I want to calculate realization ratio of a value related to a target when lower values are considered better, in one special case the goal is zero. To make my problem more clear, there is a complete use case here: http://jhealthscope.com/en/articles/80342.html Where author wants to calculate the ratio in table 2: http://jhealthscope.com/en/articles/80342.html#A80342TBL2 Author have solved the same question by dividing the target by the performance where lower is better, e.g. "The job complications of personnel" index. But I have doubt about this approach and want to find the best approach and a complete answer.


If the actual value ($a$) is never less than the goal ($g\ge 0$), you could look at $100 \frac{a-g}{a}$. For a goal of 10 and an actual value of 22 (difference of 12), this would evaluate to about 54.55%. For a goal of 0 and an actual value of 12 (the same difference), it would evaluate to 100%. The figure represents how much you would need to improve, as a fraction of the current value, to reach your goal.

UPDATE: To avoid having the ratio default to 100% when $g=0$, regardless of the value of $a$, you can adjust the denominator as follows. Make the deviation ratio $\frac{a - g}{\max\{a,\epsilon\}}$ for some $\epsilon >0$. If the $g=0$, the deviation is 1 (and realization is 0) until $a$ gets small enough that $a<\epsilon$, at which point deviation is $\frac{a}{\epsilon}$ and realization is $\frac{\epsilon - a}{\epsilon}$.

  • $\begingroup$ I think when kpi meets the target, we have to get 100% for the realization ratio and 0 deviation, and your answer is closer to the deviation than the realization ratio. $\endgroup$ Oct 2 '19 at 22:25
  • $\begingroup$ And with this approach, I cant calculate the ratio when actual value is zero. $\endgroup$ Oct 3 '19 at 7:34
  • $\begingroup$ Your question said "lower is better" and mentioned goal zero. If lower is better and outcomes cannot be negative, actual = 0 would imply goal = 0. As I said in the answer, we would just define this as 100% (which is the limit of the formula I gave as $a\rightarrow g$ when $g=0$. There is one catch: using my formula, any $a$ results in 100% when $g=0$. So the problem is not $a=0$, but you may not be happy with the formula when $g=0$. $\endgroup$
    – prubin
    Oct 3 '19 at 18:11
  • $\begingroup$ OK, if we use $(a-g) /a$ to calculate deviation then with $1-deviation=realization$ will get $realization=g/a$ that is the same formula we are using already. $\endgroup$ Oct 3 '19 at 18:45
  • 1
    $\begingroup$ You could use $\frac{a-g}{\max\{a,\epsilon\}}$ for some arbitrary $\epsilon >0$ (maybe $\epsilon = 1$). If $g=0$ and $a$ gets close to 0, it will depart from the previous formula. I've seen variations of this in other contexts. $\endgroup$
    – prubin
    Oct 3 '19 at 20:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.