When viewing scorecards in PerformancePoint you will notice not only an actual and target value for the KPI but also a status indicator. The status indicator is determined by the KPI score, shown below.

clip_image002

The KPI score is a normalized score based on the KPI settings . This score allows the KPI to be properly compared with other KPIs and aggregated into an objective. In the screenshot above, we have four KPIs that roll up to the Increase Revenue objective: Sales Amt, Sales Amt - % Growth PP, Unit Sales and Unit Sales - % Growth PP. The Increase Revenue target status is determined by the scores of the four KPIs comprising this objective.

Several steps are involved when calculating the KPI score. First, we identify the Band by Value, which is the value that we will ultimately map into the threshold ranges of the KPI to determine a score. Second, we calculate the correct normalized value of the Band by Value so that we can effectively roll up the scores across multiple KPIs into an objective score that makes sense. Outlined below are the steps for calculating the normalized KPI score.

Step 1: Calculate the Band by Value (BBV)

The BBV is calculated differently depending on the banding method and scoring pattern chosen.

Scoring Pattern: Increasing is Better, Decreasing is Better, Closer to Target is Better

Banding Method: Band by Numeric Value of Actual

BBV = value of the actual

Note: Our algorithms assume a ratio is provided for the actual value. Although it works with non-ratios, there may be some corner cases where it will fail.

Scoring Pattern: Increasing is Better, Decreasing is Better, Closer to Target is Better

Banding Method: Band by Stated Score

BBV = query result

*Query result is the result of the MDX expression entered during threshold setup where banding method and scoring pattern are chosen

Note: Our algorithms assume a ratio is provided for the actual value. Although it works with non-ratios, there may be some corner cases where it will fail.

Scoring Pattern: Increasing is Better, Closer to Target is Better

Banding Method: Band by Normalized Value of Actual to Target

BBV = (Actual Value - Worst)/(Target Value - Worst)

*Worst value is the value entered in the Threshold wizard

clip_image004

Scoring Pattern: Decreasing is Better

Banding Method: Band by Normalized Value of Actual to Target

BBV = 1 - ((Actual Value - Worst)/(Target Value - Worst) )

*Worst value is the value entered in the Threshold wizard

Step 2: Identify the In-Band Value

The In-Band Value (IBV) is a 1-based value that determines in which threshold range the BBV lies. Range 1 is always the bottom range in the threshold settings; range 2 is the range immediately above the bottom range, and so on. If you are using a three-state indicator there are three ranges, a four-state indicator has four ranges. The following screenshot shows both the BBV and IBV for the three scoring patterns:

Click on the below image to see the entire set of values.

clip_image006

Step 3: Calculate the Normalized Band by Value

The Normalized Band by Value (NBV) is the normalized score shown in the scorecard. This value is calculated to equalize the KPI scores so they can be accurately aggregated for objective KPIs. Consider the following two KPIs:

clip_image008

clip_image010

The BBV value for the first KPI is .6 (with a worst value of 50, this is (80-50)/(100-50)). The BBV value for the second KPI is 4. Because these KPIs have a different scale and threshold, we cannot simply combine them together to achieve a rational aggregate value. Instead, we must calculate normalized values that can be proportionally mapped to the same range. For a three-state indicator, this range is 100/67/33/0; for a four-state indicator, this range is 100/75/50/25/0.

In the first example above, we map the KPI's threshold ranges of 120/80/40/0 to the standard scale of 100/67/33/0 and determine a normalized (NBV) value to represent the raw (BBV) value on this standard scale. In the second example, we map the KPI's threshold ranges of 10/5/2/0 to the standard scale of 100/67/33/0 and determine a normalized (NBV) value to represent the raw (BBV) value on this standard scale.

Using the simple example shown in the first KPI above, we convert .6 to 60%. Sixty percent is at the exact midpoint of its threshold range (120-0), so it's mapped to 50% on the standard range (100-0). In the sections below, we will provide the mathematical equations to illustrate this concept in greater detail.

Step 3a: NBV for Increasing is Better

NBV is calculated using the following formula:

NBV = ((Upper - Lower) * (d/T)) + Lower

With the following values:

Upper = In-Band/# of Bands

Lower = (In-Band-1)/# of Bands

d= (BBV - lower band threshold)

T= (upper band threshold - lower band threshold)

Using the following KPI settings:

clip_image011

Scoring Pattern: Increasing is Better

Banding Method: Band by normalized value of Actual/Target

Worst Value: 50

And the following values:

BBV = (80-50)/(100-50) = .6

In-Band = 2

Upper = 2/3 = .67

Lower = 1/3 = .33

d = (.6 - .4) = .2

T = (.8 - .4) = .4

The calculation is as follows:

NBV = ((.67-.33)*(.2/.4)) + .33 = .50

And the score is shown in the scorecard below:

clip_image013

Let's walk through another example with a KPI banded by actual:

With the following settings:

clip_image010[1]

Scoring Pattern: Increasing is Better

Banding Method: Band by numeric value of Actual

Worst Value: 0

Using the following values:

BBV = 4

In-Band = 2

Upper = 2/3 = .67

Lower = 1/3 = .33

d = (4-2) = 2

T = (5 - 2) = 3

The calculation is as follows:

NBV = ((.67-.33)*(2/3)) + .33 = .556

And the score is shown in the scorecard below:

clip_image015

Step 3b: Calculate the Normalized Band by Value for Closer to Target is Better

The NBV calculation for closer to target KPIs is identical to the NBV for Increasing is Better KPIs, with one exception. Once we have the NBV from the above Increasing is Better calculation, we make the following adjustment:

If In-Band < (# of Bands/2) then

NBV = NBV * 2

Otherwise

NBV = 2 * (1 - NBV)

Let's look at an example for this calculation.

With the following settings:

clip_image017

Scoring Pattern: Closer to Target is Better

Banding Method: Band by numeric value of Actual

Using the following values:

BBV = -1

In-Band = 2

Upper = 2/6 = .33

Lower = 1/6 = .17

d = (-1-(-2)) = 1

T = (0 - (-2)) = 2

The calculation is as follows:

NBV = ((.33-.17)*(1/2)) + .17 = .25

In-Band < (6/2) so we multiply NBV by 2 to yield .50

And the score is shown in the scorecard below:

clip_image019

Let's walk through another example where the actual is 6. Using the following values:

BBV = 6

In-Band=6

Upper = 6/6 = 1

Lower = 5/6 = .8333

d = (6-5) = 1

T=(10-5)=5

The calculation is as follows:

NBV = ((1-.8333)*(1/5)) + .83 = .866

In-Band >= 3 so we calculate NBV as:

NBV = 2(1-.866) = .2668

And the score is shown in the scorecard below:

clip_image021

Step 3c: Calculate the Normalized Band by Value for Decreasing is Better

Although the NBV calculation for Decreasing is Better KPIs is identical to the NBV calculation for Increasing is Better KPIs, the BBV is calculated differently.

With the following settings:

clip_image023

Scoring Pattern: Decreasing is Better

Banding Method: Band by numeric value of Actual

Worst Value: 200 (as entered in threshold settings)

Using the following values (note the BBV calculation):

BBV = 1 - ((80-200)/0-200)) = .4

In-Band = 3

Upper = 3/3 = 1

Lower = 2/3 = .667

d = (.4 - .4) = 0

T = (0 - .4) = -.4

The calculation is as follows:

NBV = ((1-.667)*(0/-.4)) + .667 = .667

And the score is shown in the scorecard below:

clip_image025

Let's walk through another example where the actual is 100. With the following settings:

Scoring Pattern: Decreasing is Better

Banding Method: Band by numeric value of Actual

Worst Value: 200 (as entered in threshold settings)

Using the following values (note the BBV calculation):

BBV = 1 - ((100-200)/0-200)) = .5

In-Band = 2

Upper = 2/3 = .667

Lower = 1/3 = .333

d = (.5- .8) = -.3

T = (.4 - .8) = -.4

The calculation is as follows:

NBV = ((.667-.333)*(-.3/-.4)) + .333 = .583

And the score is shown in the scorecard below:

clip_image027

Scoring Roll-Up to Objectives

Once each KPI has the score determined the objective score is then calculated. The objective score is a simple average calculation of the KPIs underneath the objective (unless weighting has been applied to one or more of the KPIs, in which case the weighting is factored into the average). The first screenshot in this post shows an objective KPI with the average score.

Look for future blog topics on how the threshold settings and worst values can affect your scores.

Alyson Powell Erwin (alysonp@microsoft.com)