources for IT Professionals Resources for Developers

www.microsoft.com/learning

http://microsoft.com/technet http://microsoft.com/msdn

Learning

Related Content

BIU302 – Enriching Microsoft PowerPivot for Microsoft Excel Applications Using DAX

BIU05-INT – DAX Patterns in PowerPivot

BIU02-HOL – Defining DAX Calculations with MS PowerPivot for MS Excel 2010

Product Demo Stations

Complete an evaluation on CommNet and enter to win!

© 2010 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to

be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.