Measures Architecture 2009
The surface view of a measure is its recent value -- "the number" -- but a measure has many other components that define its meaning and management significance. IS&T measures are supported by these meta-data fields:
field |
purpose |
example |
Name |
What we tend to call it in conversation |
Help Desk Cases Resolved in more than 6 days |
Value |
the latest recorded measurement, for the last month, say, or quarter |
30 |
Units |
Percent, cases, hours, seconds, etc. |
% |
Target |
What is the current "goal" value for this measure? Setting this target is a key task for the relevant management team. (In the same units as Value.) |
15 |
Gap |
The difference between Value and Target. (In the same units as Value and Target.) |
15 |
Trend |
Left to continue as it has been, what is the value likely to be in N time periods |
26 |
Is Up Good |
a Yes/No view of what managers consider the desirable direction. Puts "Trend" into context. |
No |
Threshold |
What is the size of the Gap or the level of the Value that should set off warnings in the minds of management. In this case, it determines whether Alarm, below, to Yes or No. |
25 |
Alarm |
a Yes/No indicator of whether the Value has exceeded the Threshold. This variable would be used in spreadsheets to determine how to color cells in a spreadsheet dashboard or otherwise highlight out-of-range variables. |
Yes |
Response |
Written-out consideration of what is means to management when values start exceeding thresholds. What would management do (WWMD)? |
"Study the underlying issues within this subset of cases -- root cause analysis to help discover the business process breakdown that is driving it." |
Team |
What part of IS&T "owns" this data |
Helpdesk |
LOB |
Description of the business process the measure depicts |
Call Center ticket resolution |
Strategic Goal |
Description of the intended outcome for this measure when the LOB process is working as well as it can, why do we want the measure target to be what it is. |
|
Source |
What people or system contains the raw data for extracting to derive the variable. |
Casetracker logs |
Contact |
Who to talk to about missing data, weird data, changes to processes reflected in the variable, etc. |
Goguen or ccf@mit.edu |
Average |
Time series average of all the data points |
27 |
Median |
Time series median value for all the data points |
25 |
Can Use XmR |
Yes/No setting for whether statistical process control analysis seems appropriate. |
Yes |
Moving Range |
time series mR value (the range of change in time series data points |
8 |
Process Limit |
a statistical interpretation of the moving range, roughly equal to three standard deviations of the moving range. |
12 |
NUPL |
Natural Upper Process Limit of the time series. Assuming the process is currently in control, this is the natural maximum for the variable. It is defined as the variable's average plus the Process Limit. The relationship between Target and NUPL can inform the need for possible process changes to bring behavior in line with target. The NUPL shows how the process _is_ behaving, not how management would specify that it _should_ behave. |
37 |
NLPL |
Natural Lower Process Limit of the time series. Assuming the process is currently in control, here's the natural minimum for the variable. |
13 |
In a spreadsheet of this information, the metadata would be coupled with columns of periodic measurements, for each of at least 15 months, say. Additional columns would automatically compute quarterly averages and quarterly aggregates, as well as all the XmR parameters. Finally a graphical presentation would accompany each line of values, to allow for quick visual assessments of the variable's behavior over time.
Our current mockup along these lines covers public-facing variables at http://web.mit.edu.ezproxyberklee.flo.org/ist/about/measures