Measuring QSM's repository and analysis tool
- By James Heires
Slim-Metrics, a new software data collection, analysis and reporting tool from
QSM Inc., McLean, Va., is one of three software metrics tools offered by QSM. While Slim-Metrics is designed to
reveal an organization's ability to develop software, it also allows any organization that is not yet collecting
historical software project data to begin to do so immediately. QSM's other two products, Slim (a project estimation
tool) and Slim-Control (a project tracking tool), can use Slim-Metrics' historical data as a "sanity check"
to seamlessly integrate the measurement phases of a software project.
Software development ability is primarily measured in terms of size, effort, schedule and quality. Initial estimates
and final actuals can also be compared to help improve estimation ability. But an organization's ability to stay
competitive in its
industry is an important facet of a measurement program.
Slim-Metrics is integrated with annually updated industry-trend data representing more than 3,000 worldwide
software development projects covering all major industries. Although individual project data is not available,
these trend lines are helpful when benchmarking an organization against the competition. This capability, like
all of Slim-Metrics' features, is accessed by a no-nonsense user interface.
Slim-Metrics comprises two distinct programs: Data Entry and Analysis. The Data Entry program allows users to
view and edit the contents of an existing database or enter new data. Existing databases can be read directly from
any of the following file formats: Slim-Metrics (.smp), Slim estimate (.sle), Slim history (.slh), Pads database
(.pd4) and Slim-Control (.scp). The database is also ODBC-compliant, making it easy to create seamless interfaces
to other ODBC-compliant tools.
QSM has included a sample project database with Slim-Metrics that allows users to appraise the product's functionality
without first collecting data. Users can select a project from a database or add a new project to reveal a multipage
data entry screen (see Fig. 1). The data fields on the first page of this screen constitute the Software Engineering
Institute's Initial Core Measures of size, effort, schedule and defects. Users can enter new data from this screen
or view existing data; the remaining pages allow more detailed information to be entered.
|Fig. 1 Slim-Metrics lets users appraise functionality without collecting data. A multipage data
entry screen can reveal a project from a database.
The remaining portion of Data Entry consists of more than 50 additional measures, including cost, requirements
size, application type, design complexity, environmental factors, overrun/slippage, growth/reduction and custom
fields. Custom fields -- such as compliance with certain standards, or the use of certain tools or methods -- are
quite useful for defining measures particular to your organization or industry. But users should take note: Because
each field is a potential discriminator when it comes to analysis, custom fields should be selected carefully.
|Fig. 2 Data points, containing detailed information about the project represented, may be graphed
according to property sheets.
Although extremely flexible, Data Entry does not allow new metrics to be formulated using combinations of other
measures. For example, to define a metric such as hogsheads/fortnight, consider a spreadsheet-like formula capability.
Assume for a moment that Beer Consumed is a custom field (quantifying the amount of beer consumed, measured in
hogsheads, by a programming staff over the life of a project). Hogsheads/fortnight could then be computed in another
custom field by the following formula: Hogsheads/(Schedule/12 months per year * 26 fortnights per year), where
Schedule is measured in calendar months. Selected projects could then be analyzed with this newly created metric.
Buttons at the bottom of the Data Entry screen let users navigate through projects in a database. Projects can
also be added and deleted in the same manner. After data entry is completed, users should switch to the Analysis
program in order to experience the analytical power of Slim-Metrics.
MINING FOR DATA
The Analysis program contains a statistics engine that encompasses min., max., mean, standard deviation, coefficient
of determination, slope, intercept and a number of observations. Histograms, scatter plots and bar charts can also
be generated to help users discover the inherent trends in their data. The standard measures and metrics on each
graph can also be compared with industry-trend lines for competitive positioning studies.
Individual studies can be grouped together to form a view that contains up to 16 graphs or tables in an adaptable
layout (see Fig. 2). Views can be designed around the metrics dashboard approach or around business goals (productivity,
quality and time-to-market). Views with multiple graphs or tables can be "zoomed" so that a single item
fills the screen.
Slim-Metrics' graphs, however, are far more than just static pictures. Graphs also allow instant access to data,
with each data point containing detailed information about the project represented. Project name, organization,
size, effort and other project detail can be obtained instantly about any data point on any graph. Each graph also
has an associated property sheet that defines axis metrics, titles, data sets, variation lines and report format.
One power feature is the ability to present different data trends on the same graph. This capability can be
helpful when users need to show the differences between organizations, application types or time periods. Trend
lines for each data set can be displayed as well, showing mean and standard deviation.
Slim-Metrics requires Windows 95 or NT, a Pentium CPU and 30Mb of hard disk. License options begin at $4,995.
Training/support is also available from QSM.
James Heires is a 12-year veteran of the software industry, primarily at a Fortune 100 aviation electronics company located in the Midwest. His current professional interest lies in software quality improvement based on the analysis of software project data.Rational Rose 98.