Creating KPIs - The Practice of Cloud System Administration - page 392

benefit themselves. Do not be upset or surprised when people conform to what you wrote,
rather than what you intended.
Returning to our earlier security example, the easiest way to achieve the goals as stated
is to turn off any mechanisms that detect such breaches. This is probably not the intended
reaction. Thus, it should be revised.
19.2.4 Step 4: Revise and Select
Based on what we've learned in Step 3, we revise the KPI. We may select one KPI over
another or loop back to Step 1 and start over.
Seekconfirmingandnon-confirmingdata.Forexample,calculatetheKPIbasedonpast
or current metrics. If such metrics are not available, test against simulated data.
Seek the opinions of others. Ask how they would behave under such KPIs. You will get
different answers depending on whether the person has a vested interest in the KPI. If the
person's work will be judged by the KPI, filter what he or she says based on the individu-
al'spotentialbiastowardpersonalgainorpreferringtobejudgedlessharshly.Iftheperson
would benefit from the KPI's goal succeeding, he or she will have the opposite bias but
may not be in tune with the internal processes that can be used to subvert the KPI.
Asking for feedback on KPIs that have not been announced creates the potential that
rumors will spread about the new KPI. Like the children's game “Telephone,” where the
message becomes more and more misconstrued as it gets passed along, the rumor of the
new KPI will invariably portray it as harsher as it moves from person to person. This will
hurt morale and will result in misinformation that confuses people when the real KPI is
announced. The worst situation is where the draft KPI is tossed out but the rumors persist.
ThereforediscussingpotentialKPIsshouldbedonewithanunderstandingofwhotheycan
be shared with.
It is sometimes possible to test the KPI on one team or one project before applying it
to all the others. This gives us real-world experience with how people react. Alternatively,
it can be employed during a trial period so that people expect some revision before it be-
comes policy.
We don't have many chances to revise the KPI after it becomes policy. If a KPI is an-
nounced and has unintended negative side effects, you should modify the KPI to fix any
bugs. However, if the KPI changes again and again, management looks like they don't
knowwhattheyaredoing.Employeeslosetrust.Moralewillsuffergreatlyiftheruleskeep
changing. People will feel like the rugisbeing pulled outfromundertheir feet ifthey have
just finished adjusting their behavior to align with one KPI when it changes and requires
opposite adjustments.