Metric drives behavior

From Pearl Language
Revision as of 19:30, 6 August 2014 by Martien (talk | contribs) (Quotes += Seth Godin » If you're not prepared to change your diet or your workouts, don't get on the scale.)
Jump to navigation Jump to search


To do: Add quotes on behavior and control from don’t just do something, stand there!

It’s better to be roughly right than precisely wrong
Maynard Keynes
Everything that can be counted does not necessarily count; everything that counts cannot necessarily be counted.
Albert Einstein
What gets measured is what gets done.
unknown
You can game any metric.
unknown
Don't measure anything unless the data helps you make a better decision or change your actions.
Seth Godin, via Seth Godin » Analytics without action
If you're not prepared to change your diet or your workouts, don't get on the scale.
Seth Godin, via Seth Godin » Analytics without action
Measure Everything That Results In Customer Satisfaction

In other words, metrics drive behavior, also known as the Hawthorne Effect and possibly related to the Observer Effect.

Therefore:

  • Secure that all metrics lead to customer delight and value creation, potentially boosting the net promoter score, either, directly or indirectly.
  • Pick and design your metrics with great care as they drive human behavior, and therefore that of the organization.
  • Use metrics to guide improvements that accelerate the organization or ecosystem as a whole, not to measure activities of people. This is, pick metrics that optimize the whole on multiple levels of scale or granularity, e.g. epic, feature, and story level.
  • Limit metrics to numbers that quantify a certain outcome or the quality of certain input that is key for the quality of an outcome.
  • Select two to three universal metrics that apply to the whole unit, division or all teams.
  • Use ockham's razor to measure only the necessary things.
  • Aim the metric to maximize value creation.
  • Capture the metric in planguage to quantify quality.
  • Keep your metric elegant, terse, to the point.


  • Metrics without goals are naked—Metrics should always give you a clue on where you are regarding your goals. So, find out where you are, how you measure that, and where you want to be. Set up a metric that tracks your progress.
  • Balanced Metrics—Many metrics only focus on operational excellence. This creates a biased view on reality and distorts and deforms the organization as a whole. Therefore, balance metrics across the following dimensions:
    1. Operational Excellence
      • Velocity
      • Burn rate
      • Predictability
      • Sustainability
      • Meeting deadlines
      • Stay within budget
      • Meet quality requirements
      • Cohesive set user stories in a sprint
    2. User Orientation
    3. Business Value
      • Business value per € development
      • Business value realization
    4. Future Orientation
      • Enthusiasm and motivation
      • happiness index
      • Educational opportunities
      • Vision on future development
      • Innovative governance
    • Have participants brainstorm on metrics and put each of them in the most appropriate category. Next, pick one or two from each catagory to create balance.
    • Consider using planguage to capture metrics in a solid, comprehensive and consistent way.

Useful Metrics

Donald Reinertsen, author of Managing the Design Factory states that a good, useful metric is:

  • Simple—The ideal metrics are self-generating in the sense that they are created without extra effort in the normal course of business.
  • Relevant—One test of relevance is whether the metrics focus on things that are actually controllable by the people being measured. Psychologists have found that when people think that they can control something they are more motivated to control it. Measuring people on things they can not control simply causes stress, dissatisfaction, and alienation. Also, metrics must be relevant towards the end goal.
  • Leading—Managers like leading indicators, and prefer an imperfect forecast of the future to a perfect report on the past. It is better to measure the size of a test queue than it is to measure the processing times of individual tests because test queue size is a leading indicator of future delays in test processing. Accountants like lagging indicators that can be measured very accurately, but these point to things that have past.
  • Self-generating—Metrics are created without extra effort in the normal course of business, as in spin-off of daily activities.

KPIs

KPI's:

  • help to give direction
  • are meaningful for everyone
  • focus on trends rather than absolute numbers
  • focus on quality (speed will follow)
  • drive improvements in the way of working
  • challenge everyone to improve performance
  • are tied to a strategic objective
  • contribute to vision and mission
  • provide neutral and objective information, not judgement
  • flow top down and bottom up
  • are concrete but not a target
  • have at least one bound
  • are leading indicators
  • gives insight in and answers the question “Are we heading in the right direction?” at strategic (company), tactical (unit) and operational (team) level;
  • are reflected in the various flow state admission criteria (e.g. Definition of Ready, Definition of Done).

Good KPIs are:

  • accessible
  • transparent—visibile to everyone
  • simple
  • understandable
  • actionable at the lowest organizational levels (team, individual)

KPIs DO NOTs:

  • Do not use KPIs to compare teams or units;
  • Do not use or design KPIs to judge;
  • Do not create KPIs that threaten or scare people;

Potential KPI trends you may want to track:

  • ka-ching moments—every time an item is put in production a point is scored; higher is better; trend should be upwards as team speeds up, gets gelled;
  • average lead time distribution:
    • average time between moment item is pulled into team and ka-ching moment; shorter is better; trend should be downwards;
    • number of outliers (should decrease)
  • throughput—running average of completed items per time period (week, month, quarter, year); as team speeds up, trend should follow upwards;
  • happiness index—drives speed improvements; higher is better; trend should be upwards;
  • due date performance:
    • For the most recent month and for the year to date;
    • Optional year-on-year (or 12 months ago);
  • flow efficiency:
    • sum of work time divided by sum of waiting time for all items
    • good indicator of the waste in the system;
  • defect rate (a.k.a. bugs):
    • Defects represent opportunity cost and affect the lead time and throughput of the system.
    • Report the number of escaped defects as a percentage against the total WIP and throughput.
    • Keeping the number of bugs between 0 and 20 is a good policy for most projects.
    • Key questions:
      • Why is the number of new defects increasing? Did you relax some QA policies?
      • How did the high level of bugs in week 20 affect cycle time?
      • What was the impact on the cumulative flow diagram when the number of bugs increased?
      • Over time, work to make the defect rate fall to close to zero.
      • less is better;
      • trend should be downwards;
  • blocked items:
    • Blocked items have serious long term effects on the systems.
    • A team’s ability to quickly solve issues says a lot about the team’s performance and effectiveness.
    • Blocked items should always be visible on the board.
    • Tracking the status over time is usually a good way of knowing whether the team is moving in the right direction.
    • less blockers is better; downward trend;
  • failure load:
    • Failure Load (amount of rework) is a good indicator that you are improving as a whole organization and thinking at a system level.
    • Failure load tracks how many work items you process because of earlier poor quality—how many work items are production defects or new features that have been requested through your customer-service organization because of poor usability or a failure to anticipate user needs properly.
    • Ideally, Failure Load should fall over time.
    • less rework is better; downward trend preferred;

The operations review is a monthly feedback loop for a number of these metrics.

Sources