Starting a Measurement Programme

A measurement programme is part of a means to an end (one or more business objectives). To deliver any benefit the objective(s) must be clearly understood first and then the measurement programme must be designed to support them. Once running, a programme can be tuned or changed to meet new or evolving objectives. SMS can supply all the services, consultancy and training to start a measurement programme.

Starting a measurement programme includes activities to:

• Identifying the business objective(s), their sponsors or owners and other key stakeholders
• Translating the objectives into measureable goals, developing questions which test progress towards the goals and the measures that provide numerical answers to the questions
• Defining:
  • what data needs collecting;
  • the calculations necessary to produce useful measurements from the data
  • appropriate analysis and reports to interpret results
• Discovering what measurements and raw data are already produced and what additional data collection is needed
• Planning the programme, including the data and measurements to manage and assess the programme itself
• Setting up the data collection, analysis and reporting infrastructure
• Running the programme for a number of measurement cycles to establish the processes and their correct operation. This may include some prototyping cycles used to support the above activities.

The 'start' phase of a measurement programme continues until at least one cycle of measurement, analysis and reporting has completed without significant change requests from the sponsor(s) or key stakeholders. Once this phase is complete SMS can provide any help required for Supporting a Measurement Programme.

What should be measured?

Many software metrics are standard measures that would be applied to any manufacturing or engineering business, such as labour (staff effort), time to market, number of defects.

Others are software or systems specific measures such as:

• functional size
• effort by phase, for example: design, build, testing
• effort by activity, such as: estimating, data modelling, use case analysis, system testing.

Some sort of sizing measure for the work product, like function points, is essential when derived measures like productivity, quality, speed of delivery are required for comparison across different systems, projects, releases etc. See Software Size Measures

How and by Whom?

Ideally, to be successful collecting the data should be easier than not doing so - it should be 'the path of least resistance'. In practice this is not so easy, but some important factors to consider are:

• wherever possible make data collection a by-product of an existing process, preferably some automated data extraction from an existing system
• when that is not possible make the burden on the people outside the measurement team as light as possible
• ensure that the measurement team not only undertake the bulk of the burden but are seen to be adding value.

Where there is a large volume of historical data or a need for additional resources SMS can also provide a Data Collection service to help jump start the measurement programme.